APBRmetrics

Author Message
Mountain



Joined: 13 Mar 2007
Posts: 1527


PostPosted: Thu Nov 27, 2008 6:06 pm Post subject: Non-Inside Shot Team Effective FG% and more Reply with quote
Anybody calculate this before? I hadn't, at least that I can recall at the moment. But it seems a necessary stat to know. It is not enough to know the parts, need to know the product of them.

ATL 49%
NYK 48%
SAS 48%
MIA 48%
PHO 47%
NJN 47%
NOH 47%
TOR 46%
POR 46%
CLE 45%
LAL 45%
SAC 44%
ORL 43%
UTA 43%
IND 43%
DAL 43%
BOS 43%
MIL 43%
DET 42%
CHI 42%
CHA 42%
PHI 41%
MEM 41%
HOU 41%
LAC 40%
MIN 40%
DEN 39%
WAS 39%
GSW 38%
OKC 36%

Correlation of Non-Inside Shot Team Effective FG% and Wins as of a few days ago: +.51.

Correlation of 2pt Jumpers % of Shots - 3ptr % of Shots: -.44.

Pretty telling.

Top11 all currently playoff seeded then other factors play an important role. Denver the lowest playoff seed on this list followed by Houston.

% of inside shots + % sent to line correlation with winning is +.55. Surprisingly % of inside shots * Inside Shot FG% correlation with winning is only +.14.

How strongly does Non-Inside Shot Team Effective FG% correlate with road or playoff wins? Heightened? Anybody want to check these?



Somewhat related Houston aside:
Yao lowest shots per minute since rookie season. Artest taking far too many but actually lowest for him since his second season. If he doesn't care how many of his shots fall if the team wins is he willing to give up more of those shots for the same reason in a contract year? Debatable whether McGrady is taking too many given OK TS% but the rate is lowest since his 3rd season. One too many moderate volume shooters?

Landry usage down 6% from last season despite having the best offensive rating of any player in the league? Accurately understand him or pigeon hole him as a role player but there are different levels of role players and why not push it?

http://www.basketball-reference.com/lea ... aders.html

Every Rocket (in top 9 rotation with scores) interior player positive on early adjusted +/-. Every perimeter player negative lead by Brooks despite his individual offensive game.


Somewhat related Thunder aside:

Delonte West can't fit the PJ / Presti culture or plan but gets himself right and gets the right role with Cavs and is burning it with the 3rd best offensive rating in the league with the 3 ponter falling . Nah he doesn't fit the now Thunder design with the 3 ponter dramatically curtailed. Give him away as a trade throw in after paying dearly in the original trade with some hope for him I thought. Draft expert Presti chooses instead now to build with Westbrook. Of course it might work with time but right now 18th worst offensive efficiency of 264 players who play 15+ minutes. No real good choice between Watson (3rd worst on off. eff.) and Ridnour (44th). Except again there was the West option if you could have figured him and the situation out, if that was possible. Might have helped to try harder in a non PJ way, Green 42nd worst, Durant 59th. Well at least under Presti's tank or just not right team of PJ/Westhead.

Stuckey 32nd worst, so other GMs are somewhere on the making "investments" or making mistake spectrum.


A lot of the better teams have 5 or more guys who take 2+ 3 point attempts or aren't far off that design. The way to maximize attempts and open good attempts probably.
Celtics at 3 but one is House and he ain't hitting. They may really miss Posey in playoffs on this aspect despite not generally. I'd guess Ainge & Co. do something about this before then.

Last edited by Mountain on Sun Dec 14, 2008 2:53 am; edited 1 time in total
Back to top
View user's profile Send private message
Mountain



Joined: 13 Mar 2007
Posts: 1527


PostPosted: Mon Dec 01, 2008 4:41 am Post subject: Reply with quote
I have a feeling that the inefficiency of the mid-range game has long been masked by the offsetting boost of the inside shot FG% when folks looked at overall 2 pt shooting vs 3 point. Now 82 games provides the 3 part breakout.

This roll-up I think gives a different perspective on the perimeter game as a whole and who is doing it well or not. A low mark may not be fatal but it is a significant drag. Improving a low mark here given the high number of perimeter shots per game could produce nice point gain. And reducing mid=range shots probably would produce more easier than trying to invent the good percentage midrange shot that few find regularly.
Back to top
View user's profile Send private message
NickS



Joined: 30 Dec 2004
Posts: 384


PostPosted: Mon Dec 01, 2008 12:29 pm Post subject: Re: Non-Inside Shot Team Effective FG% Reply with quote
This is a really interesting bit of data. It would be interesting to combine this with the percentage of non-inside shots that are three pointers. Team's that shoot a lot of three-pointers (and shoot them at all well) will look good on this statistic, but is there any additional benefit to an offense from having the ability to shoot 2-point jumpshots well?

One of my theories, for which I haven't yet developed a statisitical model, is the way in which the contest between the offense and defense determines what shot the offensive team takes.

The offensive team's ideal outcomes for a possession would generally be, in order of decreasing preference:

Inside Shot
Draw Foul
Open 3 point shot
Open 2 point shot
Contested 3 point jumper
Contested 2 point jumper
Turnover.

The defense's preferences would be in referese order.

I feel like there is frequently a tension in building a team between having a team who has the mindset that they will force the defense to let them take the shot that they want (for example, PHI last season who was able to have a decent offense without having good shooters by managing to get a lot of inside shots) versus having a team who is able to be efficient while taking the shots that the defense allows them (Detroit last season could feel a little bit like that -- they were willing to take jump shots but, at least, would make sure they didn't turn the ball over and would take open jump shots).

This is just a personal theory, but seeing this post made me think again about ways to map out statistically which teams were better at getting the more desired shot opportunities, and which were good at taking advantage of less desirable shot oppportunties (like 2-point jump shots).

Mountain wrote:
Somewhat related Thunder aside: . . .


I have to admit, all of your criticism of the Thunder looks justified right now. They have been playing better since the coaching change and, offensively, Durant has looked better at SF, but they look like a team that is a long ways away from having a good mix of pieces.
Back to top
View user's profile Send private message
Harold Almonte



Joined: 04 Aug 2006
Posts: 616


PostPosted: Mon Dec 01, 2008 2:03 pm Post subject: Reply with quote
Quote:
Team's that shoot a lot of three-pointers (and shoot them at all well) will look good on this statistic....

The quid is to track the contested shots separately from open ones. The two teams that are allowing more three pointers actually are looking very good on this respectively defensive statistic.
Back to top
View user's profile Send private message
Mountain



Joined: 13 Mar 2007
Posts: 1527


PostPosted: Tue Dec 02, 2008 3:49 am Post subject: Reply with quote
Thanks Nick.

In Mike's 08-09 player projections thread I slid this in mostly out of place before creating this thread...

"Ran correlations for this early season data for team % 3pt shots. FG%, pts and more, likewise for 2ptrs, inside shots and % fouled against actual win %, expected win%, pace and offensive efficiency. The strongest correlations:

Over .5
3pt % of shots and offensive efficiency,
pts from 3 ptrs and offensive efficiency,
inside FG% and offensive efficiency,
inside points and pace

over .45
actual wins and 3pt FG%
3pt FG% and offensive efficiency,
actual wins and pts from 3ptrs
actual wins and 2pt FG% (corrected below)
pace and inside shot % of shots

The most negative correlation was between 2pt % of total shots and pace at -.38 followed very close by 2pt % of total shots and actual and expected wins. "


I actually made a mistake on one item relevant to your question. Actual wins and 2pt FG% is not better than +.45 correlated. That correlation was actually registered for 2pt shot assisted rate and wins (an argument against or at least not in favor of "shot creation"?) The correlation of 2 pt FG% and actual and expected wins is just +.17-,18. Not much case for importance.


Preference order:

"Inside Shot
Draw Foul
Open 3 point shot
Open 2 point shot
Contested 3 point jumper
Contested 2 point jumper
Turnover. "

I'd have to see the contested 3 pointer vs open 2 pt shot data before I'd check off on that priority. It might be pretty close to even and would probably depend on 2 pt distance and degree of 3 pt contest of course. Probably some open 2 ptrs would be better but not ready to concede it in general. Need to see the data as Harold recommends. Ed Peterson charted contested shots but not by 2 vs 3. Tall 3 pointers can be contested but still get off pretty clean looks with their height and elevation so that type of 3 point shooter might have extra value (Dirk, R Lewis).


The correlations also suggest that good inside shots are helpful but just more inside shots have no impact or negative impact. But may be more the flipside of the strong correlation of those who use the 3 pointer well.
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3615
Location: Hendersonville, NC

PostPosted: Tue Dec 02, 2008 5:22 am Post subject: Reply with quote
Intuitively, I guess an inside FGA is more likely to include a foul (good or bad) or a turnover (bad); while a 3-pt attempt is least likely on both; and an outside 2 is in the middle frequencies.

And these vary by team mechanics, as well as by individual. Oh, and you're more likely to be injured attacking the basket. Lots more to consider than just FG/FGA.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Mountain



Joined: 13 Mar 2007
Posts: 1527


PostPosted: Tue Dec 02, 2008 11:19 am Post subject: Reply with quote
That is basically true http://www.82games.com/locations.htm
but 2 pointers draw only a couple % more fouls than 3s (except corner 3s which I guess are quick hitting, hard to foul but should be defender tougher and perhaps fouled more often is can do it before the shot) and this is offset by the higher turnovers. The main effect is that the 3 pointer raw FG% raises the pts per possession return dramatically (first table). In a game of inches this is a huge difference. I'd take as many 3s as I could (while also maxing good inside shots and especially foul shots) that would beat the 2 point jumper paltry returns (or least the lower half of those shots on quality and who takes them) which I'd think would be almost any decent look a decent 3 pt shooter could get (over 27-30% raw FG% or 40-45% eFG%).

Jerry Sloan lauds the importance of the 15 footer http://blogs.sltrib.com/jazz/2008/12/mo ... -range.htm
and it probably does play an important in defining the title winner over the contenders... but it is still the worst shot in the game (see the FG% in the above link for zones 7 & 9)
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3615
Location: Hendersonville, NC

PostPosted: Tue Dec 02, 2008 2:31 pm Post subject: Reply with quote
Mountain wrote:
That is basically true http://www.82games.com/locations.htm
but 2 pointers draw only a couple % more fouls than 3s...

Eh? According to that link -
Quote:
... 90% of the shooting fouls do occur in the paint, but not with a similarly lopsided turnover rate,..

-- while 7% occur between the 'paint' and the arc; just 1% occur on FGA beyond the arc. Unless I'm reading it wrong.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Mountain



Joined: 13 Mar 2007
Posts: 1527


PostPosted: Wed Dec 03, 2008 2:58 am Post subject: Reply with quote
I was using table 1 and looking at total fouls.

You are using table 2 and shooting fouls.

Both useful but different.

Last edited by Mountain on Wed Dec 03, 2008 8:35 pm; edited 2 times in total
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3615
Location: Hendersonville, NC

PostPosted: Wed Dec 03, 2008 10:09 am Post subject: Reply with quote
Well, if we are talking about which shots draw fouls, then they would seem to be shooting fouls. The fouls that may occur in various 'zones' on the court are not all the result of shot attempts.

It seems a scoring attempt between the paint and the arc is about 4 times as likely to draw a foul, as is one from beyond the arc. Within the paint, about 6 times likelier still: some 26x as likely as shooting the 3.

I'm startled to see the very low TO rate -- 8% -- on possessions deep in the paint. Maybe that doesn't include TO incurred while attempting to get it down there? This line isn't convincing to me:
Quote:
Turnovers are infrequent deep in the paint (since there's not a lot of dribbling)..


The corner 3 is highly effective once a player has the ball there. One reason that player tends to be open is that he's somewhat removed from the rest of the action: he's not in position to set a pick, get an OReb, interfere with the opponent's fast break, etc.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Mountain



Joined: 13 Mar 2007
Posts: 1527


PostPosted: Wed Dec 03, 2008 8:37 pm Post subject: Reply with quote
Table 2 as for total shots not per shot like table 1 but I get your points. Shooting fouls are more relevant but prorate them to per shot and the difference between 2 and 3 pointer shooting foul data is not huge.
Back to top
View user's profile Send private message
Mountain



Joined: 13 Mar 2007
Posts: 1527


PostPosted: Sun Dec 07, 2008 1:52 am Post subject: Reply with quote
If strong midrange 2 point jumpshooting is quite important for winning a championship then PHO CLE DAL LAL SAS
ATL UTA NOH TOR ORL have that right now. POR
DET DEN HOU among those who don't, with Houston last.

Last season Boston was 5th best, Lakers 9th.

Cavs were 30th last season, 3rd this season so far. How much it is because of the offensive system overhaul and how much is Mo Williams and his impact would take more work though clearly he is a chunk of it. How much is random hotness, time will tell. As good as strong midrange shooting is they still score about 17% points more for each 3 pointer launched over the midrange jumper (putting fouls and turnovers aside).



Turns out that the correlation of midrange FG% and overall FG% is less than .3.

The average midrange FG% of the top 15 on overall FG% is 40%, just 1 percentage point higher than league average. The % of mid-range shots has far more impact on overall FG%.

I am not familiar with the coaching literature on the mid-range shot and its connection to everything in reality beyond a few anecdotes . If any are and wish to comment that would be useful and appreciated.
Back to top
View user's profile Send private message
Crow



Joined: 20 Jan 2009
Posts: 821


PostPosted: Sun Jul 12, 2009 7:06 pm Post subject: Reply with quote
In addition to looking at Non-Inside Shot Team Effective FG% it would also be useful to look at largely a flip of this, namely Non=3point TS%- that is the efficiency of all possessions used that weren't a 3 point attempt.

(I know that this does address fouls on attempted 3 point launches but I don't have that. I guess it could be generically adjusted or maybe it could from the right play by play database.)

This isn't a convoluted, esoteric stat; this is the essentially efficiency of doing anything other than shoot the 3. You might intend to get a shot inside or get fouled or shoot a mid-ranger (or even pass) but what you end up getting might be different and be one of these other things.

Even though this roll-up can explain the actions of players crossing the 3 point line and semi-excuse mid-range shots as part of that mix that can look ok to very good on overall efficiency for this subset of possessions, for almost all or all players they'd be wiser instead of shooting that mid-ranger (or at least most of them) to continue to strive to get one of the other more desirable offensive opportunities outcome or- if there is enough time- to pass and try to get an average opportunity elsewhere- inside or from 3- from somebody else.


For all players who played more than 500 minutes here are the top third on non3ptA TS%:

Player Team PS Scoring Index
dampier,erick dal C 0.664
przybilla,joel por C 0.654
jones,solomon atl C 0.653
ginobili,manu san SG 0.651
hilario,nene den C 0.647
singleton,james dal PF 0.639
landry,carl hou PF 0.635
allen,ray bos SG 0.632
o'neal,shaq pho C 0.624
ming,yao hou C 0.617
gasol,pau lal PF 0.617
andersen,c den C 0.617
stoudemire,a pho C 0.617
calderon,jose tor PG 0.616
barry,brent hou SG 0.612
james,lebron cle SF 0.610
maggette,corey gsw SF 0.608
johnson,amir det PF 0.608
paul,chris nor PG 0.607
wallace,gerald cha SF 0.605
howard,dwight orl C 0.601
collison,nick okl PF 0.601
miller,mike min SG 0.601
williams,deron uta PG 0.600
oberto,fabricio san PF 0.600
oden,greg por C 0.599
nash,steve pho PG 0.599
smith,craig min PF 0.599
bynum,andrew lal C 0.598
gordon,eric lac SG 0.598
barbosa,leandro pho SG 0.598
noah,joakim chi C 0.595
hill,grant pho SF 0.595
martin,kev sac SG 0.593
perkins,k bos C 0.593
moore,mikki bos PF 0.593
lee,david nyk PF 0.591
gasol,marc mem C 0.591
powe,leon bos PF 0.591
wade,dwyane mia SG 0.589
harris,devin njn PG 0.589
kleiza,linas den SF 0.589
armstrong,h nor C 0.589
balkman,renaldo den SF 0.587
biedrins,andris gsw C 0.586
bogut,andrew mil C 0.586
lowry,kyle hou PG 0.586
moon,jamario mia SF 0.586
jordan,deandre lac C 0.585
posey,james nor SF 0.583
iguodala,andre phi SF 0.583
terry,jason dal PG 0.583
maxiell,jason det PF 0.583
billups,c den PG 0.583
brewer,ronnie uta SG 0.582
chandler,tyson nor C 0.581
pierce,paul bos SF 0.581
nelson,jameer orl PG 0.581
okafor,emeka cha PF 0.581
robinson,nate nyk PG 0.579
williams,marvin atl SF 0.579
millsap,paul uta PF 0.579
wright,brandan gsw PF 0.578
bass,brandon dal PF 0.577
bosh,chris tor PF 0.576
gortat,marcin orl C 0.575
pachulia,zaza atl C 0.575
roy,brandon por SG 0.574
granger,danny ind SF 0.574
okur,mehmet uta C 0.574
scola,luis hou PF 0.574
fernandez,rudy por PG 0.573
redd,michael mil SG 0.573
songaila,darius was PF 0.571
lopez,brook njn C 0.570
lopez,robin pho C 0.570
ariza,trevor lal SF 0.570
murphy,troy ind PF 0.569
korver,kyle uta SF 0.569
butler,caron was SF 0.569
bryant,kobe lal SG 0.569
weaver,kyle okl SG 0.569
durant,kevin okl SF 0.568
lewis,rashard orl PF 0.567
varejao,a cle C 0.567
nowitzki,dirk dal PF 0.567
salmons,john chi ?? 0.566
horford,al atl C 0.566
kirilenko,a uta SF 0.566
garnett,kevin bos PF 0.565
turiaf,ronny gsw C 0.563
bonner,matt san PF 0.563
miller,brad chi C 0.562
novak,steve lac PF 0.562
smith,j.r. den SG 0.562
jack,jarrett ind SG 0.562
parker,tony san PG 0.561
richardson,j pho SG 0.561
warrick,hakim mem PF 0.561
ratliff,theo phi C 0.560
szczerbiak,w cle SF 0.560
allen,tony bos SG 0.559
camby,marcus lac C 0.558
duhon,chris nyk PG 0.557
battier,shane hou SF 0.556
young,thaddeus phi SF 0.556
barnes,matt pho PF 0.556
redick,j.j. orl SG 0.556
watson,c.j. gsw PG 0.556
gordon,ben chi SG 0.556
batum,nicolas por SF 0.556
jamison,antawn was PF 0.555
williams,mo cle PG 0.554
miles,c.j. uta SG 0.554
dudley,jared pho SF 0.554
miller,andre phi PG 0.553
haslem,udonis mia PF 0.553
kaman,chris lac C 0.552
diaw,boris cha PF 0.552
odom,lamar lal PF 0.551
augustin,d.j. cha PG 0.551
duncan,tim san C 0.550
graham,joey tor SF 0.549
harrington,al nyk PF 0.548
hickson,j.j. cle PF 0.548
crawford,jamal gsw ?? 0.548
rondo,rajon bos PG 0.548
love,kevin min PF 0.547
chalmers,mario mia PG 0.546
brown,kwame det C 0.546
speights,marrees phi PF 0.545
thompson,jason sac PF 0.545
turkoglu,hedo orl SF 0.544
jefferson,richar mil SF 0.544
foster,jeff ind C 0.544
west,david nor PF 0.543
murray,ronald atl PG 0.543
bogans,keith mil SG 0.542
cardinal,brian min PF 0.541
douglas-roberts njn SG 0.541
smith,josh atl PF 0.541
elson,francisco mil C 0.541
koufos,kosta uta PF 0.540
carney,rodney min SG 0.540
bargnani,andrea tor PF 0.540


These are the guys who deliver the most directly when doing anything other than launch a 3.


This is part of why I like Landry, Singleton, Powe and others.

Maggette does well with this and should go inside the 3 point line ahigher percentage of the time or exclusively.

Eric Gordon is looking good. Kleiza is not one-dimensional- at least in terms of offensive efficiency. Lowry far better than Brooks on this.
Artest in Houston gave / was allowed to post a score on this in bottom 10% of qualifiers. Also in the bottom 10%- McGrady. Sure wasn't efficient to have those doing anything other than shoot the 3.

Yi was in bottom 10% too. A few others who were almost: Blake and the short-lived expression of Brand in Philly. A few more in the bottom third: jump=shooting Boozer, K Martin, G Davis, Krstic and A Randolph.



You can compute this for teams too. The top 4 playoff teams finished 3rd, 5th, 7th and 8th on this. An average rank of 4.6 (a 2.5 average would be the best possible). No other partial stat I've discussed in this thread (inside shot percentage of total shots or mid-range accuracy or % fouled correlates with playoff finish as well as this composite stat. Maybe not surprising but worth saying as part of the reason for explicitly looking at this stat- and 3 point performance and of course the defense of both. A new "4 stat" line to check.

By contrast the top 4 playoff teams finished with an average rank of 15.5 on non-inside eFG%. The mid-range game, attempts and efficiency affects both while getting to the line and making it payoff affects non 3 ptA TS% has both but not non-inside eFG%. Doing well on both sets obviously is desirable. Some well perform well on both, some just one and getting to the line and making it payoff is part of the difference maker but alongside something else in case. Among the final 4 all but Denver were good from mid-range. Maybe it is still important though least efficient.

Knowing how players and teams do on non-inside eFG% and non 3 ptA TS% seems pretty important to me. And pretty basic to me.


The long twisting names won't help the cause so I'll suggest calling them NISI (Non-Inside Shooting Index) and NOSI (Non-Outside Scoring Index) if you want to be precise or just Shooting Index and Scoring Index for simplicity. When people talk of shooters and scorer I think this is largely what they mean. And you can look at how different teams field these types and produce the results they produce on the indexes as a team and in the win column, regular season and in the playoffs. Probably plenty to find with these simple tools.
Back to top
View user's profile Send private message
Crow



Joined: 20 Jan 2009
Posts: 821


PostPosted: Fri Jul 17, 2009 6:17 am Post subject: Reply with quote
Looking at scoring index there really seems to be a cutoff at .500. Score at below that rate (for non-outside shots) and with very few exceptions you don't play much across the league. That is pretty sensible.

By contrast on shooting index there are 25 big minute players with a non-inside FG% below .4 and another 45 below .45. That is something to manage around and try to minimize.

I don't have the shooting index data immediately for all players but if you looked at performance on shooting and scoring index for starting fives for good teams you could get a sense of the mix of above average performances they have on these measures and you could pretty easily see the differences between one starting lineup mix of these skills and the other teams.

Just looking quickly at top usage guys Dallas, Detroit, Atlanta, Golden State, Toronto and Washington had the greatest depth of not bad outside shooters with 4-5. That group is a mixed bag and lacks any of the very top teams. The best teams seem to do alright with 3. Of course with the full data you could analyze the average or good shooter levels more fully and understand the differences better.

And if you had the shooting and scoring indexes for all players you could check which lineup mixes did better on adjusted +/- (or just offense if that becomes available again) on average minutes weighted by team and across the league. And see how the types look by this criteria overall and how adjusted varies by player within these types by performance on these metrics. Perhaps there are varying positional patterns expected or not

Using these 2 measures you could pretty easily summarize who to play tight on the perimeter and who to make extra efforts of one kind or another to prevent from going inside. I imagine the scouting reports get this right most of the time but there might be some differences between what the metrics say and what the scouting reports say for some players.
Back to top
View user's profile Send private message
Crow



Joined: 20 Jan 2009
Posts: 821


PostPosted: Fri Jul 17, 2009 7:10 pm Post subject: Reply with quote
This post deserves a mention

http://www.3hoopsfans.com/category/fun- ... rs/page/3/

Kevin Pelton
Site Admin


Joined: 30 Dec 2004
Posts: 684
Location: Seattle

PostPosted: Fri Nov 04, 2005 2:11 am Post subject: Protrade on the NBA Draft Reply with quote
I can't seem to find a link an hour or two later, but Protrade -- the pseudo-fantasy/stock market site that incorporates statistical analysis with those two things -- had an article from Ben Alamar analyzing the 2005 NBA Draft.

Ben's been good at chatting about his methods on the comment threads at FootballOutsiders.com, and I was wondering if one of our connected folks (Roland or DeanO) could invite him here; I'd love to discuss what seems like a unique methodology for analyzing the draft.
Back to top
View user's profile Send private message Send e-mail Visit poster's website AIM Address
moneyp



Joined: 24 Feb 2005
Posts: 69


PostPosted: Fri Nov 04, 2005 12:13 pm Post subject: Reply with quote
http://www.protrade.com/insight/Insight ... Draft.html

Interesting site.
Back to top
View user's profile Send private message
gkrndija



Joined: 20 Feb 2005
Posts: 64


PostPosted: Fri Nov 04, 2005 1:01 pm Post subject: Reply with quote
That was a really bad article. He completely ignored eFG% when analyzing players and he completely ignored pace. Arguably, the 2 biggest fundamentals of apbrmetrics.

What's worse is that he seems to have read about things like skill curves and still manages to miss the point. Scoring Efficiency does not drop according to minutes played according to the book.

He also completely dismisses Salim Stoudemire from any NBA success because of his poor asst/TO ratio(0.91). That seems a little short sighted considering Reggie Miller and Alan Houston also had less than stellar ratios coming out of college.

I didn't read past the Atlanta preview.
Back to top
View user's profile Send private message
benalamar



Joined: 08 Nov 2005
Posts: 3


PostPosted: Tue Nov 08, 2005 2:23 pm Post subject: Reply with quote
Thought I'd justify my exclusion of Stoudimire a little bit.

The players mentioned about Houston and Miller, both had significantly better asst/to ratios (1.1 for Houston and 1.14 for Miller) than Stoudimire (0.91). In the 20 years of NBA draft data I have, no guard came out of college and had NBA success (by the measure used in the article) with that low a ratio. It just didn't happen.

If, however, he is included in the analysis, he would have ranked 43rd out of 43 players (due in large part to his low ratio).
Back to top
View user's profile Send private message
gkrndija



Joined: 20 Feb 2005
Posts: 64


PostPosted: Tue Nov 08, 2005 4:30 pm Post subject: Reply with quote
benalamar wrote:
Thought I'd justify my exclusion of Stoudimire a little bit.

The players mentioned about Houston and Miller, both had significantly better asst/to ratios (1.1 for Houston and 1.14 for Miller) than Stoudimire (0.91). In the 20 years of NBA draft data I have, no guard came out of college and had NBA success (by the measure used in the article) with that low a ratio. It just didn't happen.

If, however, he is included in the analysis, he would have ranked 43rd out of 43 players (due in large part to his low ratio).


Welcome to the board Ben. Your analysis is too harsh on Salim. If you are going by college stats to predict future NBA success, there is no way Stoudamire should be dead last. He has an eFG% of 65% Shocked It dwarfs your #1 guy Sean May(57%). He also has a TS% of around 68% which is insane. He doesn't even really turn the ball over that much. He doesn't get many assists. But why pass if you're that unstoppable and lead one of the best offences in the country?

Your putting too much stock on A/TO from one year. Salim was (1.41) as a sophmore. Larry Hughes is a pretty successful player and he had a (0.63) when he entered the draft. Heck, Michael Jordan left school with (0.96) and was (0.74) the year before.
Back to top
View user's profile Send private message
benalamar



Joined: 08 Nov 2005
Posts: 3


PostPosted: Tue Nov 08, 2005 4:53 pm Post subject: Reply with quote
I don't think that the ranking is that off base. If you look at PER, Stoudimire ranked 39th among players that had declared for the draft and his PER dropped against top quartile competition. Now the model does have a bias against seniors so that may drag him down the rankings, but even without the bias, he does not look good.

I thought about relying just on one season of data for the model, and tried to go back and use multiple seasons, but found almost no correlation to previous seasons with future success, once the last year in college was included. Stoudimire could end up being a statistical oddity, and he does have a small probability of actually becoming successful, but the odds are against him.
Back to top
View user's profile Send private message
gkrndija



Joined: 20 Feb 2005
Posts: 64


PostPosted: Tue Nov 08, 2005 10:15 pm Post subject: Reply with quote
PER is a cool stat, however it's not perfect. They're great at getting a general idea of who are the best players from the boxscores, but it is not the holy grail of rankings.

For Example Drew Gooden(25th) and Memo Okur(32nd) were ahead of players like Ben Wallace(41st) and Rasheed(66th) in PER last year. The Wallace brothers are better than Gooden and Okur, letalone that much worse.

PER also gives too much credit to things like defensive rebounds and assists. A team would rather take a player who averages 5 blks/gm rather than 6 assists/game. Yet the 2 players would have almost the same PER. It is also less reliable in the college game because of the varying schedules and opponents teams play.

Salim probably won't be an all-star and will face an uphill battle to make it in the league as a 2nd rounder playing on a team loaded with swingmen, but he's not the worst college player in the draft.
Back to top
View user's profile Send private message
jemagee



Joined: 05 Nov 2005
Posts: 38


PostPosted: Tue Nov 08, 2005 10:22 pm Post subject: Reply with quote
gkrndija wrote:


PER also gives too much credit to things like defensive rebounds and assists. A team would rather take a player who averages 5 blks/gm rather than 6 assists/game. Yet the 2 players would have almost the same PER. It is also less reliable in the college game because of the varying schedules and opponents teams play.


Then I fell the team who takes the 5 blocks is stupid. I think blocks are the most overrated stat in basketball. Plus let's break down blocks - blocks that lead to a change of possession and those that don't, and could we please get a stat that tracks goaltending PLEASE, somewhere, in a box score.
Back to top
View user's profile Send private message
Kevin Pelton
Site Admin


Joined: 30 Dec 2004
Posts: 684
Location: Seattle

PostPosted: Tue Nov 08, 2005 10:51 pm Post subject: Reply with quote
Can we save the PER discussion for elsewhere? The PER issues pointed out don't really explain why Stoudamire rates poorly.

I do tend to disagree with the Stoudamire discussion. Profit shot 31.2% on 3s as a senior. Dehere was much better (39.6%) but still nowhere near the 50+% Stoudamire shot. I'm more of A/TO fan than most here, but I think separating into Assist Rate and TO Rate captures the fact that Stoudamire just doesn't get a lot of assists.

I was a little surprised to see that subjective Stoudamire comp Steve Kerr had such an outstanding A/TO ratio as a senior (a little better than four to one), though he of course wasn't playing alongside Mustafa Shakur.

Quote:
And even consider that the model we used for this analysis has a built in bias towards younger players based upon their greater "potential."

Given how important measuring potential is in comparing guys who come out after their freshman and senior seasons, I'd like to hear more about how potential was accounted for.

Quote:
Using data from 1980 � 1999, players were identified as successful if they were in the top 10% of all players in either offensive or defensive efficiency.

Am I safe in assuming these are Dean's Offensive/Defensive ratings? Is this over the course of their careers?

Quote:
A binary regression was used with success as the dependent variable and available college statistics as the independent variables.

I'm curious what statistics you used or started with. Were you looking more to measure performance or skills?
Back to top
View user's profile Send private message Send e-mail Visit poster's website AIM Address
kbche



Joined: 19 Jul 2005
Posts: 51
Location: washington d.c.

PostPosted: Tue Nov 08, 2005 11:29 pm Post subject: Reply with quote
What is the origin of the assist/TO ratio significance? This ratio will approach infinity as TO approaches 0 regardless of the number of assists. Also, how does this ratio correlate with other commonly used stats? Assists are difficult to conceptually use in any predictive manner because they are highly dependent on other basketball stats.

KAB
Back to top
View user's profile Send private message Send e-mail
JonathanG



Joined: 20 Apr 2005
Posts: 29


PostPosted: Wed Nov 09, 2005 3:54 pm Post subject: Reply with quote
Coming from a more traditional scouting point of view, I really have to question ranking NBA draft prospects solely based on their statistical production. Defensively is where this kills me the most. There is just no way you can rank a guy's defense based strictly off his steals, blocks and rebounds.

I didn't dig that deep into the numbers last year (certainly not the way that a lot of you guys do), but I would imagine that using that kind of criteria, a guy like Michael Harris from Rice would rank as a lottery pick in an article like that while Deron Williams would probably go in the 2nd round. Its never that simplistic. As Kevin implied by mentioning the potential factor, players don't get drafted based on the type of players they are right now, they are drafted based on the kind of pros teams project them to evolve into. There are many examples of players who had unimpressive stats in college from an efficiency standpoint or however you want to call it, but still went on to have fantastic NBA careers in the long run.

There also seem to be quite a bit of inaccuracies in this article. It says for example that Travis Diener "shot only 42% from the field, worst in the draft." Just off the top of my head, I can tell you that Linas Kleiza shot worse at around 40% from the field, so Diener definitely isn't the worst here. There are probably others. But let me ask you, does shooting 43% from behind the arc as a junior/senior on a high number of attempts as your team's first (only?) offensive option not count for anything anymore? Especially when the same guy got to the line almost 6 and a half times per game. I'm not much of a stats guy, but in my opinion that's pretty damn impressive. This has been harped on already, but the 7 assists to 2.5 turnovers per game on a mediocre team with very little offensive firepower besides him is something that HAS to be taken into consideration here. Especially when Salim Stoudamire gets completely written off because of that same stat. I thought that Diener would be a guy that would be severely overrated by an article like this, but for some reason he gets killed.

Another example...the writer calls Julius Hodge a better shooter than Rashad McCants. Has he ever bothered watching either of these guys play? Then he goes on to write that Bracey Wright (41% FG, 33% 3P) "shoots well from the field".

All in all I think this guy really plays around with the numbers and just draws whatever is convenient for him, which is probably the #1 thing that frustrates me the most with certain people using stats to make a point.


Lemme stress that I think there is certainly value in analyzing NCAA statistics to a certain extent, but not in such a conclusive way that puts this kind of stuff in such a black and white context. For the most part we are talking about players who have not yet come into their own as basketball players. There is a way to use this data in a way that makes sense, but only as just another tool that can be used to evaluate pro potential, definitely not the only one.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
John Hollinger



Joined: 14 Feb 2005
Posts: 79


PostPosted: Wed Nov 09, 2005 4:11 pm Post subject: Reply with quote
Actually, the A/TO difference between Kerr and Stoudamire is pretty easy to figure out -- Kerr played the point, and Salim didn't.
Back to top
View user's profile Send private message AIM Address
benalamar



Joined: 08 Nov 2005
Posts: 3


PostPosted: Wed Nov 09, 2005 4:59 pm Post subject: Reply with quote
To start, I'd like to agree with Johnathan, that you cannot do drafting based on stats alone. I would never seriously suggest otherwise. Statistical analysis has alot to offer, but it is part of the story, not the whole book.

As for questions on the model:

>Given how important measuring potential is in comparing guys who >come out after their freshman and senior seasons, I'd like to hear more >about how potential was accounted for.

I simply used draft age and found that this a negative impact on the projected probability of success that differed by position.


>Am I safe in assuming these are Dean's Offensive/Defensive ratings? Is >this over the course of their careers?

Yes these were Dean's numbers, but I only used the level at year 3. Originally the analysis was planned to check success at years 1, 3 and 5 seperately and examine the differences, but some plans are grander than alotted the alotted time.

>I'm curious what statistics you used or started with. Were you looking >more to measure performance or skills?

I used a lot of different statistics including stadard per game and per minute variables, PER (not to get the PER discussion started again), physical variables such as height and weight, total games played etc. This was largely a kitchen sink type of project where I looked for the statistically significant varaibles.
Back to top
View user's profile Send private message
Kevin Pelton
Site Admin


Joined: 30 Dec 2004
Posts: 684
Location: Seattle

PostPosted: Thu Nov 10, 2005 1:22 am Post subject: Reply with quote
Ben, I forgot in my first post to thank you for dropping by. Much appreciated.

benalamar wrote:
Yes these were Dean's numbers, but I only used the level at year 3. Originally the analysis was planned to check success at years 1, 3 and 5 seperately and examine the differences, but some plans are grander than alotted the alotted time.

Hmm ... the time issue is completely understandable and you start to get too many guys out of the league when you look too far off, but I wonder if going only to the third year is enough for potential.

I wouldn't be surprised if Sean May is better than Marvin Williams two years from now; nor would I be surprised if Williams subsequently surpassed him.

To say this more systemically, in this piece I look at minutes played for high school early entrants vs. players with at least two years of college experience/age 20 (what I thought at the time the age limit would end up being). The preps-to-pros guys play less minutes in years one through three, but catch up in year four. The difference here isn't as significant, since everybody's gone through at least one year of college, but still worth thinking about.

I also wonder whether using Dean's stats favors role players at the expense of less efficient go-to guys (though Iverson was mentioned as qualifying as a "success").
Back to top
View user's profile Send private message Send e-mail Visit poster's website AIM Address
Mark



Joined: 20 Aug 2005
Posts: 807


PostPosted: Thu Jun 29, 2006 7:12 pm Post subject: Reply with quote
Salim Stoudamire's first season: 12.5 PER, but .545 TS%. Both sides of the earlier discussion can probably say they were partially right.

What did the first year performances suggest to folks about projection methodology and improvements?

What is to come from this crop in NBA? Any picks for surprise changes from 1st year results?

I'll venture that Danny Granger continues up and gets more press. His 3pt%/inside shot% combo rating was 33%/33%. If he moved closer to 40/40 that would be great for him and the team.

And for the next draft crop? Now that teams are known, any further more specific predictions?

Larry Bird also got Shawne Williams and James White, 2 other players who should go at least 30/30 and hopefully for them closer to 40/40. Toronto might want Bargnani eventually to be 40/40 along with Charlie V. but I assume his inside shot % will probably start in the 20s.
Foye and Douby probably rate somewhere in 30s/30s and perhaps others will too- but not sure what to expect from Roy, Gay, Brewer on 3pt%. How much will Morrison go inside?

I also noticed this piece today on Jordan/Morrsion pick. I heard Jordan had to warm up to the idea and the majority of the other voices were enthusiastic for Morrison earlier before he came around to agreeing.
http://www.protrade.com/insight/Insight ... 706009&x=x
I assume using a value added perspective Morrison's expected scoring trumped Thomas or Gay in other areas.

this link can be used to get back to the original 2005 draft article
http://tinyurl.com/l33jr

Generally pretty good set of projections, looking back at them. One on the other side, I expected Frye to do quite well (just on gut feel, not detailed study) but the article's assessment was more cautious.

Choose one
PTS/(FGA+.44*FTA)
27%
27% [ 3 ]
PTS/(2*(FGA+.44*FTA))
72%
72% [ 8 ]
Total Votes : 11

Author Message
KnickerBlogger



Joined: 30 Dec 2004
Posts: 101


PostPosted: Thu Jan 27, 2005 10:21 am Post subject: SoapBox: PSA, TS%, ERA??? (Standardization) Reply with quote


I think one of the things we should try to accomplish as a group is to standardize the stats we use. It's important if we want to have non-APBR people try to understand the work we've done, that they can go from one site to the next without a conversion chart.

One thing that has confused me is PSA & TS%. A year ago when I initially tried my hand at stats, I had used the equation PTS/(FGA+FTA) for PSA. I thought the name "points per shot attempt" would be just that. However what Hollinger calls PSA in his book is more accurately described as adjusted PSA (aPSA) because of the .44*FTA factor.

Next, I've thought we came to an agreement to call this TS%,because of Kevin's excellent primer. As Kevin pointed out to me, TS% is half of PSA. I know that as people develop stats on their own, everyone is going to come up with the same stat that is essentially multiples of each other (per 40 minute stats or per 48 minute stats?). However there is no reason to continue with two different stats that are identical. Imagine if baseball had ERA (ER*9/IP) and RAI (runs allowed per inning ER/IP). That's exactly how (for lack of a better word) absurd having two seperate stats are.

Everyone here may know that PSA=TS%*2 (thanks Kevin for bringing me up to speed), but it's hard when you're trying to bring the message to a wider audience. An audience who's statistical knowledge may be limited to FG% and per game stats because that's all they've seen in their basketball coverage. It's even harder when the stats have misleading names likes "points per shot attempt."

So what I'm asking is this:

1. Decide which we should use: PTS/(FGA+.44*FTA) or PTS/(2*(FGA+.44*FTA))

2. Decide what that stat is called. If we use the latter, I think TS% is a fine choice. If the former, we need to come up with a more precise name. (My choice - aPSA)

[I know we had a similar discussion about this, but I really think we went one step too quickly.]
Back to top
View user's profile Send private message Visit poster's website AIM Address Yahoo Messenger
Dan Rosenbaum



Joined: 03 Jan 2005
Posts: 295
Location: Greensboro, North Carolina

PostPosted: Thu Jan 27, 2005 10:50 am Post subject: Reply with quote
I would prefer True Shooting Percentage, i.e. dividing by two. That puts this measure in direct competition with field goal percentage, which is where I think it belongs.

Another option would be to not divide by two, but multiply by 100. That would put it on the same scale as offensive and defensive ratings. That said, my preference is still for True Shootng Percentage.
Back to top
View user's profile Send private message Send e-mail Visit poster's website Yahoo Messenger
Ed Küpfer



Joined: 30 Dec 2004
Posts: 310
Location: Toronto

PostPosted: Thu Jan 27, 2005 11:57 am Post subject: Reply with quote
I also vote for division by 2. We already have a points per attempt-ish number -- Offensive Rating. Let's not confuse the issue: make it a percentage that ranges from zero to 100%, and that's something most people understand intuitively.
_________________
ed
Back to top
View user's profile Send private message
WizardsKev



Joined: 03 Jan 2005
Posts: 309
Location: Washington, DC

PostPosted: Thu Jan 27, 2005 12:49 pm Post subject: Reply with quote
I'll third the motion. True Shooting Percentage, divide by 2.
Back to top
View user's profile Send private message Visit poster's website AIM Address Yahoo Messenger
KnickerBlogger



Joined: 30 Dec 2004
Posts: 101


PostPosted: Thu Jan 27, 2005 12:53 pm Post subject: Reply with quote
I'm only uncomfortable using division by 2, because the Prospectus/Forecast and basketball-reference.com don't. Those are the two "mainstream" sources that use either (that I know of - 82games uses neither).

As a blogger, I have to appeal to a wide audience, and need to use stats that the common fan (read those that don't track their own stats) can identify with an understand. If they don't understand what a .60 TS% is, by not being able to look at players/eras they're familiar with, then for those readers the stat is useless. Without one or both of those switching over, I would feel highly uncomfortable using them.
Back to top
View user's profile Send private message Visit poster's website AIM Address Yahoo Messenger
Ben



Joined: 13 Jan 2005
Posts: 103
Location: Iowa City

PostPosted: Thu Jan 27, 2005 2:09 pm Post subject: Reply with quote
I like PSA because it has a direct meaning. How do you explain TS%? You explain how you estimate points per shot attempt (and what you mean by shot attempt) and then, uh, err, divide by 2. Why 2, why not divide by 3, 2.5, pi, or max PSA?

I don't really have strong feelings on the matter, but just thought I'd put forth an argument for my minority position. Wink
Back to top
View user's profile Send private message
Dan Rosenbaum



Joined: 03 Jan 2005
Posts: 295
Location: Greensboro, North Carolina

PostPosted: Thu Jan 27, 2005 2:38 pm Post subject: Reply with quote
Ben wrote:
I like PSA because it has a direct meaning. How do you explain TS%? You explain how you estimate points per shot attempt (and what you mean by shot attempt) and then, uh, err, divide by 2. Why 2, why not divide by 3, 2.5, pi, or max PSA?

I don't really have strong feelings on the matter, but just thought I'd put forth an argument for my minority position. Wink

I think you are overstating your case a bit. We divide by two so it is directly comparable to the competing field goal percentage statistic. Which is precisely the reason why I think we should divide by two. The quicker we run field goal percentage out of business as an important statistic the better.
Back to top
View user's profile Send private message Send e-mail Visit poster's website Yahoo Messenger
Ben



Joined: 13 Jan 2005
Posts: 103
Location: Iowa City

PostPosted: Thu Jan 27, 2005 2:57 pm Post subject: Reply with quote
Dan Rosenbaum wrote:
Ben wrote:
I like PSA because it has a direct meaning. How do you explain TS%? You explain how you estimate points per shot attempt (and what you mean by shot attempt) and then, uh, err, divide by 2. Why 2, why not divide by 3, 2.5, pi, or max PSA?

I don't really have strong feelings on the matter, but just thought I'd put forth an argument for my minority position. Wink

I think you are overstating your case a bit. We divide by two so it is directly comparable to the competing field goal percentage statistic. Which is precisely the reason why I think we should divide by two. The quicker we run field goal percentage out of business as an important statistic the better.


But FG% actually is a percentage. What precisely do you mean by "directly comparable"? FG% also happens to indicate 1/2 your PSA if you took no FT's or 3FGs, while TS% is 1/2 your PSA?

Of course, I agree on your last point.
Back to top
View user's profile Send private message
KnickerBlogger



Joined: 30 Dec 2004
Posts: 101


PostPosted: Thu Jan 27, 2005 4:03 pm Post subject: Reply with quote
Dan Rosenbaum wrote:
Ben wrote:
I like PSA because it has a direct meaning. How do you explain TS%? You explain how you estimate points per shot attempt (and what you mean by shot attempt) and then, uh, err, divide by 2. Why 2, why not divide by 3, 2.5, pi, or max PSA?

I don't really have strong feelings on the matter, but just thought I'd put forth an argument for my minority position. Wink

I think you are overstating your case a bit. We divide by two so it is directly comparable to the competing field goal percentage statistic. Which is precisely the reason why I think we should divide by two. The quicker we run field goal percentage out of business as an important statistic the better.


Why do we have to make it look like FG% to drive FG% out of business? If anything should replace FG% it's clearly eFG%. I feel that PSA/TS% should be it's own animal, so that it's clearly distinctive from FG%.

I agree with Ben that "PSA" is simple to understand as a function of the number of shots taken. While it's not as simple as PTS/FGA, it's easy to understand that a "PSA" of 1.00 means that if player A is ending the possession by shooting, he's likely to score one point. Similarly that a player with a 1.2 will get 12 points on 10 possessions or a player with a .8 will get 8 on those same 10. You don't get that same feel with TS%, until you multiply it by two. "PSA" is just more intuitive.

Mike
Back to top
View user's profile Send private message Visit poster's website AIM Address Yahoo Messenger
HoopStudies



Joined: 30 Dec 2004
Posts: 277
Location: Bay Area, California

PostPosted: Thu Jan 27, 2005 4:15 pm Post subject: Reply with quote
Ben wrote:
I like PSA because it has a direct meaning. How do you explain TS%? You explain how you estimate points per shot attempt (and what you mean by shot attempt) and then, uh, err, divide by 2. Why 2, why not divide by 3, 2.5, pi, or max PSA?


I personally don't use this ____ stat much because I don't like combining FT effectiveness with FG effectiveness except in occasional cases. But when I do, it's to assess a player's ability to both create their shot (FGA + 0.44*FTA, on its own) and how efficient they are. As a consequence, I think of it as a weighted average of eFG% and FT%. If you're looking for an explanation, that's what it is.
_________________
Dean Oliver
Consultant to the Seattle Supersonics
Author, Basketball on Paper
http://www.basketballonpaper.com
Back to top
View user's profile Send private message Visit poster's website
Dan Rosenbaum



Joined: 03 Jan 2005
Posts: 295
Location: Greensboro, North Carolina

PostPosted: Thu Jan 27, 2005 8:22 pm Post subject: Reply with quote
HoopStudies wrote:
I personally don't use this ____ stat much because I don't like combining FT effectiveness with FG effectiveness except in occasional cases. But when I do, it's to assess a player's ability to both create their shot (FGA + 0.44*FTA, on its own) and how efficient they are. As a consequence, I think of it as a weighted average of eFG% and FT%. If you're looking for an explanation, that's what it is.

That's interesting that you think about it that way. I don't like separating FT effectiveness from FG effectiveness. A player does not decide to shoot a free throw. His decision often is whether or to put himself in a position to draw a foul. The cost of putting himself in that position is that he probably is going to shoot a lower percentage on those field goal attempts. Personally I want to capture this tradeoff when I compare a player like Kobe Bryant who draws a lot of fouls to a player like Jamal Crawford who does not. So for me true shooting percentage is a pretty fundamental concept. In contrast, I sometimes look at eFG% and very rarely look at FT%. I almost never look at FG%, except when it is brought up by other people.
Back to top
View user's profile Send private message Send e-mail Visit poster's website Yahoo Messenger
mtamada



Joined: 28 Jan 2005
Posts: 77


PostPosted: Fri Jan 28, 2005 9:43 pm Post subject: Reply with quote
Dan Rosenbaum wrote:
I don't like separating FT effectiveness from FG effectiveness.


Yes, I agree with DanR (and with the "divide by two" crowd). I won't quote DanR's whole response but it states the case for combining FTA and FGA very well.

There certainly are times when we'll want to split scoring into two sub-components as DeanO suggests (FG shooting and getting to the FT line and making FTs), but for that matter there are plenty of times when we'll want to split scoring into 3 sub-components (3PT FG%, 2PT or overall FG%, and FTM & FTA), and or even more (fast breaks vs halfcourt, contested vs open, set shot vs shooting on the move, etc.).

But there's other times when a single "effective scoring percentage" is useful, and does away with the need to separately look at FG% stats and FTM stats.
Back to top
View user's profile Send private message
mtamada



Joined: 28 Jan 2005
Posts: 77


PostPosted: Sat Jan 29, 2005 12:02 am Post subject: Re: SoapBox: PSA, TS%, ERA??? (Standardization) Reply with quote
KnickerBlogger wrote:
I2. Decide what that stat is called. If we use the latter, I think TS% is a fine choice. If the former, we need to come up with a more precise name. (My choice - aPSA)


Although nobody voted for it in the other thread, I like to call it "shooting efficiency".

What does the "S" is "TS%" stand for? "Scoring" could imply all the things that go into creating a score: not turning the ball over, creating assists, etc. So I prefer "Shooting", because that's what that these stats look at --FG% and FG% -- while ignoring turnovers and assists.
Back to top
View user's profile Send private message
HoopStudies



Joined: 30 Dec 2004
Posts: 277
Location: Bay Area, California

PostPosted: Sat Jan 29, 2005 2:40 am Post subject: Reply with quote
Dan Rosenbaum wrote:

That's interesting that you think about it that way. I don't like separating FT effectiveness from FG effectiveness. A player does not decide to shoot a free throw. His decision often is whether or to put himself in a position to draw a foul. The cost of putting himself in that position is that he probably is going to shoot a lower percentage on those field goal attempts. Personally I want to capture this tradeoff when I compare a player like Kobe Bryant who draws a lot of fouls to a player like Jamal Crawford who does not. So for me true shooting percentage is a pretty fundamental concept. In contrast, I sometimes look at eFG% and very rarely look at FT%. I almost never look at FG%, except when it is brought up by other people.


In all these things, it's a matter of how finely you want to slice it. I like slicing these two right away typically because it's easy. I also find greater significance in my studies when slicing them. If I combine things, significance is harder to find in strategy evaluation. That doesn't mean I don't do it, but I think blending skills can introduce noise that can make it more difficult to find significance. _Can_ make it. In some way, they are the same skill (guys who penetrate), but not for guys who draw a lot of foul shots by going to the boards. And, as always, I am interested in skills and roles more than some measure of overall quality.

I don't have any problem with whatever we're calling this stat. I just choose to keep the ability to draw fouls separate from the ability to make shots most of the time.
_________________
Dean Oliver
Consultant to the Seattle Supersonics
Author, Basketball on Paper
http://www.basketballonpaper.com
Back to top
View user's profile Send private message Visit poster's website
bchaikin



Joined: 27 Jan 2005
Posts: 206
Location: cleveland, ohio

PostPosted: Sat Jan 29, 2005 5:21 pm Post subject: Reply with quote
i too like to combine the two, so you are comparing "apples to apples". after all does a coach really care how you score, just as long as you do score?...

i look at any one FGA or two FTAs as each being a single scoring opportunity, and use:

(2pters + 1.5x3pters + FTM/2)/(FGA + FTA/2)

and call this a "Scoring FG%" for lack of a better term, and:

PTS/(FGA + FTA/2)

and call this points per scoring opportunity...

Eli W



Joined: 01 Feb 2005
Posts: 391


PostPosted: Mon Dec 05, 2005 8:11 pm Post subject: Optimal ratio of twos to threes Reply with quote
Several commentators have recently pointed out that Golden State ranks #1 in 3PA per game yet is #25 in 3P%. Their conclusion is that the Warriors take too many three pointers. I wanted to determine whether this was accurate, and more generally what the optimum ratio of twos to threes was for the typical NBA team.

To do this my goal was to calculate Golden State's points per two-point shot and points per three-point shot. The big complicator is figuring how often teams are fouled on twos compared to threes (a further complicator would be their different FT% on each due to guys who take threes tending to be better foul shooters, but I'll ignore that).

A = missed two not fouled
B = missed two and fouled
C = made two not fouled
D = made two and fouled
E = missed three not fouled
F = missed three and fouled
G = made three not fouled
H = made three and fouled

Some of these can be derived from known numbers such as 2PA, 2PM, 3PA, 3PM, and opponent shooting fouls (which is listed on 82games.com).

2PA = A + C + D
2PM = C + D
3PA = E + G + H
3PM = G + H
Opponent shooting fouls = B + D + F + G

Plugging in the numbers for Golden State:

1010 = A + C + D
477 = C + D
147 = E + G + H
457 = G + H
202 = B + D + F + G

From this you can derive A = 533 and E = 110.

The final formulas I would use are these:

Points off twos = B*2*FT% + C*2 + D*2 + D*FT%
Points off twos = FT%(2B + D) + 2(C + D)
Two-point shots = A + B + C + D
Points per two-point shot = points off twos/two-point shots

Points off threes = F*3*FT% + G*3 + H*3 + H*FT%
Points off threes = FT%(3F + H) + 3(G + H)
Three-point shots = E + F + G + H
Points per three-point shot = points off threes/three-point shots

Thus the key values that must be determined are B, D, F & H (since C + D and G + H are already known). Does anyone know how to calculate these? Or is there some easy way to do this that I missed completely? I suppose since it's still early in the season I could just comb through all the play-by-plays.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
mtamada



Joined: 28 Jan 2005
Posts: 233


PostPosted: Tue Dec 06, 2005 4:30 am Post subject: Re: Optimal ratio of twos to threes Reply with quote
John Quincy wrote:
Several commentators have recently pointed out that Golden State ranks #1 in 3PA per game yet is #25 in 3P%. Their conclusion is that the Warriors take too many three pointers. I wanted to determine whether this was accurate, and more generally what the optimum ratio of twos to threes was for the typical NBA team.

[analysis deleted]

Thus the key values that must be determined are B, D, F & H (since C + D and G + H are already known). Does anyone know how to calculate these? Or is there some easy way to do this that I missed completely? I suppose since it's still early in the season I could just comb through all the play-by-plays.


Very good analysis. 82games.com probably has this info, but not on their freely available data pages. Probably your best bet is the play-by-play route.

You might get some value out of looking at the short (4 or 5 articles) thread "Three Pointers - Foul and Block %s". http://sonicscentral.com/apbrmetrics/vi ... .php?t=468
Ed Kupfer does some regression analysis to estimate percentages of 2pt and 3pt FGAs blocked, and maybe something similar could be done with FTAs on 3pt and 2pt FGAs (however there's the problem of some FTAs coming as the result of a loose ball foul or non-shooting foul in the penalty).

But his most convincing stats were obtained from the PbPs.

Also I lay out 6 steps for analyzing the relative effectiveness of 2pt vs 3pt shooting. The FT analysis that you describe is basically Step 4, although there's also Step 3 to take into account: teams which rely on 3pt FGAs have an additional advantage due to the extra shot attempts they will get by virtue of getting extra offensive rebounds. Step 5 is harder still, the TO% on plays in which a team goes for a 2pter vs a 3pter; and Step 6 harder still, the need to diversify the offense and mix in 2pters and 3pters so the defense can't concentrate on your 3pt FGAs. In the last article MikeG implicitly suggests a Step 7, roster imbalance if a team emphasizes players' 3pt shooting ability at the expense of other qualities.

However, you can do your calculation (what I'm calling Step 4) with just the information that you already have (my Steps 1 and 2) plus the missing information that you seek. I.e. you can ignore the other steps while calculating the values that you seek -- the TS% on 2pt vs 3pt FGAs. Those would be highly useful numbers to have.

But useful though they'd be, they'd ignore the value of the offensive rebounds, the possible differences in TO%, the need for an offense to feature both 2pt and 3pt FGAs, and MikeG's roster considerations (which I don't think are a large factor in the long run, but Steps 5 through 7 are pretty speculative).
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 2350
Location: Delphi, Indiana

PostPosted: Tue Dec 06, 2005 7:30 am Post subject: Reply with quote
There may be several other uncertainties. Since basketball games are seldom single-possession affairs, you want to optimize your lineup to include rebounding, defense, etc, and not just scoring.

Further, do we know the difference in OffReb% between the typical 3-pt'er and the typical 2? What about the relative frequency of putback (high-% followup) attempts? What about opponent OffRtg after a missed 3 vs. a 2?

Fouls not-in-the-act are of course not directly due to shot attempts; but how many guys are fouled while jockeying for position beyond the arc? Or when receiving a pass out there? And then, what is the real effect of causing foul trouble for the opponent?

A balanced attack is always optimal, in the long run. The threat of drives, or a post threat, sets up the 3s. The threat of the 3 opens the middle. Last year, teams ranged from 12% (Uta) to 29% (Phe) of their FGA being 3s. Both teams have sound coaching and could be assumed to be doing 'the right thing', in this regard.

The Warriors are trying 31% of their shots from 3-land. Disregarding OffReb and FT from this strategy, that is a .482 eFG%. On their 2-pt and FT shots (combined), they are converting at .533. Their OffReb% is only .265, 22nd in the league
Back to top
View user's profile Send private message Send e-mail
Eli W



Joined: 01 Feb 2005
Posts: 391


PostPosted: Tue Dec 06, 2005 12:45 pm Post subject: Reply with quote
mtamada, you're right that there are lots of other factors to consider. I seem to remember someone doing a study that contradicted the traditional wisdom that missed threes lead to more offensive boards than missed twos.

Mike G, those numbers you calculate include FTs from non-shooting fouls, right? So that's not really the eFG% for twos.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
oprice



Joined: 25 Aug 2005
Posts: 2
Location: North Carolina

PostPosted: Tue Dec 06, 2005 1:19 pm Post subject: Reply with quote
If you're seeking a ratio, then it seems that the solution would require at least some element of game theory (and I'm not sure how this can be quantified). I guess you could also use empirical data to come up with a ratio that has proven to work in the past. Otherwise, I think you're left with declaring either twos or threes as being optimal -- not a ratio.

If you're interested in the extremes of 3-point shooting, then there are a few examples of teams shooting about half their field goals from beyond the arc in NCAA Div 1 basketball (although I suspect the three is a far better prospect in college). Many of which use elements of the "Princeton Offense" (Princeton, Air Force, Samford, West Virginia, etc.) and have had success at producing an efficient offense. I can say that, from looking at these college teams, teams which shoot more threes tend to have trouble getting offensive rebounds. I think this's probably more a result of the offensive motions (pulling big guys out away from the basket) than shot allocation, but I think it's probably true that threes are less likely to generate offensive rebounds than twos. I'm not sure if this decrease is enough to offset the additional opportunities, however. These teams also shoot and make less free throws per possession than the average college team while turning it over at about the same rate or slightly less. I also believe that these teams have slightly worse defenses because the opponent has more "medium break" opportunities off of rebounds. The opponent spends less time retrieving the ball from the basket (or after free throws) than it would against an average team. Anyway, those are my amateur observations of the effect of threes on the college game.
Back to top
View user's profile Send private message Send e-mail
Ed Küpfer



Joined: 30 Dec 2004
Posts: 681
Location: Toronto

PostPosted: Tue Dec 06, 2005 2:08 pm Post subject: Reply with quote
John Quincy wrote:
I seem to remember someone doing a study that contradicted the traditional wisdom that missed threes lead to more offensive boards than missed twos.


I think DeanO did that in his book, and found no difference in the rebound rates between two- and three-pointers. I don't have time to duplicate that study (which involved pbp data, I think), but I ran a quick regression on team game-by-game stats, using team offensive rebounds as the response variable, with two- and three-point misses as the independant variables. I forced the intercept to zero.


Code:
Predictor Coef SE Coef T P

2ptMisses 0.284310 0.001526 186.36 0.000
3ptMisses 0.262618 0.005855 44.86 0.000


This shows that 2 point misses are slightly more likely to be offensively rebounded. I'm not happy with this model, as the RMSE is about 4.5 OR/game, which is larger than I expected. But it does seem to suggest that conventional wisdom may not be correct here.
_________________
ed
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 2350
Location: Delphi, Indiana

PostPosted: Tue Dec 06, 2005 2:34 pm Post subject: Reply with quote
That's nifty stuff, Ed. Which is the conventional wisdom: what we used to remember? or what we thought before that?

I may be biased against over-reliance on the 3. I lumped in all FT with the 2-pt attempts because, as I said, there isn't much fouling out beyond the arc. Being fouled before the attempt is usually (not always) closer to the basket.

The few FT that occur with 3-pt shots are likely offset by worse putback opps, etc. I guess I'm saying, Unless your 3-pt eFG% is greater than your overall TS%, you should be going inside more.

Further complicating things, a certain number of 3's are just desperation heaves that shouldn't really count against their %'s.
Back to top
View user's profile Send private message Send e-mail
John Hollinger



Joined: 14 Feb 2005
Posts: 146


PostPosted: Tue Dec 06, 2005 2:53 pm Post subject: Reply with quote
I had done some work on rebounding on 3s vs. 2s a few years ago as well, and my conclusion was that missed 3s are harder to rebound, not easier. Which makes sense -- the most dangerous offensive rebounder is the shooter, and on a 3 he's got quite a haul to get to the boards.
Back to top
View user's profile Send private message AIM Address
HoopStudies



Joined: 30 Dec 2004
Posts: 640
Location: Near Philadelphia, PA

PostPosted: Tue Dec 06, 2005 3:36 pm Post subject: Reply with quote
People simply forget that there are a TON of easy offensive boards off 2pt shots. On help defense on the interior, a man leaves his guy to stop a shot, leaving a very easy offensive board for the man he left. And if an offensive player gets inside position and misses a chippy, he's right there to try again. In this case, conventional wisdom needs modification -- 3pt jumpers are easier to rebound than 2pt jumpers (not 2pt shots).

The NBA understands this. Most teams act as though they know that 2pt shots are as easy to rebound as 3s even if people are saying otherwise. So this seems to be more a media thing than a true mistake in conventional wisdom. I never heard it from my coaches.
_________________
Dean Oliver
Author, Basketball on Paper
http://www.basketballonpaper.com
Back to top
View user's profile Send private message Visit poster's website
mtamada



Joined: 28 Jan 2005
Posts: 233


PostPosted: Tue Dec 06, 2005 10:17 pm Post subject: Reply with quote
Wait a minute, when we say "easy" or "hard" to rebound, do we mean for the shooting team or the defensive team?

Also, what is the conventional wisdom? Here is my perception: most reporters and fans think that offensive rebounding rates are lower on 3pters than on 2pters. Some hoopstatisticians have found that the rates are similar, but I don't know if there has been a definitive study. 82games.com at this point probably has enough observations to make a real good estimate. I don't know what coaches think, but I trust DeanO's perception of their thinking. Except I don't know if he's saying they think rebounds of 3pters are harder for the O or harder for the D, compared to missed 2pters.
Back to top
View user's profile Send private message
gabefarkas



Joined: 31 Dec 2004
Posts: 1086
Location: Durham, NC

PostPosted: Tue Dec 06, 2005 10:36 pm Post subject: Re: Optimal ratio of twos to threes Reply with quote
mtamada wrote:

Also I lay out 6 steps for analyzing the relative effectiveness of 2pt vs 3pt shooting. The FT analysis that you describe is basically Step 4, although there's also Step 3 to take into account: teams which rely on 3pt FGAs have an additional advantage due to the extra shot attempts they will get by virtue of getting extra offensive rebounds.


I could have sworn that we conclusively disproved that 3PTers lead to a higher % of ORebs than 2PTers, no?
Back to top
View user's profile Send private message Send e-mail AIM Address
mtamada



Joined: 28 Jan 2005
Posts: 233


PostPosted: Tue Dec 06, 2005 10:52 pm Post subject: Re: Optimal ratio of twos to threes Reply with quote
gabefarkas wrote:
mtamada wrote:

Also I lay out 6 steps for analyzing the relative effectiveness of 2pt vs 3pt shooting. The FT analysis that you describe is basically Step 4, although there's also Step 3 to take into account: teams which rely on 3pt FGAs have an additional advantage due to the extra shot attempts they will get by virtue of getting extra offensive rebounds.


I could have sworn that we conclusively disproved that 3PTers lead to a higher % of ORebs than 2PTers, no?


That's not what the claim is, at least not in the way that I think you're wording it.

Here are two similar but different questions:

1. If the Sonics miss a FGA, what % of the time do they grab the offensive rebound? (And, if we had the necessary data, we could see if this percentage is different for 2pt vs 3pt attempts.)

2. If the Sonics ATTEMPT a FGA, what % of the time do they grab the offensive rebound? (Not that this is how we would necessarily word, it, but it makes it easy to see how this is different from the question above.)

If we look at question 2, 3pt FGAs have a big advantage over 2pt FGAs. 2pter are made, what, about 47% of the time? So only 53% of the time will there be even an opportunity to grab an offensive rebound.

Whereas even a good longball team such as the Sonics will miss a lot of their 3pt FGAs -- leaving a lot of offensive rebounding opportunities.


Again, the wording above is not the clearest way to actually carry out the investigation, but meant to show where the possible advantage in offensive rebounding lies. I don't think anyone claims that under question 1, 3pt FGAs have an advantage; I don't think anyone claims that they have a higher off rebd %.

It's under question 2 where the 3pters have an advantage. NOT due to higher offensive rebound percentages, but due to higher offensive rebound OPPORTUNITIES -- because of more missed shots.
Back to top
View user's profile Send private message
gabefarkas



Joined: 31 Dec 2004
Posts: 1086
Location: Durham, NC

PostPosted: Tue Dec 06, 2005 10:58 pm Post subject: Re: Optimal ratio of twos to threes Reply with quote
mtamada wrote:

It's under question 2 where the 3pters have an advantage. NOT due to higher offensive rebound percentages, but due to higher offensive rebound OPPORTUNITIES -- because of more missed shots.


More opportunities, but not a higher % of opportunities, so what's the point? The defense is also getting more DRebs.

So, it's tough to look at just that in a vacuum. We need to consider if teams are good O-Rebounders to begin with, right?
Back to top
View user's profile Send private message Send e-mail AIM Address
mtamada



Joined: 28 Jan 2005
Posts: 233


PostPosted: Tue Dec 06, 2005 11:53 pm Post subject: Re: Optimal ratio of twos to threes Reply with quote
gabefarkas wrote:
mtamada wrote:

It's under question 2 where the 3pters have an advantage. NOT due to higher offensive rebound percentages, but due to higher offensive rebound OPPORTUNITIES -- because of more missed shots.


More opportunities, but not a higher % of opportunities, so what's the point? The defense is also getting more DRebs.


You're not taking into account the possessions which are lost due to made FGs (if a team successfully scores, they automatically have to give the ball to the opponent).

And 2pt shots attempts result in a higher percentage of made FGs -- and thus a higher percent of balls being given to the opponent.

Whereas 3pt shots, with their higher number of misses, give the shooting team more opportunities to get the ball back. They give the defensive team FEWER opportunities to get the ball back -- sure, the defensive rebound opportunities are higher, but the automatic possessions gained after made FGs are fewer.


Example: assume that a team shoots 50% on 2pters, and 33.3% on 3pters. Assume that it grabs 33.3% of offensive rebound opportunities on both (note that no one is claiming that 3pters have a higher offensive rebound %). For simplicity, assume that the team never commits any TOs, nor shoots any FTs.

The team will have a TS% of 50% on both 2pters and 3pters.

But it will have 1 point per possession if it does nothing but shoot 2pters (obviously I'm ignoring the game theoretic aspects and assuming that all percentages stay constant).

And, if I've calculated the series correctly, it will have 9/7 = 1.29 points per possession if it does nothing but shoot 3pters, thanks to the additional offensive rebounds and resulting additional FGAs it gets.

Working out the first few iterations:

1/3 of the time it will successfully sink a 3pter, scoring 3pts and giving the ball to the opponent (i.e. ending the possession).

2/3 of the time it will miss; 2/3 of these (so 4/9 of the time overall) the defensive team will grab the rebound, ending the possession (with no point scored).

But 1/3 of those misses (so 2/9 of the time overall) the shooting team will grab the offensive rebound and be able to shoot another 3pter. It will go in 1/3 of the time, so 2/27 of the time overall it will score 3pts on its second shot and end the possession.

The other 4/27 of the time it misses this second shot. 2/3 of these misses, so 8/81 overall, get rebounded by the opponent, ending the possession.

But 1/3 of them, so 4/81 overall, get rebounded by the shooting team, giving them a third shot. 1/3 of them go in, 4/243 overall.

Etc.

The resulting Points Per Possession is the sum of (2/9)^i, with i ranging from 0 to infinity, which sums to 9/7 if I've done the math correctly.

Obviously a more realistic calcuation has to take into account FTs (which JohnQ originally asked about, because the data are not readily available), TOs, etc.

Plus the items that MikeG raised:

1. Possible differences in the OR% of missed 3pters vs missed 2pters. And possible resulting fast breaks (or medium breaks as MikeG put it) if the opponent grabs the defensiv erebound.

2. Possible differences in TS% or expected points per play after grabbing the offensive rebound.

3. Effect of additional fouls putting you into the bonus sooner, or getting an opposing player into foul trouble.

Etc.
Back to top
View user's profile Send private message
HoopStudies



Joined: 30 Dec 2004
Posts: 640
Location: Near Philadelphia, PA

PostPosted: Wed Dec 07, 2005 10:53 am Post subject: Reply with quote
Offensive rebounding percentage is higher off 2pt attempts than off 3pt attempts. It's been shown (though not a huge difference).

There are broadcasters who imply the opposite.
_________________
Dean Oliver
Author, Basketball on Paper
http://www.basketballonpaper.com

gabefarkas



Joined: 31 Dec 2004
Posts: 664
Location: NYC

PostPosted: Thu Jul 06, 2006 1:03 pm Post subject: Will a new game ball affect play? Reply with quote
http://www.nba.com/news/blackbox_060628.html?loc=bullet

Quote:
The NBA is introducing a new Official Game Ball for play beginning in the 2006-07 season. The new ball, manufactured by Spalding, features a new design and a new material that together offer better grip, feel, and consistency than the current leather ball.


Is it presumptive to say that something like this will reduce turnovers, increase assists, or anything else significant like that?
_________________
Statistics are like a woman's bikini. What it reveals can be fascinating, but what it conceals is ultimately critical!
Back to top
View user's profile Send private message Send e-mail AIM Address
HoopStudies



Joined: 30 Dec 2004
Posts: 433
Location: Near Philadelphia, PA

PostPosted: Thu Jul 06, 2006 2:09 pm Post subject: Re: Will a new game ball affect play? Reply with quote
gabefarkas wrote:
http://www.nba.com/news/blackbox_060628.html?loc=bullet

Quote:
The NBA is introducing a new Official Game Ball for play beginning in the 2006-07 season. The new ball, manufactured by Spalding, features a new design and a new material that together offer better grip, feel, and consistency than the current leather ball.


Is it presumptive to say that something like this will reduce turnovers, increase assists, or anything else significant like that?


It's a great ball. The feel is very nice. My shot was horrible the day I used it, but I blame stress for that. I can't imagine that we'll see any effect large enough to blame on the ball.
_________________
Dean Oliver
Author, Basketball on Paper
http://www.basketballonpaper.com
Back to top
View user's profile Send private message Visit poster's website
Dan Rosenbaum



Joined: 03 Jan 2005
Posts: 461
Location: Greensboro, North Carolina

PostPosted: Thu Jul 06, 2006 3:35 pm Post subject: Reply with quote
With what I think was the new basketball, I was able to palm the ball and flip it up to the basket back-handed (palm facing the ground). I have always been able to palm the ball, but this was amazing.
Back to top
View user's profile Send private message Send e-mail Visit poster's website Yahoo Messenger
bchaikin



Joined: 27 Jan 2005
Posts: 440
Location: cleveland, ohio

PostPosted: Thu Jul 06, 2006 4:09 pm Post subject: Reply with quote
in terry pluto's Loose Balls (1990) about the old ABA (1967-1976), it talks about the first season of the league and the red, white, and blue ball used. it mentions that a number of players complained about the ball being slippery early on that first season, but as steve jones (aba veteran and long time nba commentator) surmised, it was because all of the balls were in fact new, and that typically in the nba balls that had been used for awhile were chosen for game balls. guess we'll see during this upcoming exhibition season if these new ones are thought by players to be too slippery too...
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Mark



Joined: 20 Aug 2005
Posts: 670


PostPosted: Wed Jul 12, 2006 7:13 pm Post subject: The new ball- a list of possible impacts tossed up Reply with quote
Will "beter grip" mean NBA players will be able to go into a one handed dunk scoring move off the dribble in traffic with more security? Is a squeeze enough? Alley oop passing and finishing now a little easier?

Will rebounds be sticker to the first hand that touches it? Advantage defense?

Tip put backs and dunk shooting %s likely to go up?

Stickiness hurt or help 'long practicing might not like change' "pure shooters" more or less than scorers or not as patterned and perfectionist their way shooters?

Potential impacts on passing, receiving, dribbling, loose ball situations.

Quickier to be into shooting motion and release because of quicker grip control? More 3 pointers? Or do you have to consciously apply a light touch?

How similar in feel to college ball? Do rookies have less adjustment than in past and is that a slight uptick for their stock?

Are small hands for a big man now less of a liability? Are big hands a bigger advantage or can too big be a problem as reported with Rondo for his shot?

Ball behavior on the rim? Back board?


Some of these may come into play, others not. But near 200 possessions in a game and 82 games I expect some impact. I havent felt the ball yet to know if it is a big deal, but if the difference is instantly noticeable to those who have, it might do more than they thought. If there are micro but meaningful impacts the teams /player that recognize them and adjust fastest and best might gain a slight edge.

Last edited by Mark on Wed Oct 04, 2006 12:15 am; edited 5 times in total
Back to top
View user's profile Send private message
HoopStudies



Joined: 30 Dec 2004
Posts: 433
Location: Near Philadelphia, PA

PostPosted: Thu Jul 13, 2006 12:05 pm Post subject: Reply with quote
Good questions, but I think the new ball feels a lot like the best of the old balls, which were usually the ones that players chose for a game. I think I remember that Gary Payton liked a slicker one, which messed other people up more than it messed him up. So that inconsistency would upset some people. But I don't expect to see much difference from it.
_________________
Dean Oliver
Author, Basketball on Paper
http://www.basketballonpaper.com
Back to top
View user's profile Send private message Visit poster's website
Mark



Joined: 20 Aug 2005
Posts: 670


PostPosted: Thu Jul 13, 2006 1:10 pm Post subject: Reply with quote
Since the topic was brought up I thought I'd catalogue the possible impacts and then watch for any signs of them. Maybe it ends ups a non-issue. The condition of the ball does matter as well as you note.

Last edited by Mark on Tue Oct 03, 2006 11:29 pm; edited 1 time in total
Back to top
View user's profile Send private message
Mark



Joined: 20 Aug 2005
Posts: 670


PostPosted: Tue Oct 03, 2006 2:38 pm Post subject: Shaq no fan of NBA's new ball Reply with quote
Yahoo article with negative reactions to ball from Heat players.
http://tinyurl.com/ln6vu
And later in another article Riley backs Shaq up: "I'm right with him," Heat coach Pat Riley said. "I think it's horrible..."

Article in original post said all players got a new ball earlier this summer but wonder how many took advantage of the time to really get use to it. Wade hadn't but he had something of an excuse with WC and Fiba ball.


Didnt really hear any commotion about it coming out of summer legue but few media reps play attention to that. A good amount of negative reaction to any change is to be expected. It would be a modest problem if an impact lasted more than a few weeks, more serious if it lasted beyond a season.

"Steve Nash said the ball has a tacky feel that's making shooting and certain types of passes tricky." Maybe there will be a middle ground like with the old ball. Break it in some, keep it dry. If it is more uniform, maybe they could substitute a completely dry ball instead of just wiping it for a common towel? Maybe they might also look at special treated moisture wipes or dehumidifiers.

And I'd look into changing the rims too to try to reduce damage to fingers on dunks. Maybe apply a polymer coating to the outside or light padding to reduce shock impact while maintaining equivalent dimensions.

Last edited by Mark on Sun Nov 05, 2006 5:12 pm; edited 2 times in total
Back to top
View user's profile Send private message
Mark



Joined: 20 Aug 2005
Posts: 670


PostPosted: Wed Oct 18, 2006 3:16 pm Post subject: Reply with quote
"There is no question the ball is different. Its dimensions are the same, but the feel of the cover is not. And the ball reacts differently, too, say a couple of local physicists.

Jim Horwitz is chair of the physics department at UT-Arlington, and Kaushik De is the project leader and a physics professor at UTA. They were asked by Mavericks owner Mark Cuban to do a study about the differences between composite and leather balls.

What the physicists found was that the new ball, when dropped from a height of five feet, bounced an average of four inches lower than the old ball. In addition, the synthetic ball was much slower to absorb moisture, a key complaint among players who seem to be constantly losing their grip on the new spheroid in the preseason.

"When water came in contact with both balls, the leather ball absorbed it more quickly and was more easily gripped," Horwitz says. "We suspect that will be the biggest difference in the physical characteristics of the balls."

Horwitz and De are not being paid by Cuban, and their research is ongoing. They plan to research with friction tests how the ball reacts to hardwood flooring and human skin, as well as spin control.

Another complaint of players is that the ball does not react the same off the backboard, making bank shots more adventurous.

from dallasnews.com
http://tinyurl.com/y4wpcm
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 1139
Location: Delphi, Indiana

PostPosted: Wed Oct 18, 2006 4:04 pm Post subject: Reply with quote
I guess I've never really thought about what happens to 'moisture' when it gets on the ball. If it's absorbed, the ball must get heavier. And all along, I've thought those sweaty, humid days just made the ball FEEL heavier, because your energy is sapped.

It makes perfect sense, though. And I think it's a heavy argument in favor of the old ball. Whatever advantages might be found in a non-absorbing ball, once it is wet those advantages are gone.
_________________
40% of all statistics are wrong.
Back to top
View user's profile Send private message Send e-mail
Climate



Joined: 18 Oct 2006
Posts: 2


PostPosted: Wed Oct 18, 2006 6:22 pm Post subject: Reply with quote
Are they using this new ball for the preseason games?

That would be an easyway to gather info on the new ball.

Last edited by Climate on Thu Oct 19, 2006 1:23 am; edited 1 time in total
Back to top
View user's profile Send private message
Mark



Joined: 20 Aug 2005
Posts: 670


PostPosted: Wed Oct 18, 2006 7:47 pm Post subject: Reply with quote
Yes it was in summer league and now is being used in preseason.

Last edited by Mark on Sun Nov 05, 2006 5:09 pm; edited 1 time in total
Back to top
View user's profile Send private message
gabefarkas



Joined: 31 Dec 2004
Posts: 664
Location: NYC

PostPosted: Thu Oct 19, 2006 10:32 am Post subject: Reply with quote
Mike G wrote:
I guess I've never really thought about what happens to 'moisture' when it gets on the ball. If it's absorbed, the ball must get heavier. And all along, I've thought those sweaty, humid days just made the ball FEEL heavier, because your energy is sapped.

It makes perfect sense, though. And I think it's a heavy argument in favor of the old ball. Whatever advantages might be found in a non-absorbing ball, once it is wet those advantages are gone.


I think it's more like added moisture will change the friction factor between the hand and the ball, and thus change how well the ball can be gripped.
_________________
Statistics are like a woman's bikini. What it reveals can be fascinating, but what it conceals is ultimately critical!

Author Message
John Quincy



Joined: 01 Feb 2005
Posts: 270


PostPosted: Wed Apr 19, 2006 11:51 pm Post subject: New book by Dave Berri Reply with quote
I recently heard a talk by Malcolm Gladwell where he mentioned in passing a book on evaluating NBA players that he was reviewing. It turns out it's a new book by David Berri and two other econ professors called Wages of Wins, that looks at all sports including basketball. It looks like it comes out in May, but there's already a website out with some sample chapters available:

http://wagesofwins.com
Back to top
View user's profile Send private message Send e-mail Visit poster's website
davis21wylie2121



Joined: 13 Oct 2005
Posts: 327
Location: Atlanta, GA

PostPosted: Thu Apr 20, 2006 12:23 am Post subject: Reply with quote
Sounds really cool, especially the part about the NFL. Let's hope that his new "Win Score" system doesn't rank Dennis Rodman as the most valuable player in the NBA again, though...
Back to top
View user's profile Send private message
bchaikin



Joined: 27 Jan 2005
Posts: 497
Location: cleveland, ohio

PostPosted: Thu Apr 20, 2006 12:28 am Post subject: Reply with quote
they must have been reading old APBRmetrics postings...

The story of Antoine Walker illustrates our point. According to the NBA Efficiency model, Walker was the 35th most productive player in the NBA in 2004–05. Given that more than 450 players played that season, if we believe the NBA model, Walker was ranked in the top 10% of all players. A key reason for Walker’s lofty ranking was his scoring totals. With an average of 19.1 points per game, he also ranked in the top 40 among scorers. Although Walker achieved a high scoring average, a bit of inspection reveals he was not a very efficient producer of points. In the 2004–05 campaign, 35 players took more than 300 shots from the three-point range. Walker took 341, and with a three-point field goal percentage of 32%, his accuracy from this distance among the 300 plus shooters ranked 34th, or second to last. His two-point field goal percentage was also relatively poor. Walker was one of twenty players who took at least 1,000 shots from two-point range. Given a shooting percentage of 45%, he ranked fifteenth—out of twenty—in accuracy among prolific shooters from two-point range. So Walker shot relatively poorly from both inside and outside the arc. Still, his percentages exceeded the thresholds imposed by the NBA model, and consequently Walker earned high marks in NBA Efficiency despite his inefficient scoring.

What happens when we view Walker through the lens of the productivity model we employ? The Win Score measure we propose imposes the cost of the shot attempt regardless of whether or not the shot is made. As a result, a player must connect on at least 50% of two-point shots and 33% of three pointers for the benefit of shooting to equal the cost. Walker failed to achieve these levels; consequently, his Win Score was below the average player at his position. Specifically, Walker posted a Win Score of 493.5 in 2004–05, while an average player at his position playing his minutes would have posted a Win Score of 635. So the NBA Efficiency model, which does not value shot attempts correctly, argues that Walker is an above average performer. When we take into account the cost of Walker’s many field goal attempts, we see that Walker is actually below average.

Back to top
View user's profile Send private message Send e-mail Visit poster's website
Doc319



Joined: 02 Feb 2005
Posts: 126


PostPosted: Thu Apr 20, 2006 5:04 am Post subject: Reply with quote
No one would argue that Walker is efficient, but Pat Riley, one of the winningest coaches in NBA history, liked his game enough to acquire Walker when Riley was in the GM/President role, and then to play him in the sixth man role when Riley became Heat coach. Kobe used the "love me or hate me" line in his commercials, but that could be applied at least as much to Walker. I understand his deficiencies but he played a key role on Boston's playoff teams in both of his stints there and has contributed to Miami's division title this year (obviously the bulk of the credit goes to Wade and Shaq). Pierce had a very good season without Walker this time around, better than he did without Walker last time, but Boston slid back to being a non-playoff team. At least part of the reason is that Al Jefferson and others were not able to replace what Walker--statistical warts and all--supplied.

Here is a link to NBA.com's Sixth Man Award page, followed by that page's text about Walker:

http://www.nba.com/awards2006/thecase_sixth_walker.html

Antoine Walker - Miami Heat
Sixth Man Award

2005-06:
Leads the team in three-pointers made (133) ... Ranks fourth on the team in scoring (12.2), fifth in rebounds (5.2) and fifth in assists (2.0) ... Scored in double-figures 54 times this season ... Scored 20-plus points nine times ... Recorded four double-doubles ... Recorded five or more assists seven times ... Scored 32 points, grabbed eight boards and dished out five assists in a win over Toronto 4/11 ... Scored 25 points and grabbed 16 boards in the first game of the season, 4/2 at Memphis ... Is the only Heat player to have played in all 81 games so far.

Our take:
When he and Celtics coach Red Auerbach invented the concept of sixth man in the late '50s, Frank Ramsey set in stone the role sixth men still fill today: instant offense through many shots.

Which brings us to Miami's Antoine Walker as a sixth man. Why didn't anyone think of this before?

A career 19.0 points-per-game scorer, unapologetic gunner and a starter in every NBA game in which he ever played until this season, Walker fills the sixth man role perfectly for the Heat. With two go-to guys in the starting lineup (Dwyane Wade and Shaquille O'Neal), Walker keeps the pressure on when he enters the game with his 12.2 points and 5.2 boards per game. He also proves to be a nice outside complement to Shaq in the paint and the slashing Wade.

While some would argue that Walker's more instant shooting than instant offense, you can't score if you don't shoot. It's nice to see Walker keeping the tradition alive.
-- Rob Peterson

From the coach:
If he's put in that role every single night, he'd average 20 points per game. You sort of forget a little bit how good he can be when he gets a number of good opportunities ... I don't think there's any doubt what kind of skill level he has.
-- Pat Riley, after Walker scored 32 points against Toronto on 4/11.

A fan's take:
'Toine, a lifetime starter, has accepted his role and has delivered well. He has made numerous big shots and has had a good amount of double-digit scoring fourth quarters. His 32 point, eight rebound, five assist game should seal the deal on the award, considering all the players who were out.
-- Sam - West Hartford, Conn.

The Numbers:
G GS MPG PPG RPG APG SPG BPG FG% 3P% FT%
81 18 26.9 12.2 5.2 2.0 0.57 0.36 .436 .354 .622

Sixth Man Award - The Candidates:
Speedy Claxton - New Orleans/Oklahoma City Hornets
Derek Fisher - Golden State Warriors
Mike Miller - Memphis Grizzlies
Jerry Stackhouse - Dallas Mavericks
Antoine Walker - Miami Heat

Others worth consideration: Earl Boykins, Channing Frye, Devin Harris, Eddie House, Bobby Jackson, Fred Jones, Alonzo Mourning, Ronald Murray, Charlie Villanueva, Maurice Williams


--David Friedman
http://20secondtimeout.blogspot.com/
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Nikos



Joined: 16 Jan 2005
Posts: 338


PostPosted: Thu Apr 20, 2006 9:13 am Post subject: Reply with quote
http://www.wagesofwins.com/WOWCh8.htm

Check out the Jordan Reg Season vs Playoffs comparison -- reminds me of Mike G's Reg vs Playoff Rates. Why are they saying MJ got worse in the playoffs when he clearly got better when statistically compared to the average playoff performer tendency to DECLINE? Did he not account for the fact that all playoff performers decline and that team effieicncies decline and can only be measured against the opponent?
Back to top
View user's profile Send private message
Ben



Joined: 13 Jan 2005
Posts: 190
Location: Iowa City

PostPosted: Thu Apr 20, 2006 9:48 am Post subject: Reply with quote
davis21wylie2121 wrote:
Sounds really cool, especially the part about the NFL. Let's hope that his new "Win Score" system doesn't rank Dennis Rodman as the most valuable player in the NBA again, though...


I think we can safely say it won't be Dennis Rodman. It might be leading rebounder Kevin Garnett though. Maybe Shawn Marion.
Back to top
View user's profile Send private message
John Quincy



Joined: 01 Feb 2005
Posts: 270


PostPosted: Thu Apr 20, 2006 11:11 am Post subject: Reply with quote
Nikos wrote:
http://www.wagesofwins.com/WOWCh8.htm

Check out the Jordan Reg Season vs Playoffs comparison -- reminds me of Mike G's Reg vs Playoff Rates. Why are they saying MJ got worse in the playoffs when he clearly got better when statistically compared to the average playoff performer tendency to DECLINE? Did he not account for the fact that all playoff performers decline and that team effieicncies decline and can only be measured against the opponent?


If you read the paragraph just before the table it sounds like they did account for that, though their method isn't perfectly clear to me.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Mike G



Joined: 14 Jan 2005
Posts: 1431
Location: Delphi, Indiana

PostPosted: Thu Apr 20, 2006 5:46 pm Post subject: Reply with quote
I'll read the article when I have the time; but if they haven't accounted for the many stifling defensive series between the Bulls and the Pistons/Knicks/Jazz/Sonics, then they could well assume Jordan's numbers 'declined' in his playoffs. If they observe that virtually everyone else in those series declined even more, then they should notice some simple ratios.
_________________
40% of all statistics are wrong.
Back to top
View user's profile Send private message Send e-mail
Mike G



Joined: 14 Jan 2005
Posts: 1431
Location: Delphi, Indiana

PostPosted: Thu Apr 20, 2006 7:31 pm Post subject: Reply with quote
OK, I read enough of it. It's ugly writing, to put it mildly. It also seems to be written by someone(s) who have no inkling of the game of basketball. It could even be said to give statistical analytics a bad name.

Citing the 1998 Finals between the Bulls and Jazz, the authors reveal that M. Jordan didn't shoot particularly well. Here are the TS% for the Bulls, along with their PPG:

Code:

Chicago Bulls Eff% PPG
Toni Kukoc .556 15.2
Dennis Rodman .543 3.3
Steve Kerr .528 3.8
Michael Jordan .514 33.5
Scottie Pippen .501 15.7
Luc Longley .490 5.0
Scott Burrell .450 3.5
Ron Harper .417 5.3


Jordan averaged 33.5 in games that averaged scores of 88-80. He scored 42% of what the opposition scored. This might be an alltime high.

Rodman and Kerr were not going to score scads of points, and Kukoc was somewhat an offensive specialist. Jordan took no more shots than he should have taken; he outscored Karl Malone by 8.5 PPG (up from 1.7 in the season); 'helped' (to put it mildly) hold Stockton and Hornacek to 4.2 PPG less than their season averages. And so on.

Dispelling myths is one thing. Picking some stats and ignoring others (to make a point?) might be accidental or deliberate confusion. Any earnest fan who watched the '98 Finals knows that Jordan dominated throughout and had the final say. It wasn't anything new; there is hardly a playoff series of the '90s he didn't dominate.

I guess I have to say these guys have demonstrated how analysis can take the reader further from the truth, rather than closer. Maybe they should try telling the Utah players how badly Jordan played those 6 games. Responses might prove enlightening.
_________________
40% of all statistics are wrong.
Back to top
View user's profile Send private message Send e-mail
Nikos



Joined: 16 Jan 2005
Posts: 338


PostPosted: Thu Apr 20, 2006 8:14 pm Post subject: Reply with quote
Mike G wrote:
OK, I read enough of it. It's ugly writing, to put it mildly. It also seems to be written by someone(s) who have no inkling of the game of basketball. It could even be said to give statistical analytics a bad name.

Citing the 1998 Finals between the Bulls and Jazz, the authors reveal that M. Jordan didn't shoot particularly well.

Jordan averaged 33.5 in games that averaged scores of 88-80. He scored 42% of what the opposition scored. This might be an alltime high.

Rodman and Kerr were not going to score scads of points, and Kukoc was somewhat an offensive specialist. Jordan took no more shots than he should have taken; he outscored Karl Malone by 8.5 PPG (up from 1.7 in the season); 'helped' (to put it mildly) hold Stockton and Hornacek to 4.2 PPG less than their season averages. And so on.

Dispelling myths is one thing. Picking some stats and ignoring others (to make a point?) might be accidental or deliberate confusion. Any earnest fan who watched the '98 Finals knows that Jordan dominated throughout and had the final say. It wasn't anything new; there is hardly a playoff series of the '90s he didn't dominate.

I guess I have to say these guys have demonstrated how analysis can take the reader further from the truth, rather than closer. Maybe they should try telling the Utah players how badly Jordan played those 6 games. Responses might prove enlightening.


Yeah I figured you would say this, but I still wanted to make sure it was said. Well done!
Back to top
View user's profile Send private message
Dan Rosenbaum



Joined: 03 Jan 2005
Posts: 491
Location: Greensboro, North Carolina

PostPosted: Thu Apr 20, 2006 9:35 pm Post subject: Reply with quote
I really wish that the academic sports economists would spend more time in forums like this. I have learned a lot here, and I am sure they would too.

Berri and his co-authors do make an adjustment for playoff basketball, but they assume that the effect is a fixed amount for every player. A better assumption probably would be that the drop in performance is proportional to their Win Score measure. Under that assumption, Jordan's performance improved in all but two of those playoffs.

But that probably is understating Jordan's performance. In general, as teams go deeper in the playoffs, the teams they play become better defensively. So the dropoff in performance probably should have been greater for Jordan than for the average playoff performer just due to the better defensive competiton. Account for this as well, and I would imagine that Jordan may have improved every year in the playoffs.

I have only seen the snippets of this book on the web, and while I think books like this in general are a good thing, it is unfortunate that they do not seem to recognize the good work that people have done on these same topics (and I am not talking about me). Now many of these people doing this work are not academics, but academics do not have a monopoly on the right way to do things.
Back to top
View user's profile Send private message Send e-mail Visit poster's website Yahoo Messenger
Mike G



Joined: 14 Jan 2005
Posts: 1431
Location: Delphi, Indiana

PostPosted: Fri Apr 21, 2006 4:43 am Post subject: Reply with quote
I was probably a bit harsh, but since Nikos quotes me in toto, I can't edit my phrasing. It's ironic that the passage I read brought back the Bulls/Jazz Finals. During one of those ('97 or '98) I saw a sports reporter on TV, on separate nights, interview Malone and Rodman.

To Malone, he said something like, "Well, your numbers are down, your team isn't winning, blah blah..." Malone left before punching the guy out. With Rodman, it was like, "Well, your team is winning ugly, your numbers are down, blah blah..." Rodman bolted, in tears.

Since the world's greatest rebounder/defender was guarding the world's greatest (forward) scorer/rebounder -- wouldn't it make perfect sense that numbers would be down for both guys? Could it possibly be otherwise? Do people crave attention so desperately, they'll resort to just being stupid, just to make someone mad on TV?

When two 66-win teams meet at the Finals, they aren't both going to win 80% of their games. And their players' stats are going to look like less than those of 66-game-winning teams. On average, they'll look like the stats of average players.

Dig: In those 2 Finals, we saw the greatest scorer ever, the greatest rebounder, the greatest passer; the greatest power forward ever; and Pippen. There were many classic moments -- enjoyed by some, derided by others.
_________________
40% of all statistics are wrong.
Back to top
View user's profile Send private message Send e-mail
Nikos



Joined: 16 Jan 2005
Posts: 338


PostPosted: Fri Apr 21, 2006 6:25 am Post subject: Reply with quote
Well the 'ugly' part might have been harsh, but most of what you posted is on point.

Quote:
Dig: In those 2 Finals, we saw the greatest scorer ever, the greatest rebounder, the greatest passer; the greatest power forward ever; and Pippen. There were many classic moments -- enjoyed by some, derided by others.


Hopefully Tim Duncan will have something to say about that in the next few years.
Back to top
View user's profile Send private message
94by50



Joined: 01 Jan 2006
Posts: 386
Location: Phoenix

PostPosted: Fri Apr 21, 2006 9:58 am Post subject: Reply with quote
Nikos wrote:
Well the 'ugly' part might have been harsh, but most of what you posted is on point.

Quote:
Dig: In those 2 Finals, we saw the greatest scorer ever, the greatest rebounder, the greatest passer; the greatest power forward ever; and Pippen. There were many classic moments -- enjoyed by some, derided by others.


Hopefully Tim Duncan will have something to say about that in the next few years.

Some people might say he's there already. Duncan's already established himself in the game's Pantheon - two MVPs, a perennial all-league first-teamer, the best player on three championship teams. If he matches Malone's longevity, then I doubt there would be an argument.
Back to top
View user's profile Send private message
John Quincy



Joined: 01 Feb 2005
Posts: 270


PostPosted: Mon May 22, 2006 11:35 am Post subject: Reply with quote
Here's Malcolm Gladwell's review of the book:

http://www.newyorker.com/critics/books/ ... rbo_books1

Not surprisingly (given Gladwell's background), he takes the Win Scores measure at face value. However, it sounds like it may have some trouble passing the laugh test, as it calculates that "journeyman forward Jerome Williams was actually among the strongest players of his generation."

hpanic7342



Joined: 16 Feb 2005
Posts: 201


PostPosted: Mon May 22, 2006 1:07 pm Post subject: Reply with quote
The review also says that the book says that Ray Allen has had as good a career as Kobe Bryant.

Is there anyone else in the world that thinks this?
Back to top
View user's profile Send private message Send e-mail
kjb



Joined: 03 Jan 2005
Posts: 864
Location: Washington, DC

PostPosted: Mon May 22, 2006 1:11 pm Post subject: Reply with quote
I was a little surprised to see how uncritically Gladwell seemed to accept the book. Gladwell didn't seem aware of other statistical approaches, which may disagree with some of Berri et al.'s conclusions.
Back to top
View user's profile Send private message AIM Address Yahoo Messenger
bchaikin



Joined: 27 Jan 2005
Posts: 685
Location: cleveland, ohio

PostPosted: Mon May 22, 2006 1:24 pm Post subject: Reply with quote
the last paragraph of gladwell's article appears to contain a conclusion from berri's book, and sounds like the passage quoted in the "Reaction to a Passage?" APBR thread posted here on 4/24/06...

“One can play basketball,” the authors conclude. “One can watch basketball. One can both play and watch basketball for a thousand years. If you do not systematically track what the players do, and then uncover the statistical relationship between these actions and wins, you will never know why teams win and why they lose.”

i haven't read the book but based on this article i'd say berri's right about iverson, walker, and gordon, but (if the quote is correct) is reaching on jerome williams. the junkyard dog has had a couple of really good seasons statistically on a per minutes basis, especially the 99-00 season with the pistons (great rebounding, high steals, high ScFG%)....

but in 9 years jerome williams played more than 1700 total minutes in a season just twice. his touches/min were very low throughout his career, never as much as even 0.70, and thus his impact on offense was nothing special (his career ScFG% was less than 2% above the league average during the time that he played). he was an excellent offensive rebounder, and got fouled a ton per touch, but couldn't hit his FTs (less than 65% lifetime)...

he did grab alot of steals per minute, especially for a PF, but was not much of a shot blocker, and even if he was a bruce bowen/ron artest type defender, which i don't believe he was, i think its quite a stretch to call him "...one of the strongest players of his generation..."...

unless they are refering to his weight-lifting abilities...
Back to top
View user's profile Send private message Send e-mail Visit poster's website
HoopStudies



Joined: 30 Dec 2004
Posts: 705
Location: Near Philadelphia, PA

PostPosted: Fri Jun 02, 2006 8:28 am Post subject: Reply with quote
Dave Berri is on AM 610 in Philadelphia 6/2 at 9:30 am. As he says, it's part of his $18.06 book tour.
_________________
Dean Oliver
Author, Basketball on Paper
The postings are my own & don't necess represent positions, strategies or opinions of employers.
Back to top
View user's profile Send private message Visit poster's website
parinella



Joined: 16 Dec 2005
Posts: 10


PostPosted: Tue Jun 06, 2006 3:11 pm Post subject: Reply with quote
Discussion about this book and article appear on a statistics blog here. Anyone interested in regression coefficients should chime in.

The author has linked to several of the members here.
Back to top
View user's profile Send private message
KD



Joined: 30 Jan 2005
Posts: 163


PostPosted: Wed Jun 07, 2006 3:28 am Post subject: Reply with quote
Mike G wrote:
OK, I read enough of it. It's ugly writing, to put it mildly. It also seems to be written by someone(s) who have no inkling of the game of basketball. It could even be said to give statistical analytics a bad name.

Citing the 1998 Finals between the Bulls and Jazz, the authors reveal that M. Jordan didn't shoot particularly well. Here are the TS% for the Bulls, along with their PPG:

Code:

Chicago Bulls Eff% PPG
Toni Kukoc .556 15.2
Dennis Rodman .543 3.3
Steve Kerr .528 3.8
Michael Jordan .514 33.5
Scottie Pippen .501 15.7
Luc Longley .490 5.0
Scott Burrell .450 3.5
Ron Harper .417 5.3


Jordan averaged 33.5 in games that averaged scores of 88-80. He scored 42% of what the opposition scored. This might be an alltime high.

Rodman and Kerr were not going to score scads of points, and Kukoc was somewhat an offensive specialist. Jordan took no more shots than he should have taken; he outscored Karl Malone by 8.5 PPG (up from 1.7 in the season); 'helped' (to put it mildly) hold Stockton and Hornacek to 4.2 PPG less than their season averages. And so on.

Dispelling myths is one thing. Picking some stats and ignoring others (to make a point?) might be accidental or deliberate confusion. Any earnest fan who watched the '98 Finals knows that Jordan dominated throughout and had the final say. It wasn't anything new; there is hardly a playoff series of the '90s he didn't dominate.

I guess I have to say these guys have demonstrated how analysis can take the reader further from the truth, rather than closer. Maybe they should try telling the Utah players how badly Jordan played those 6 games. Responses might prove enlightening.


This post needs to be distributed about the internet. Toni had a fab 1998 Finals, but he was a bit above average for each of the games save Game 5 -- where he shot a ridiculous mark from the field. It was his third straight disappointing postseason, with the previous two coming after nasty injuries (1996, back; 1997, plantar issues).

Gladwell's posts and the subsequent hype this book has received (haven't read it, so I'm giving it the Joe Morgan treatment) seem to be doing a disservice to the scene, maaaan.

RocketsFan



Joined: 23 Oct 2005
Posts: 148


PostPosted: Tue Apr 25, 2006 5:24 pm Post subject: Kevin Garnett's Win Shares Reply with quote
Comparing KG to Richard Jefferson, I noticed that Garnett led in Stops, PProd, and ORtg, yet ranked two win shares behind Jefferson, 33-31. That just doesn't seem right to me.
Justin uses a marginal points produced system to estimate offensive win shares, which works fine. But defensive win shares are estimated through stops, which means there are no players to zero out, since, near as I can tell, it is impossible to have negative stops.
I wondered if it would be possible to create a marginal stops system. It is easy enough to estimate the league stop% as 1-floor%. From there, you could say that a marginal player should have a stop percentage of .92*lgstop%. For the league, this is about .46.
You could also crudely estimate player defensive possessions through Min/TmMin*TmPoss. Marginal Stops then, would be Stops - DPoss*.92*lgstop%. Anyone below the margin would be zeroed out.
What is interesting here is that the T-Wolves had 7411 minutes accounted for by players who were worse than the margin defensively. KG had to carry that team defensively, a lot more than I think the original WS system accounts for.
I must confess that I am not sure what the league marginal stop% should be, if .92 is the right factor or not, because with the adjustments made here, Garnett is given an incredible 33.62 DWS. I can't say I'm convinced the jump in his defensive credit should be that high.
Is this a legitimate way of thinking about the issue? Perhaps the numbers could use some tweaking, like .92, but does it make sense?
Back to top
View user's profile Send private message Yahoo Messenger
94by50



Joined: 01 Jan 2006
Posts: 258
Location: Phoenix

PostPosted: Tue Apr 25, 2006 6:30 pm Post subject: Reply with quote
In the past, I've converted defensive ratings into marginal points allowed, which kind of helps... but probably doesn't help enough.
Back to top
View user's profile Send private message
jkubatko



Joined: 05 Jan 2005
Posts: 299
Location: Albuquerque, NM

PostPosted: Tue Apr 25, 2006 9:41 pm Post subject: Reply with quote
One item that plays a huge factor in this is that the Nets won more games than their statistics suggest they should have. Based on their marginal points, we would predict 32.28 wins for the Timberwolves, and they actually won 33. However, the predicted number of wins for the Nets based on their marginal points is 41.74, and they actually won 49 games. So a marginal point was worth more to the Nets than it was to the Timberwolves. There are basically two options: (a) throw those "extra" wins away and attribute them to luck or (b) credit them to the players. I choose (b).
_________________
Regards,
Justin Kubatko
Basketball Stats!
Back to top
View user's profile Send private message Send e-mail Visit poster's website
jkubatko



Joined: 05 Jan 2005
Posts: 299
Location: Albuquerque, NM

PostPosted: Tue Apr 25, 2006 10:04 pm Post subject: Re: Kevin Garnett's Win Shares Reply with quote
Re: Calculating marginal defense for individual players, I think you would want to do something like this:

Code:

margDef = (MP/tmMP)*oppPoss*(1.08*(lgPTS/lgPoss) - (DRtg/100))


You would zero out any players with a negative value. I just came up with this quickly off the top of my head, so I may have missed something important. Regardless, I'll look into this.
_________________
Regards,
Justin Kubatko
Basketball Stats!
Back to top
View user's profile Send private message Send e-mail Visit poster's website
jkubatko



Joined: 05 Jan 2005
Posts: 299
Location: Albuquerque, NM

PostPosted: Tue Apr 25, 2006 10:09 pm Post subject: Reply with quote
RocketsFan, good catch. Making the adjustment to marginal defense that I described above, Garnett vaults ahead of Jefferson, 35 WS to 33 WS. That seems much more palatable to me. I think I will go ahead and make that change, unless someone sees something terrible wrong with my logic above.
_________________
Regards,
Justin Kubatko
Basketball Stats!
Back to top
View user's profile Send private message Send e-mail Visit poster's website
94by50



Joined: 01 Jan 2006
Posts: 258
Location: Phoenix

PostPosted: Wed Apr 26, 2006 2:20 am Post subject: Re: Kevin Garnett's Win Shares Reply with quote
jkubatko wrote:
Re: Calculating marginal defense for individual players, I think you would want to do something like this:

Code:

margDef = (MP/tmMP)*oppPoss*(1.08*(lgPTS/lgPoss) - (DRtg/100))


You would zero out any players with a negative value. I just came up with this quickly off the top of my head, so I may have missed something important. Regardless, I'll look into this.

I have used

Code:

margDef = (MP/tmMP)*oppPoss*(1.08*(tmPTSA/tmPoss) - (DRtg/100))


- in other words, team points allowed per possession rather than league points per possession.

Last edited by 94by50 on Wed Apr 26, 2006 4:33 pm; edited 1 time in total
Back to top
View user's profile Send private message
RocketsFan



Joined: 23 Oct 2005
Posts: 148


PostPosted: Wed Apr 26, 2006 2:42 pm Post subject: Re: Kevin Garnett's Win Shares Reply with quote
jkubatko wrote:
Re: Calculating marginal defense for individual players, I think you would want to do something like this:

Code:

margDef = (MP/tmMP)*oppPoss*(1.08*(lgPTS/lgPoss) - (DRtg/100))


You would zero out any players with a negative value. I just came up with this quickly off the top of my head, so I may have missed something important. Regardless, I'll look into this.


I think the formula would be (MP/tmMP)*oppPoss*(1.08*(lgPTS/lgPoss) - (DRtg/100)*(MP/tmMP)*oppPoss). In other words, just adding on the player defensive possessions estimate at the end.
Back to top
View user's profile Send private message Yahoo Messenger
jkubatko



Joined: 05 Jan 2005
Posts: 299
Location: Albuquerque, NM

PostPosted: Wed Apr 26, 2006 3:30 pm Post subject: Re: Kevin Garnett's Win Shares Reply with quote
RocketsFan wrote:
I think the formula would be (MP/tmMP)*oppPoss*(1.08*(lgPTS/lgPoss) - (DRtg/100)*(MP/tmMP)*oppPoss). In other words, just adding on the player defensive possessions estimate at the end.


No, I believe what I had was correct. This part:

Code:

1.08*(lgPTS/lgPoss) - (DRtg/100)


estimates the difference per possession between a marginal defender and the given defender, and this part:

Code:

(MP/tmMP)*oppPoss


estimates possessions for the given defender.
_________________
Regards,
Justin Kubatko
Basketball Stats!
Back to top
View user's profile Send private message Send e-mail Visit poster's website
RocketsFan



Joined: 23 Oct 2005
Posts: 148


PostPosted: Wed Apr 26, 2006 3:47 pm Post subject: Reply with quote
I was looking at the parentheses wrong. My mistake.
Back to top
View user's profile Send private message Yahoo Messenger
jkubatko



Joined: 05 Jan 2005
Posts: 299
Location: Albuquerque, NM

PostPosted: Wed Apr 26, 2006 8:25 pm Post subject: Reply with quote
RocketsFan and 94by50,

Thanks to your suggestions, I have modified the way defensive Win Shares are allocated to individual players. I made sure to credit both of you for motivating the work in this blog post. Thanks again for your help.
_________________
Regards,
Justin Kubatko
Basketball Stats!
Back to top
View user's profile Send private message Send e-mail Visit poster's website
edijorj



Joined: 21 Apr 2006
Posts: 30


PostPosted: Wed Apr 26, 2006 11:56 pm Post subject: Reply with quote
So because of the change in how WS is calculated, Shaq's win share is now 2 lower than Udonis Haslem's, instead of 3 lower, which is an improvement but still ridiculous.
Back to top
View user's profile Send private message
94by50



Joined: 01 Jan 2006
Posts: 258
Location: Phoenix

PostPosted: Thu Apr 27, 2006 1:34 am Post subject: Reply with quote
edijorj wrote:
So because of the change in how WS is calculated, Shaq's win share is now 2 lower than Udonis Haslem's, instead of 3 lower, which is an improvement but still ridiculous.

Ridiculous? I'm not sure I agree. If I may:

Shaq missed a third of the season. Shaq's 18 WS came in 1800 minutes of play, while Haslem had to play 2500 minutes to get his 20 WS.

Also, Shaq's offensive efficiency is no longer what it used to be - he peaked at 117, but now it's dropping below 110, and that's bound to make a difference.

Third, more of a minor point, Dwyane Wade assumed a slightly greater role in the offense (Usage Rate 30.1), and Shaq slightly less so (Usage Rate 26.7). While it might not be a significant difference, Shaq simply isn't the first option on offense anymore. Combine that with his decreasing efficiency, and Shaq isn't the player he once was. He's still either the #1 or #2 center in the league, depending on what you think of Yao Ming.

Fourth, check Shaq's 1996 and 1997 seasons, in which he played about as much as he played this season. His WS totals in those years were only somewhat higher than they were this year, and in 1996 he ended up fifth on Orlando in WS, mainly because he missed so much time.

Finally, the defensive ratings in particular aren't perfect. They're forced to rely on team defensive efficiency to generate individual ratings, so it may give Haslem more credit (and Shaq less) than each should get. (On the other hand, the reverse might be true.)

Obviously, WS isn't perfect. Justin would agree with me. But in this case, it's not out of the question that Haslem, on the whole, made a larger contribution (albeit not much larger) to Miami's success than Shaq this season, even though Shaq is still the better and more efficient player.
Back to top
View user's profile Send private message
jkubatko



Joined: 05 Jan 2005
Posts: 299
Location: Albuquerque, NM

PostPosted: Thu Apr 27, 2006 9:43 am Post subject: Reply with quote
edijorj wrote:
So because of the change in how WS is calculated, Shaq's win share is now 2 lower than Udonis Haslem's, instead of 3 lower, which is an improvement but still ridiculous.


If you're going to call a result "ridiculous," you should at least spend a few seconds thinking about what may have contributed to that result. The difference -- as 94by50 stated -- is due to playing time. Shaq earned 10 Win Shares per 1000 minutes played, Haslem 8.
_________________
Regards,
Justin Kubatko
Basketball Stats!
Back to top
View user's profile Send private message Send e-mail Visit poster's website
edijorj



Joined: 21 Apr 2006
Posts: 30


PostPosted: Thu Apr 27, 2006 11:35 pm Post subject: Reply with quote
Quote:
But in this case, it's not out of the question that Haslem, on the whole, made a larger contribution (albeit not much larger) to Miami's success than Shaq


If we look at the 23 games that Shaq missed, Miami went 10-13. Excluding the last two games of the season, which Dwyane Wade also missed, Miami went 10-11. So without Shaq, Miami was looking like it would win 38 (38.1) of its first 80 games. Based on PF (2010) and PA (1992), they were looking like they would win 43 (42.5) of their first 80 games. In fact they won 52 games. So it seems like Shaq added about 12 wins. Even if this is an overestimate, I think it's pretty clear that Udonis Haslem did not add anything close to this number.

Quote:
If you're going to call a result "ridiculous," you should at least spend a few seconds thinking about what may have contributed to that result. The difference -- as 94by50 stated -- is due to playing time.


Did you really think I didn't know that Haslem played more minutes?
Back to top
View user's profile Send private message
jkubatko



Joined: 05 Jan 2005
Posts: 299
Location: Albuquerque, NM

PostPosted: Fri Apr 28, 2006 11:18 am Post subject: Reply with quote
edijorj wrote:
So it seems like Shaq added about 12 wins.


In 2000, O'Neal played 79 games and played 3163 minutes. He had 58 Win Shares that season, tied for third-highest of all time. Let's assume Shaq was worth 12 wins this season, or 36 Win Shares. If we pro-rate his 2006 numbers to 3163 minutes, that's 63 Win Shares. In other words, Shaq in 2006 was better per minute than he was in 2000. That doesn't make sense. When you compare 2006 to 2000, Shaq's Assist Ratio is lower, his Turnover Ratio is higher, his Usage Rate is lower, his Rebound Rate is lower, and his PER is lower.

Also, I'm not out on a limb when I have Haslem with more wins credit than O'Neal. Looking at Dean Oliver's Player Wins, Haslem has 7.5 and O'Neal 6.6.

Quote:
Did you really think I didn't know that Haslem played more minutes?


I have no way of knowing that, so I made an educated guess.
_________________
Regards,
Justin Kubatko
Basketball Stats!

Mike G



Joined: 14 Jan 2005
Posts: 3491
Location: Hendersonville, NC

PostPosted: Fri Apr 28, 2006 8:08 pm Post subject: Reply with quote
I'm still hammering out some of the details for this season, but tentatively I've got Shaq with 9.5 eW, Haslem with 5.6

Quote:
Looking at Dean Oliver's Player Wins, Haslem has 7.5 and O'Neal 6.6.


eW differs from PW and WS in a couple of ways:

- Player eW do not add up to team wins (nor to pythagorean expectations). Miami's player eWins total about 47. This being 6 games over .500 (41/82), eXpected Wins = 41 + 2*6 = 53 (They won 52; pyth = 52.2). The idea is that better players make better teams, but better teams don't make the players better.

- eW allocates little (or nothing) to marginal players, and the lion's share to superstars. Wade had 16.0 this year. Replacing Wade with a 'replacement player' for all his 2896 minutes causes Miami to win 32 fewer games. This rather extreme prediction -- that a team with Shaq could go 20-62 -- assumes the RP doesn't get better as he plays, and no one else can step up.

So middling starters (like Haslem, Walker, JWilliams) playing 30-32 mpg generally get 4-5 eW. A player of this caliber makes a team 8-10 wins better than if they were starting a player like Jason Kapono or Shandon Anderson.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
edijorj



Joined: 21 Apr 2006
Posts: 30


PostPosted: Tue May 23, 2006 1:49 am Post subject: Reply with quote
Quote:
Let's assume Shaq was worth 12 wins this season, or 36 Win Shares. If we pro-rate his 2006 numbers to 3163 minutes, that's 63 Win Shares. In other words, Shaq in 2006 was better per minute than he was in 2000. That doesn't make sense.


I was never arguing that Shaq deserves 36 win shares. I was arguing that Shaq's contribution to Miami's success (which I estimated at 12 wins added) is much larger than Haslem's contribution could believably be. Also, per minute wins added makes little sense; Shaq averaged only 30 mpg this year, those minutes were approximately the minutes that were the most important minutes for Shaq to be in the game. Its unlikely that the minutes Shaq played in 2000, when he averaged 40 mpg, were on average quite as important.

Quote:
Also, I'm not out on a limb when I have Haslem with more wins credit than O'Neal. Looking at Dean Oliver's Player Wins, Haslem has 7.5 and O'Neal 6.6.


It is not at all surprising that WS and PW agree since both are based on the very flawed ORtg and DRtg.
Back to top
View user's profile Send private message
jkubatko



Joined: 05 Jan 2005
Posts: 702
Location: Columbus, OH

PostPosted: Tue May 23, 2006 8:07 am Post subject: Reply with quote
edijorj wrote:
It is not at all surprising that WS and PW agree since both are based on the very flawed ORtg and DRtg.


Could you explain why you feel ORtg and DRtg are "very flawed?" That's a pretty strong statement to make without any explanation. Do you have any suggestions for improving them? Do you have a system that is better?
_________________
Regards,
Justin Kubatko
Basketball-Reference.com
Back to top
View user's profile Send private message Send e-mail Visit poster's website
94by50



Joined: 01 Jan 2006
Posts: 499
Location: Phoenix

PostPosted: Tue May 23, 2006 9:47 am Post subject: Reply with quote
jkubatko wrote:
edijorj wrote:
It is not at all surprising that WS and PW agree since both are based on the very flawed ORtg and DRtg.

Could you explain why you feel ORtg and DRtg are "very flawed?" That's a pretty strong statement to make without any explanation. Do you have any suggestions for improving them? Do you have a system that is better?

Yeah, that is pretty strong. I know DRtg has things that can be improved, but I didn't think ORtg had too many holes in it.
Back to top
View user's profile Send private message
Mark



Joined: 20 Aug 2005
Posts: 807


PostPosted: Tue May 23, 2006 12:37 pm Post subject: Reply with quote
There was a mini-review of Dean O style offensive and defensive ratings here http://sonicscentral.com/apbrmetrics/vi ... .php?t=727

Really I think we need meta-ratings for player impact and that no one or two advanced stats are "enough". A meta-method rating that captured true team scoring offense and defensive ratings or team +/-, counterpart matchups (both would be beter if adjusted) and I think probably some separate treatment for the four factors would be superior to any one of these alone or Dean O offensive and defensive ratings.

I have mentioned this before in several other threads.
http://sonicscentral.com/apbrmetrics/vi ... ight=#8185
http://sonicscentral.com/apbrmetrics/vi ... ight=#6997


The weights for the method components probably should be found be regression for best fit. Putting together this in a form that many would accept and use would be an advance over critiquing each of these current stats that are all insufficient on their own. I am willing to discuss further if any others wish to try a joint construction of such a meta-method rating. I know some prefer to have a handful of separate measures and then evalaute and blend them afterwards into a conclusion. I'd like to see how far you could get blending the number from the methods. It still would come down to judging how you felt about the outcome, the parts and the other things not explicitly accounted for (consistency, pace issues, big game behavior, starter vs subs in the stats, the versatility of multi-skill players to meet different emerging needs, quality as a teammate, responsiveness to coach, player winning strength as much as past performance can give clues and on and on)

And individual player weaknesses can be accepted /perhaps diminished if in a team construction if you have excess capacity to cover their shortfall, as alluded to with big men providing defense and teams surviving their offensive shortfall in another recent thread. Some qualities can come from anywhere, some are best when they come from a big or a guard. It is ultimately team quality I am mostly after, so a meta-team rating is also needed (I tried a preliminary four factor approach here http://sonicscentral.com/apbrmetrics/vi ... ight=#6885) and you can crosswalk between both the meta-player ratings and the individual method ratings to the team ratings depending if you are focused on the overall quality or are tinkering with one specific area to establish your recipe for success. Then throw on additional layers of complexity with the whole set of issues of moneyball to get the players or the performance you want and coaching.

Last edited by Mark on Mon May 29, 2006 12:03 pm; edited 1 time in total
Back to top
View user's profile Send private message
Pinot



Joined: 23 May 2006
Posts: 21


PostPosted: Tue May 23, 2006 1:26 pm Post subject: Reply with quote
jkubatko wrote:
RocketsFan and 94by50,

Thanks to your suggestions, I have modified the way defensive Win Shares are allocated to individual players. I made sure to credit both of you for motivating the work in this blog post. Thanks again for your help.


I feel like a kid in a candy factory meeting Mr. Wonka himself!

I've been reading, using and attempting to distribute your analysis for two years now. Thank you so much!!!
________
Merkur XR4Ti

Last edited by Pinot on Thu Feb 03, 2011 3:07 pm; edited 1 time in total
Back to top
View user's profile Send private message AIM Address
jkubatko



Joined: 05 Jan 2005
Posts: 702
Location: Columbus, OH

PostPosted: Tue May 23, 2006 1:31 pm Post subject: Reply with quote
Pinot wrote:
I feel like a kid in a candy factory meeting Mr. Wonka himself!


This made me chuckle for several reasons, one being that I just watched Charlie and the Chocolate Factory for the first tme this weekend.
_________________
Regards,
Justin Kubatko
Basketball-Reference.com
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Pinot



Joined: 23 May 2006
Posts: 21


PostPosted: Tue May 23, 2006 1:40 pm Post subject: Reply with quote
jkubatko wrote:
Pinot wrote:
I feel like a kid in a candy factory meeting Mr. Wonka himself!


This made me chuckle for several reasons, one being that I just watched Charlie and the Chocolate Factory for the first tme this weekend.


Ah yes, myself finding this site at this time was meant to be! Hopefully I can wow you guys with some weighted 3-year projections, a la Marcels, if I can finish it before preseason.
Back to top
View user's profile Send private message AIM Address
edijorj



Joined: 21 Apr 2006
Posts: 30


PostPosted: Mon May 29, 2006 6:09 am Post subject: Reply with quote
Quote:
Could you explain why you feel ORtg and DRtg are "very flawed?" That's a pretty strong statement to make without any explanation. Do you have any suggestions for improving them? Do you have a system that is better?


I think ORtg and DRtg are very good ways of looking at team offense and team defense. The problem is when ORtg and DRtg are used for players, since it is too difficult, if not impossible, to determine how many possessions a player is responsible for. For example, if Magic gets a rebound, Nixon gets an assist, Wilkes makes the FG, Kareem is doubled teamed and Chones is unguarded, I see no non-arbitrary way to determine player possessions. Dean Oliver's system is probably as good as any rating built around player possessions.

Quote:
The weights for the method components probably should be found be regression for best fit. Putting together this in a form that many would accept and use would be an advance over critiquing each of these current stats that are all insufficient on their own.


I think the best player rating system would be the regression of adjusted on/off player ratings on player box stats and possibly on team box stats. Dan Rosenbaum did something very similar to this. Unfortunately, he ignored certain box stats (he used FTAs and 3pAs but not FTs or 3pts) and weighted crunch time higher than non-crunch time.
Back to top
View user's profile Send private message
Ben



Joined: 13 Jan 2005
Posts: 264
Location: Iowa City

PostPosted: Tue May 30, 2006 10:33 am Post subject: Reply with quote
edijorj wrote:
Quote:
Could you explain why you feel ORtg and DRtg are "very flawed?" That's a pretty strong statement to make without any explanation. Do you have any suggestions for improving them? Do you have a system that is better?


I think ORtg and DRtg are very good ways of looking at team offense and team defense. The problem is when ORtg and DRtg are used for players, since it is too difficult, if not impossible, to determine how many possessions a player is responsible for. For example, if Magic gets a rebound, Nixon gets an assist, Wilkes makes the FG, Kareem is doubled teamed and Chones is unguarded, I see no non-arbitrary way to determine player possessions. Dean Oliver's system is probably as good as any rating built around player possessions.

Quote:
The weights for the method components probably should be found be regression for best fit. Putting together this in a form that many would accept and use would be an advance over critiquing each of these current stats that are all insufficient on their own.


I think the best player rating system would be the regression of adjusted on/off player ratings on player box stats and possibly on team box stats. Dan Rosenbaum did something very similar to this. Unfortunately, he ignored certain box stats (he used FTAs and 3pAs but not FTs or 3pts) and weighted crunch time higher than non-crunch time.


I believe that one thing that Rosenbaum found was that players didn't need a very high true shooting percentage to add value. This isn't the case with win value/win shares. I guess that's what always struck me about win value - it seems to put too much weight on efficiency over the ability to create shots. I couldn't tell you what the values should be, but I find whenever the results seem puzzling, it's always an efficient offensive player that seems overrated by the system in my mind. I can find a few examples if you like, but 2005 Miami is one where I really noticed this.

Mark



Joined: 20 Aug 2005
Posts: 670


PostPosted: Wed May 17, 2006 5:04 pm Post subject: Actual / Projected Stats for the 05-06 Rookie Class Reply with quote
While looking for something else I stumbled across this title
http://www.82games.com/pelton6.htm
and decided to look and see how well Kevin did projecting rookie class stats. I think he did well overall. On about half he was quite close on either everything including minutes or after adjusting hard to predict minutes. On the most of the rest he was low mainly on points but did better on rebounds, assists. Paul and Frye perhaps beat the projection by the most, Bogut was a little short.

Kevin, are you interested in discussing your methods? (there have been a few posts that have shown interest in college to NBA stat translation in recent months) Do you plan any changes to methodology or weighting for the next class of rookies? I assume you plan on waiting until they have been drafted and have gone thru the summer league before posting?

Last edited by Mark on Wed May 17, 2006 7:43 pm; edited 1 time in total
Back to top
View user's profile Send private message
davis21wylie2121



Joined: 13 Oct 2005
Posts: 183
Location: Atlanta, GA

PostPosted: Wed May 17, 2006 6:04 pm Post subject: Reply with quote
Speaking of which, I projected PER's for rookies here before the season. I have no idea if I did well or not, though (seeing as no one else to my knowledge did projected PER's), but since Mark brought up the topic of rookie projections, I thought I'd chime in...
Back to top
View user's profile Send private message Visit poster's website
Mark



Joined: 20 Aug 2005
Posts: 670


PostPosted: Wed May 17, 2006 7:01 pm Post subject: Reply with quote
I did a quick check of about 20 and it looks like you did a good job as well, generally +/- 1-3 on PER. You had Paul at a modest PER15, Frye a little higher at 16 (I didnt make a full set of projections but I was pretty high on him before the draft).

I'll leave it to the two of you to more fully analyze your own results or make side by side comparisons if you want.

Or go ahead next year?
Back to top
View user's profile Send private message
davis21wylie2121



Joined: 13 Oct 2005
Posts: 183
Location: Atlanta, GA

PostPosted: Wed May 17, 2006 7:35 pm Post subject: Reply with quote
I'm sure Kevin's methodology was a lot more scientific than mine. I simply went to the various draft sites (ESPN.com, Hoopshype, NBADraft.net, DraftExpress, etc.) and compiled a list of comparable players, then checked how those comps did at the same age. As you probably know, many of the comps can be outlandish (for example, practically every European SF gets compared to Dirk Nowitzki), so I scaled down as I saw fit, and for those w/o adequate comps I just went on gut instinct. Probably not textbook APBRmetrics, to be sure... If he is interested in posting it, I would love to hear Kevin's methodology as well, as it will likely be of benefit to my "system".
Back to top
View user's profile Send private message Visit poster's website
Mark



Joined: 20 Aug 2005
Posts: 670


PostPosted: Wed May 17, 2006 8:12 pm Post subject: Reply with quote
A few comments on a couple of the guys on Kevin's list of 12:

Ray Felton March 17pts 7 assists 4 rebs and good shooting %s
April 17pts 10 assists 3 rebs and terrible shooting %s
his minutes and average more than doubled from November to April.
If he plays those minutes next year he may compete with Paul on glossy offensive numbers. His one on one defense looks better than Paul on counterpart PER especially at SG.

Jarret Jack closed strong in April 12pt 5 assists in 26 minutes and 55%FG.
I expect him to have a strong 2nd year.

Ike Diogu also closed strong with 12 pts 6 rebs in 22 minutes in April. Will they create more playing time for him? I would.

How close does Charlie V get to 20 pts 10 rebs? Based on improvement shown in final months I expect pretty close. Anything less than 17pts 9 rebs would disappoint me.

Deron Williams probably shows some pts /assists improvement and even more on PER. Maybe he gets some better national press too, especially if the Jazz makes playoffs as they probably should.

What happens to Paul's stats depends heavily on what the Hornets and Claxton do about his free agency. If Claxton leaves, Paul get bigger numbers but Hornets chances for the playoffs probably take a step back.
If Paul goes over 20 pts a night he may remain the most talked about but with as his fellow rookies getting closer to the same level of minutes and advance further in their transition (Paul was quickest with this) it will be a more competitive comparison among them.
Back to top
View user's profile Send private message
Ed Küpfer



Joined: 30 Dec 2004
Posts: 522
Location: Toronto

PostPosted: Wed May 17, 2006 8:48 pm Post subject: Reply with quote
Kevin projected the following stats:
MPG, PPG, RPG, APG, SPG, BPG, and TOPG, along with some fantasy-style stats that need not concern us. First thing's first: we have MPG, so the other per game stats can be made into per 48 minutes stats to facilitate comparison. Here's how Kevin's projections look:

Code:
MPG PTS REB AST STL BLK TO
Actual Pelton Actual Pelton Actual Pelton Actual Pelton Actual Pelton Actual Pelton Actual Pelton
Andrew Bogut 29 30 15.7 18.4 11.7 13.1 3.9 3.2 1.0 1.1 1.4 2.1 2.6 3.0
Chris Paul 36 28 21.5 13.9 6.8 5.1 10.4 9.4 3.0 2.7 0.1 0.0 3.1 3.3
Marvin Williams 25 28 16.5 15.3 9.4 11.0 1.5 1.5 1.2 1.9 0.6 0.9 2.0 2.7
Channing Frye 24 18 24.5 16.0 11.4 9.3 1.6 2.9 0.9 1.3 1.4 2.7 3.0 2.1
Jarrett Jack 20 20 15.8 14.2 4.8 5.3 6.6 6.2 1.2 2.2 0.1 0.0 3.1 3.8
Joey Graham 20 20 16.2 18.0 7.4 7.7 1.8 3.1 1.1 1.2 0.4 0.2 2.8 3.4
Raymond Felton 30 25 18.9 12.3 5.3 5.2 8.9 10.4 2.0 2.5 0.2 0.4 3.6 4.4
Danny Granger 23 12 16.0 19.6 10.5 11.2 2.5 3.6 1.6 2.8 1.7 2.4 2.2 3.2
Ike Diogu 15 18 22.5 19.2 10.6 10.1 1.3 1.9 0.7 0.5 1.4 2.4 3.6 3.7
Charlie Villanueva 29 25 21.4 15.7 10.6 11.9 1.8 2.3 1.2 1.0 1.3 2.7 2.0 3.5
Deron Williams 29 30 18.0 11.7 4.0 4.0 7.5 9.6 1.2 1.1 0.4 0.2 3.0 3.2
Sean May 17 26 22.7 20.9 13.1 15.1 2.6 3.0 2.0 1.7 1.4 1.5 4.0 3.7



So how good are those projections? First of all, you need a baseline, something to compare those projections to. Typically for a baseline, you use a naive prediction, something uncomplicated, that represents the bare minimum you would expect. Lets use league average as our baseline projection, and compare it to each of Kevin's projections using RMSE to compare accuracy.

Code:
RMSE
MPG PTS REB AST STL BLK TO
PELTON 5.4 11.6 17.5 20.5 23.7 24.2 22.2
LgAVG 7.2 9.8 17.4 22.0 24.0 24.5 22.6


Looks pretty even. Kevin's prjections of points scoring rates was worse than the league average baseline, but his projection for minutes per game was much better. If you do this again using per game stats, you'll see that the main difference between Kevin's projections and the baseline is his accuracy in MPG. Since MPG is much more variable than any one of the those other stats, accuracy here trumps accuracy elsewhere.

If you wanted to, you could improve the baseline by including additional information. The most relevant would probably be draft position. But then you'd end up with a projection model that probably looks a lot like the one Kevin already has. (I did something like this, and managed to outperform Kevin's projections in every other stats category except MPG. But, once again, Kevin's accuracy in that category trumps his being outperformed elsewhere.)

My purpose here wasn't really to talk about Kevin's projections, per se, but to note that talking about accuracy in projections is meaningless without some kind of baseline against which to measure.
_________________
ed
Back to top
View user's profile Send private message
Mark



Joined: 20 Aug 2005
Posts: 670


PostPosted: Wed May 17, 2006 11:33 pm Post subject: Reply with quote
The baseline comparison is a helpful addition of rigor. I thought the topic was worth a brief mention and I am glad it drew some additional comments and more depth. Sometimes hard to know what will catch interest. Some posts attempt to be comprehensive, this one was intended as a starter. My comments about the minutes being pretty accurate and the points low were brief but they end up still being main summary points I take away. We all make choices when to be brief and when to elaborate, as you choose to emphasize the baseline and preferred to go relatively brief on description of your own prediction results and method. More detail is often "worth it" but I accept your choice.

P.S. Daviswylie: I looked at a couple of your pages

the team stats pages indexed here

http://www.prism.gatech.edu/~gtg319y/0506NBA.html

show defensive PER calculations for players. Thanks. Might be of interest to others)


and a 2007 PER projection sheet

http://www.prism.gatech.edu/~gtg319y/06 ... ctions.txt

I dont know if you wish to discuss or if it is done. Was this by formula by age and and a general career curve and/or did you look at comparables and make refinements for each player? Looking at the gain/loss lists below the teams I have pretty major concerns when I see C Webber as a projected major gainer and Kobe Bryant as the biggest projected loser for next year but I will wait to hear more about method / willingness to discuss it before I go further.
Back to top
View user's profile Send private message
davis21wylie2121



Joined: 13 Oct 2005
Posts: 183
Location: Atlanta, GA

PostPosted: Thu May 18, 2006 12:07 pm Post subject: Reply with quote
Thanks, Mark... After taking a CS class, I finally learned how to post pages like that up on the internet!

Regarding the Defensive PER's: sadly, they are done using the same methodology as the B-R.com individual DRtgs, which means I had to estimate from team stats everything but blocks, steals, PF's, and DReb's. However, I hope to soon put up a more accurate defensive statistic that I created using some 82games numbers. I will post a topic on it before I do so, because I want to get input on the methodology -- currently, it ranks Vince Carter and Brent Barry as among the best defensive SG's in the NBA, and I'm almost positive that cross-matching with Jason Kidd and Bruce Bowen, respectively, is at fault, but I'm not sure how to fix it. More details to come...

The projections were built with the comparables from B-R.com (using the similarity scores), plus an age adjustment that goes something like this:


They were something I just did as a lark, so I wouldn't read a lot of accuracy into them. CWebb's massive gains come because his top three Age-32 comps were Karl Malone (Age 33 PER: 28.9), Charles Barkley (23.0), and Hakeem Olajuwon (25.5). Not really sure how to correct for that... Kobe's decline I can see more easily -- he's coming off a season he will likely never repeat, and he's crashed from a huge (26+) PER season down to about 23.3 in the past as well. I'd expect him to be around 23, not the 21.6 I've got projected. Most of the gainers/losers are your basic regression to the mean cases.

Thanks for exploring the site, by the way!
Back to top
View user's profile Send private message Visit poster's website
admin
Site Admin


Joined: 30 Dec 2004
Posts: 594
Location: Seattle

PostPosted: Fri May 19, 2006 12:37 am Post subject: Reply with quote
Ed Küpfer wrote:
Looks pretty even. Kevin's prjections of points scoring rates was worse than the league average baseline, but his projection for minutes per game was much better.

Hmm. Disheartening.

davis21wylie2121 wrote:
I would love to hear Kevin's methodology as well, as it will likely be of benefit to my "system".

I built a database of players who went straight from college to the NBA from ... I think it was like 2000-01 to 2003-04, something like that. I compared their performance during their last college year to their NBA rookie seasons, and applied that to the college crop, with an adjustment for schedule strength.

This has a little more and you can follow the links.

Charles



Joined: 16 May 2005
Posts: 48


PostPosted: Mon May 22, 2006 7:31 pm Post subject: PER vs OnCourt/OffCourt Reply with quote
Through the Piston's first twelve playoff games Ben Wallace is averaging just 4 points, 11 rebounds and 2 assists. His shooting percentges are very poor (FG= 38%, FT= 24%) Even Ben's blocks (1.5) and steals (1.2) are nothing special.

Naturally, the PER numbers at 82games reflect these realities. In fact, Ben has terrible ratings in both his Own PER and Opponent's PER.

Code:

------- PER ------
Own Opp Net
Billups 22.9 9.8 +13.1
Hamilton 18.2 13.6 +4.6
R.Wallace 19.5 15.0 +4.5
Prince 18.0 18.2 -0.2
B.Wallace 11.7 20.1 -8.4


On the other hand, the players most likely to be on the court when Ben is off have excellent Net PERs -- Rasheed (+4.5) and McDyess (+5.9.) Therefore, it would be natural to assume that Ben's presence on the court must be having a huge negative impact... Not so.

Code:
OnCourt/OffCourt

B.Wallace +23.0
Billups +17.1
Hamilton +8.5
Prince +7.4
R.Wallace -6.2

In the series that just ended the Pistons were +79 when Ben was on the court compared with -39 when he was off. He has now had a better plus/minus than Rasheed in nine consecutive games. He has also had a better plus/minus than Billups in eight of those nine games.

So, is Wallace performing badly - as the PER numbers suggest? Or do individual stats fail to capture Ben Wallace's true impact? I'm curious how people here view this contradiction.

Is Ben Wallace helping the Pistons or hurting them?
Back to top
View user's profile Send private message
mathayus



Joined: 15 Aug 2005
Posts: 180


PostPosted: Mon May 22, 2006 7:51 pm Post subject: Reply with quote
Well, first off let me say that +/- stats are notoriously high in deviation, and hence you can't really just look at a handful of games and conclude too much from them.

That said, yes, I believe +/- captures things PER doesn't, and that is why I always look at +/- stats when evaluating a player (not the only thing I use though).

Game 6 for example was an excellent example of how Ben's impact is not captured in traditional stats. He recorded 0 blocks, and yet his impact as a shotblocker saved the Pistons from losing (along with his impact on rebounding). Several times down the stretch LeBron would drive and get fouled by one of Ben's teammates...and he wouldn't even be slowed down by the foul. They might have been better off just letting him walk past. But after the foul, Ben would come in and either make the unofficial block, or alter LeBron's shot to the point where he didn't have a look at the basket. One time, LeBron was still going to be able to go up for strong dunk after the foul (who knows what kind of momentum that would have given the Cavs) and Ben got his hand on the ball and pushed LeBron back away from the basket.

I was so impressed by Ben in that game, without him the Pistons go home after that, and yet neither the box score or recaps talk at all about the impact he made. And I say this as someone who was cheering for the Cavs because I wanted to see some history.
Back to top
View user's profile Send private message
Mark



Joined: 20 Aug 2005
Posts: 807


PostPosted: Mon May 22, 2006 8:06 pm Post subject: Preferring a combination formula Reply with quote
By Roland Rating, Ben Wallace is having virtually no net impact- but it is no longer clear to me what formula is being used for that right now (it is not the simple combination of counterpart and team +/- on/off it started out as) and so I can't say if I agree with the weights or the Ben Wallace answer it produces. (Can anyone clarify?) Rather than seeing the two methods as separate and at variance and also as a possible alternative to whatever the existing mixed ingredient Roland Rating formula is, here is another try to blend counterpart and team +/-:

Method
1) credit players with say 70% of their individual offense when on court (or whatever % you felt was appropriate as the individual credit share for that scoring on average) and distribute the remaining 30% to the 5 on the court in proportion to their individual offensive winshares compared to the total offensive winshares on the court at that time;

2) take them and then subtract from it 70% of points allowed charged to their position while on court and then take the remaining 30% of points allowed and assign them to the players on the court inversely based on their relative amount of individual defensive winshares (perhaps from regular season to set it) compared to the total defensive winshares on the court at that time and subtract them too from the offensive number the player earned from his individual and team scores;

3) and then the net result would be a net counterpart and on/off +/- combination s.

I think is worthwhile to have something in this spirit, whatever the details.
I see no definition of current Roland rating on the site right now, perhaps something like the adjustment I describe is being done? Fair value salary may be superceding it for next year but no detailed explanation of that formula is available yet either as it is apparently still under refinement. That is a valuable way to look at players, but I still think some form of combination rating of these two major player evaluations (counterpart and team +/-) would be useful to see before salary gets brought into it so it can be used as a reference on overall player impact in terms of game score.

My new twist is using the offensive and defensive winshares as a tool in that formula and eliminating the impact on off the court numbers in a player's score. Giving key players some of the credit for other players numbers get done in discussions- this is a crude firstcut method to try to capture that impulse. Previously I proposed changing the original Roland Rating formula to give some more weight to +/- on and some less to +/- off (a 3-1 ratio instead of equal) but I might prefer to do away with "off" entirely. True +/- adjusted would be preferrable of course but this isnt available in public / current time yet for all players.

Ben's role is not as a scorer so he will look bad on counterpart scoring. His role is defense and shows up in team +/- but it isnt really fair to see that value and call the full amount a measure of his team impact (in his case almost all on defense) because of course that number is the product of all the actions of 5 guys on offense and defense and not just him.
Back to top
View user's profile Send private message
Mark



Joined: 20 Aug 2005
Posts: 807


PostPosted: Mon May 22, 2006 10:10 pm Post subject: Reply with quote
I thought it might also be interesting to share main Piston playoffs to regular season value/performance by Roland Ratings. Obviously playoffs affected by matchup. Cleveland stronger in frontcourt than backcourt. Milwaukee was generally stronger perimeter but also got strong net counter production from center but not PF. Cleveland series was longer and thus had more weight.

Roland Rating

Playoffs Regular Net
Billups 14 11.9 2.1
Prince 1 7.7 -6.7
B.Wallace -0.6 6.9 -7.5
R.Wallace 1.3 6.1 -4.8
Hamilton 5.4 5.4 0
McDyess 1.9 -3.5 5.4

Ben doesnt seem to be near the impact he was in the regular season (where Roland Rating had him as a strong positive) so far in the playoffs but now comes the Shaq test and whoever is after him.
Back to top
View user's profile Send private message
Pinot



Joined: 23 May 2006
Posts: 19


PostPosted: Tue May 23, 2006 1:34 pm Post subject: Reply with quote
How much weight can you really give the Roland rating though? It's not like a player goes one on one in an isolated match up every time. I give Roland ratings about as much (a little more) weight relative to player quality as I do to Wins for pitchers in baseball. There are so many other factors at work which are not, maybe cannot, be accurately factored in. I followed them closely this year assuming they'd improve greatly as the season went on, but was not impressed with the results compared to how accurately PER or WS calculate what they intend to.
Back to top
View user's profile Send private message AIM Address
davis21wylie2121



Joined: 13 Oct 2005
Posts: 327
Location: Atlanta, GA

PostPosted: Tue May 23, 2006 1:39 pm Post subject: Reply with quote
Well, PER itself is a part of the Roland Rating. Also, the +/- component is designed to filter out the perception many stats give that basketball is just a series of one-on-one matchups.
Back to top
View user's profile Send private message
Pinot



Joined: 23 May 2006
Posts: 19


PostPosted: Tue May 23, 2006 1:45 pm Post subject: Reply with quote
davis21wylie2121 wrote:
Well, PER itself is a part of the Roland Rating.


Of course. Do you know why do 82games PER values differ from Hollinger PER values so greatly, i.e. Joe Johnson 19.2 vs. 17.95 (JH)?


Quote:
Also, the +/- component is designed to filter out the perception many stats give that basketball is just a series of one-on-one matchups.


I'm really not sure what that can mean in all practicality.
Back to top
View user's profile Send private message AIM Address
Mark



Joined: 20 Aug 2005
Posts: 807


PostPosted: Tue May 23, 2006 1:48 pm Post subject: Reply with quote
Hi, welcome.

Care to share some more detail on your evaluation of the relative performance of the different statistical measures?

What basis should be used to evaluate their results?

Did the team with the highest team PER (sum of starting five or all the individual PERs weighted by minutes) win the most games? Will they win the championship? How well did team summary of player win % do? Did the team with the highest average roland Rating do worse than these? How about compared to good old point differential? Or a new weighted team 4-factor evaluation method? I'd like to the results.

Has anyone calculated these for this year or for multi-years to justify higher regard for one over the other? Or willing to split up the work to accomplish this?

Would a weighted combination of these do better explaining the past or predicting the future? It could for any one year but are there weights stable and successful over time?

Will we push for this?
Back to top
View user's profile Send private message
alphamale



Joined: 03 May 2006
Posts: 67


PostPosted: Wed May 24, 2006 12:27 pm Post subject: Reply with quote
Did you guys see the great job that ben did on both Shaq and Zo last night?

He was the lone bright spot, every other piston stunk.
Back to top
View user's profile Send private message
Mark



Joined: 20 Aug 2005
Posts: 807


PostPosted: Wed May 24, 2006 1:04 pm Post subject: Reply with quote
Looking at the top 4 teams remaining...

record and expected wins both show the Pistons superior to Miami but the gap in expected wins was only 2/3 thirds the gap in actual wins. Average the PW% of the top 6 guys weighted by minutes and the Pistons are better than Miami and at about the same level of superiority as shown by expected wins. But the minutes weighted average PER of the top 6 guys is almost a dead heat between them. It is mostly offense and so misses much of Detroit's better defense. The simple sum of top 6 Roland Ratings however gives the edge to Miami.

In the Mavs / Suns series the Mavs are superior on all five measures, though some narrowly. The expected win gap is only half the size of actual. PW% is nearly even, while the Mavs have a modest lead on PER, but a large one on Roland rating.

If the Pistons and Mavs meet the Pistons are superior on four but the Mavs have a higher average PEr for top 6- again expected becuase of the power of their offense. Normalized and a simple sum of these five measures the Pistons look the best.

Not that this will "prove' anything but because of the separation in how the stats rate the two teams in each series we can casually look at which stats serves better as a predictor.

Really these quality ratings have to be recognized as general quality ratings and we must remember the powerful importance of matchup. While a few of right summary stats can do surprisingly well to predict winners, the game is actually determined by the interaction between dozens if not hundreds of function curves and some strengths and some weaknesses will end up larger than expected and others smaller.
That makes the game exciting.

Trying to capture a little about matchup and playoff basketball, I previously suggested that the strongest contenders for the title would be top 7 on defense first and then top 2 among them on offense or an outside shot for the next team on the balance of the two. By this system and using the playoff to date stats raw, unadjusted for pace there is some concern whether Detroit has enough offense. This may have been heavily influence by playing the Bucks and the Cavs and regular season the Pistons had enough offensive efficeincy so I tend to assume they will be ok. The Mavs appear to satisfy the desired defense/offense quality pattern pretty well too. Miami comes in third. Phoenix looks last.

Unless the era of defense mattering more relatively in the playoffs has changed and made this formula outdated. 4 of the 6 top regular season offensive efficency teams made the final 4. Only Detroit did of the top 6 on defense. Last year it was only 2 of the top 6 offensive efficiency teams and 3 of the top 6 defensive teams in the final 4. Random or a meaningful shift?
Back to top
View user's profile Send private message
davis21wylie2121



Joined: 13 Oct 2005
Posts: 327
Location: Atlanta, GA

PostPosted: Wed May 24, 2006 3:29 pm Post subject: Reply with quote
Pinot wrote:

Of course. Do you know why do 82games PER values differ from Hollinger PER values so greatly, i.e. Joe Johnson 19.2 vs. 17.95 (JH)?


I've wondered that myself. I assume you got 17.95 from ESPN Insider, and I have JJ at 17.89 (meaning that John hasn't tweaked the formula significantly since last year), so I don't know why 82games' PER's are off by so much. My guess is that 82games uses their unofficial scorekeeping data (which also includes the on-off and counterpart numbers, jump shot %'s, etc.) to calculate PER, and that since the scorers are volunteers (at least they were a couple of years ago, correct me if I'm wrong), mistakes happen.
Pinot wrote:

davis21wylie2121 wrote:
Also, the +/- component is designed to filter out the perception many stats give that basketball is just a series of one-on-one matchups.


I'm really not sure what that can mean in all practicality.

Some people argue that traditional stats like points, blocks, rebounds, etc. (and the stats built from them, like PER) fail to capture the essence and flow of basketball, instead treating the game like a continual series of five one-on-one matchups. +/- stats were designed to try and take a new tack on player performance evaluation, in case some players do things that don't show up in the traditional statistics. For example, it is often said that Steve Nash does things to help the Suns that don't show up in stats like PER, so +/- numbers may offer an alternative measure to capture those "intangibles" (although in Nash's case, he still doesn't come off looking like the MVP in the +/- either).
Back to top
View user's profile Send private message
Mark



Joined: 20 Aug 2005
Posts: 807


PostPosted: Wed May 24, 2006 4:07 pm Post subject: Reply with quote
Isolating 2 of the 4 factors of offense (eFG% and and FT/FG) in the form of points per shot, regular season Miami 1 and Dallas 2 were best while Phoenix was 12th Detroit 23rd.

How these teams rate on the other two factors is also important. For the Suns and the Pistons low turnovers helps, while Miami and Dallas are closer to average on this. Dallas is aided by strong offensive rebouding, while Miami is not as much as you might think- only 17th best.

The final 8 teams in the playoffs this year included 7 of the 8 best playoff terams on points per shot with NJ the only exception. Further evidence of a rise in importance of superior ability to score?


On defense points per shot Detroit is 2nd, Miami 6th, Phoenix 7th, Dallas only 12th.

How good Dallas' defense really is will be heavily scrutinized.
Regular season they were 10th on eFG% allowed, 11th on turnovers forced, 2nd best at minizing opponent offensive rebounds, but 3rd highest in foul shots permitted. In playoffs, among just 16, they are 10th on eFG%, 7th on turnovers forced, 2nd lowest /best on off reb allowed, but 2nd highest /worst in foul shots permitted. The patterns are similar, but of course this is not adjusted for quality on the characteristics of the opponents.
Back to top
View user's profile Send private message
Mark



Joined: 20 Aug 2005
Posts: 807


PostPosted: Wed May 24, 2006 11:28 pm Post subject: Reply with quote
Dallas in general doesnt care much about pace, on average they win about the same against high. average and low pace teams.

But Suns is somewhat worse at slow pace.

In 5 games this year the Suns have average about 114 pts against Dallas.
Dallas won the slower 2 and lost the faster 3.
I'd slow it down. If they can. Maybe they can't control pace against the Suns.
Back to top
View user's profile Send private message
Charles



Joined: 16 May 2005
Posts: 48


PostPosted: Fri May 26, 2006 12:26 am Post subject: Reply with quote
mathayus wrote:
Well, first off let me say that +/- stats are notoriously high in deviation, and hence you can't really just look at a handful of games and conclude too much from them.


Fair enough. Let's take the Piston's example a little further.

Detroit has had the same starting lineup for three seasons now which gives us six sets of data -- three regular seasons and three playoff seasons. In those six "seasons" Chauncey Billups has had the highest Net PER all six times and Ben Wallace has had the lowest Net PER all six times. The differences are not small.

Code:
2006 Playoffs
Billups +13.1
R.Wallace +4.5
Hamilton +4.6
Prince -0.2
B.Wallace -8.4

2006 Regular Season
Billups +13.6
Hamilton +4.7
Prince +3.7
R.Wallace +2.8
B.Wallace +1.4

2005 Playoffs
Billups +7.7
Hamilton +5.3
R.Wallace +3.5
Prince +2.6
B.Wallace +2.4

2005 Regular Season
Billups +7.3
Prince +3.3
Hamilton +3.0
R.Wallace +2.9
B.Wallace +2.9

2004 Regular Season
Billups +7.3
R.Wallace +7.0
Prince +4.0
Hamilton +2.5
B.Wallace +1.7

2004 Playoffs
Billups +7.9
Prince +7.4
R.Wallace +4.6
Hamilton +0.1
B.Wallace -4.1

However, despite Billup’s overwhelmingly superiority in PER, Ben Wallace has had a larger OnCourt/OffCourt impact two of the three regular season and two of the three playoff seasons.

This constitutes a major difference between the two approaches.

So, how do you reconcile this disparity? Do you believe Ben Wallace has as much or more impact than Billups? If so, doesn’t that make the widespread use of PER in player assessment somewhat of a sham?
Back to top
View user's profile Send private message
Mark



Joined: 20 Aug 2005
Posts: 807


PostPosted: Fri May 26, 2006 12:10 pm Post subject: Reply with quote
Until there is a meta-rating for combining counterpart data with team +/- and perhaps other things the best stat for comparing player impact may be winshares or EWins. Assuming the critieria for assigning a win is even between offensive and defensive wins. Not alleging they are not, just putting out that caution/question and asking if the authors feel that their system is fully adjusted to be "fair" to ewach side of the ball contrubtuions. By allowing "wins" to be earned many ways they may capture the diversity of win significant contributions I was trying to capture with meta-method ratings.

Billups 47 winshares, B Wallace 32, R Wallace 28, R Hamilton 24, Prince 22, MCDyess 15, M Evans 10 seems like a pretty good value of contribution distribution to me, balanced for offense/defense, one on one counterpart and help/team defense. B Wallace might seem a little high but would the Pistons without him and instead with a journeyman replacement be elite? I dont think so. Detroit is well-designed with 4 capable scorers, 5 above average defenders on starting five. That is pretty close to as good as you can get. With 4 capable scorers, Ben not scoring is manageable and doesnt have to seen as a big negative at the team level. But if he can get 10 points, they are even stronger as in last game win.

Playing with two weaker than average scoring positions may be too much these days. San Antonio regular season did that with Bowen and a true center Mohammed or Nesterovic. But against Dallas San Antonio went Duncan at center 80% of the time to get Finley Horry Barry in the game more for their hoped for scoring. Cleveland certainly didnt get enough points from other positions.

mathayus



Joined: 15 Aug 2005
Posts: 194


PostPosted: Fri May 26, 2006 2:50 pm Post subject: Reply with quote
Charles wrote:

Fair enough. Let's take the Piston's example a little further.

Detroit has had the same starting lineup for three seasons now which gives us six sets of data -- three regular seasons and three playoff seasons. In those six "seasons" Chauncey Billups has had the highest Net PER all six times and Ben Wallace has had the lowest Net PER all six times. The differences are not small.

Code:
2006 Playoffs
Billups +13.1
R.Wallace +4.5
Hamilton +4.6
Prince -0.2
B.Wallace -8.4

2006 Regular Season
Billups +13.6
Hamilton +4.7
Prince +3.7
R.Wallace +2.8
B.Wallace +1.4

2005 Playoffs
Billups +7.7
Hamilton +5.3
R.Wallace +3.5
Prince +2.6
B.Wallace +2.4

2005 Regular Season
Billups +7.3
Prince +3.3
Hamilton +3.0
R.Wallace +2.9
B.Wallace +2.9

2004 Regular Season
Billups +7.3
R.Wallace +7.0
Prince +4.0
Hamilton +2.5
B.Wallace +1.7

2004 Playoffs
Billups +7.9
Prince +7.4
R.Wallace +4.6
Hamilton +0.1
B.Wallace -4.1

However, despite Billup’s overwhelmingly superiority in PER, Ben Wallace has had a larger OnCourt/OffCourt impact two of the three regular season and two of the three playoff seasons.

This constitutes a major difference between the two approaches.

So, how do you reconcile this disparity? Do you believe Ben Wallace has as much or more impact than Billups? If so, doesn’t that make the widespread use of PER in player assessment somewhat of a sham?


Sham? Well I wouldn't use such terms, and I wouldn't claim that +/- is a panacea. I absolutely do look at PER and other such rating system and value them. But yes I'm in the +/- camp. My background: I had my own player rating system based around traditional stats a few years ago. I won't claim it was good as PER, but the core principles were the same. Eventually I came to the conclusion, that there were important factors that you could get from traditional stats.

Clearest example for me was Nash. I saw the impact he was having, and I saw that with my system he wasn't rated as being anywhere near the MVP of his team. For a while, I tried to rationalize it away, saying that people were getting overexcited about Nash. But eventually I had to conceed that my system made it essentially impossible for a pass-first point guard to look as valuable as big men. At that point, you either have to believe that the position known for being the coach on the floor, and called floor general or quarterback, simply isn't that valuable, or you have to conclude that your methodology is fatally flawed. Seeing as how however valuable a big man is, a team of 5 all-time great big men would get killed by a balanced team, hence meaning that at least some little guy was out playing big men, I eventually came to the conclusion that the problem was with my stat and not the world.

I like +/- because it has no bias toward offense or toward the guy with the ball. Simply measure how the team does with you. Now, there's a whole lot of things you need to take into consideration to really eliminate the biases of +/-, and even then, the deviation is significant, so it's definitely not perfect. But when I see a guy consistently have strong +/- numbers (not some scrub who plays limited minutes, and honestly not just a guy who has good +/- stats one year), then I don't care about his PER if his PER is weak. He's obviously doing something that PER doesn't account for.
Back to top
View user's profile Send private message
Mark



Joined: 20 Aug 2005
Posts: 807


PostPosted: Mon May 29, 2006 11:06 pm Post subject: Reply with quote
Regarding whether Detroit has enough offense, so far this series it has been an issue, averaging not much more than 85-86 points. And giving up 91-92.

Don't have that strong post up game (and Shaq's there).
After hitting over 40% from 3 pt land in first two series they slipped to just a little over 30% in first three games. Rasheed trying it maybe too much at 5-18. Hamilton getting too few looks at just 3 for 10 for series. Wade's defense? Conversely Wade may face Detroit's weakest defender and he is shooting 70+%FG. Have they tried a switch of Billups much on Wade? I assume A Walker on the court prevents switching Prince to Wade. Lindsey Hunter's name came up but that would only be for a few more minutes, not the bulk of the remedy. Maurice Evans sometimes plays good D and his offense has been flowing.
Back to top
View user's profile Send private message
Gary C



Joined: 14 Apr 2006
Posts: 69


PostPosted: Wed May 31, 2006 10:52 am Post subject: Reply with quote
Quote:
Playing with two weaker than average scoring positions may be too much these days. San Antonio regular season did that with Bowen and a true center Mohammed or Nesterovic. But against Dallas San Antonio went Duncan at center 80% of the time to get Finley Horry Barry in the game more for their hoped for scoring.


SA did not *choose* to go to Duncan with no other center, Dallas forced them to with matchups. And something fascinating happened: SA's defnese suddenly looked very ordinary, giving up a bunch of layups and dunks. I'm sure some of that was that Dallas is a good offensive team, but I take out of the whole thing that SA's defensive success in this era has been at least as much about Pop's system as it has about the players plugged into that system.
Back to top
View user's profile Send private message
Mark



Joined: 20 Aug 2005
Posts: 807


PostPosted: Wed May 31, 2006 11:50 am Post subject: Reply with quote
Yes Dallas forced San Antonio to go Duncan at center a lot. Dallas changed San Antonio and not the other way around. Dampier and Diop didnt handle Duncan that well but it allowed Dirk to focus on offense. In past playoffs when asked to guard the best big opponent and score he didnt seem to be able to fully handle both jobs.

Dallas' roster has offensive and defensive strength. Will defense make the difference against Phoenix?

If San Antonio and Detroit, two of the strongest defensive bias teams that use a really defenisve biased player in the starting lineup go out before the finals this year that is a pretty important event to think about.

Bowen- counterpart playoffs he isnt shown as getting it done there with over 50% eFG% allowed for an average of 22 pts per 48. Flawed , yes ... Ok, Team FG% allowed actally higher with him on court than off. Points allowed only down 1 while offense was down 16 points. Last year he was much better counterpart but far worse on team defense measures on/off. Don't really have a handle on him and his stats. But not sure I'd stay with him for as many minutes in the future, at least in playoffs. I'd also try to get a athletic PF/C who could score more than Mohammed or Nesterovic. Tough task. Robertas Javtokas appears to be lined up to come over now but I dont know how good he translates here.

In the free agent class I think Al Harrington at PF / Duncan to center could also be interesting option to add. In a more flexible world Harrington is three position capable depending on matchup and San Antonio could use that versatility. Maybe Parker for Harrington and other stuff. Or Harrington for return of Mohammed (who might in his return fit what Atlanta needs now. Other trade possibilities? Brad Miller might be possible or Z from the Cavs)

Ben Wallace seemed a lock to get a big new contract but I could now see both sides thinking more seriously about options. (How would he look on the Suns, Kings or Rockets? I think pretty good.)

Detroit regular season with 5th best defensive efficiency but looking at the partsmaybe it isnt as impressive: 10th on eFG% allowed, 10th of turnovers forced 25th on offensive rebounds 1st on FT/FG allowed. But Miami is getting more free throws than any team in the playoffs and get 28.5 a game against Detroit. Take away that strength and the rest of the defense isnt enough, especially when the offense isnt firing that well.

You could argue combined strength, team offensive/defensive balance is in with Dallas and Miami. That isnt new, the Bulls and the Shaq/Kobe/Jackson Lakers were more that than recent San Antonio and Detroit. With a superstar and a fair number of two way capable players and a mix of inside/outside. Detroit and Phoenix strived for that balance but may prove to have not quite enough of one to succeed all the way to championship. Detroit with Flip seemed like it would work this year. Eddie Johnson may be right that Flip didnt use his bench enough this year and perhaps showcased that better offense too much. Miami appears rebuilt right to make it real tough for them. Miami's lowest in the league off. reb allowed seems to prevent that from becoming a series turner for Detroit as it could be against other teams (it was one major factor for success in 2003-4 finals)
Back to top
View user's profile Send private message
Charles



Joined: 16 May 2005
Posts: 134


PostPosted: Thu Jun 01, 2006 1:23 pm Post subject: Reply with quote
Mark wrote:
Until there is a meta-rating for combining counterpart data with team +/- and perhaps other things the best stat for comparing player impact may be winshares or EWins...

I am not familiar with winshares or ewins, but one problem with aggregated ratings is that -- although they may give you a reasonable quick-and-dirty estimate for “average” players -- they are more likely to break down when applied to players who fall at the high end of the performance spectrum. Shouldn’t that be expected when you apply a linear model to data that may well be non-linear?

Combining ratings is an interesting idea, but I worry that both the statistical measurement and the On/Off measurement be tainted.

I am not promoting plus/minus as a replacement for individual statistics. However, I think the Pistons offer a sharp example of just how cautiously formulas like PER should be treated in player assessment. This season Chauncey Billups received a significant vote for MVP despite having the second worst OnCourt/OffCourt impact of the starters on his own team.

Code:
OnCourt/OffCourt: 2006 regular season
Prince +16.6
B.Wallace +14.2
R.Wallace +13.7
Billups +10.2
Hamilton +6.0


These numbers are unadjusted, but Billups low impact was not because he was being compared to good backups. To the contrary, Billups backups had very low PERs.
Code:

--- NetPER ---
Own Backups
Billups +13.6 -3.2
B.Wallace +1.4 +4.7


Keep in mind that, although there may be considerable variance in On/Off data, Billups has never had the best or even the second best On/Off differential among the Piston’s starters in any regular season or any playoffs.

Which brings me back to the original question which is: What do you do when the PER data consistently contradicts the On/Off data? In my case, if the On/Off is long term, consistent and has a rational explanation, I treat the PER (or other ratings based on individual data) with extreme suspicion.
Back to top
View user's profile Send private message
Mark



Joined: 20 Aug 2005
Posts: 807


PostPosted: Thu Jun 01, 2006 1:34 pm Post subject: Reply with quote
".... but one problem with aggregated ratings is that -- although they may give you a reasonable quick-and-dirty estimate for “average” players -- they are more likely to break down when applied to players who fall at the high end of the performance spectrum. Shouldn’t that be expected when you apply a linear model to data that may well be non-linear?"

I proposed a linear type meta-rating as a starting point but there is no reason that is has to stay linear. Use the highest and best statistical approach.

"Combining ratings is an interesting idea, but I worry that both the statistical measurement and the On/Off measurement be tainted."

PER, team on/off, counterpart, etc. they measure what they measure, the way they measure it and each has concerns and arent flawless. They dont capture all of reality or all of reality that has meaning (as if we know for sure everything to look for). But together they may capture more and by "triangulation" pin down a location more exactly than any one line would.


"Which brings me back to the original question which is: What do you do when the PER data consistently contradicts the On/Off data? In my case, if the On/Off is long term, consistent and has a rational explanation, I treat the PER (or other ratings based on individual data) with extreme suspicion."

I understand that, but I'd also look at how good the situation when that player gets to be on the court is. Treat both as useful but requiring caution in interpretation. And being better undestood in the context of 3. 6 or more other measures.

We all want adjusted on/off data, a few folks know how to do it right but it is currently treated as proprietary (understable, not meant as criticism, just lamentable). Dan gave us an interesting glimpse. Someone, perhaps Ed or perhaps 82games eventually, could open this door and provide easier access to that data on a current and on-going basis for the rest of us, someday. That would an outstanding advance.

I wonder if a modest hack at adjusted +/- off could be achieved using known data for time on court together player pairs and those +/- ratings as well as the overall ones but havent tried this very rough approach as it is obviously a weak method. Add in gameflow level data and it would be better. But then you are half way there so why not crunch the data properly to get statistically respectable adjusted data? Maybe I'll try something for a team or a few teams if we are still in an unadjusted +/- world much longer.
Back to top
View user's profile Send private message
Jakedfw



Joined: 27 May 2006
Posts: 31


PostPosted: Mon Jun 05, 2006 11:52 pm Post subject: B Wallace Reply with quote
Quote:
So, how do you reconcile this disparity? Do you believe Ben Wallace has as much or more impact than Billups? If so, doesn’t that make the widespread use of PER in player assessment somewhat of a sham?


Not a statistician, so don't flay me too much...

But could this mean that things like rebounds aren't weighted highly enough in the PER calculations?
Back to top
View user's profile Send private message Send e-mail
deepak



Joined: 26 Apr 2006
Posts: 664


PostPosted: Tue Jun 06, 2006 12:47 am Post subject: Re: B Wallace Reply with quote
Jakedfw wrote:
Quote:
So, how do you reconcile this disparity? Do you believe Ben Wallace has as much or more impact than Billups? If so, doesn’t that make the widespread use of PER in player assessment somewhat of a sham?


Not a statistician, so don't flay me too much...

But could this mean that things like rebounds aren't weighted highly enough in the PER calculations?


I wouldn't say that. Actually, looking at the OnCourt/OffCourt data for Ben Wallace, the Piston's rebounding was actually worse with him on the court, on both ends. And the shot blocking is about the same.

The difference is the Pistons force more turnovers with him on the court, they foul much less (give up less free throws), and the opponent shoots worse from the field. Ben Wallace does things that aren't captured in a box score (and thus, not captured by PER) which helps the Pistons as a team perform better in those areas defensively.
Back to top
View user's profile Send private message
ziller



Joined: 30 Jun 2005
Posts: 126
Location: Sac Metro

PostPosted: Tue Jun 06, 2006 5:45 pm Post subject: Reply with quote
Quote:
We all want adjusted on/off data, a few folks know how to do it right but it is currently treated as proprietary (understable, not meant as criticism, just lamentable). Dan gave us an interesting glimpse. Someone, perhaps Ed or perhaps 82games eventually, could open this door and provide easier access to that data on a current and on-going basis for the rest of us, someday. That would an outstanding advance.


Available adjusted +/- is the biggest need currently in abermetrics, in my opinion.

Obviously, strides in relativity and comparability are the pieces missing from all these readily available measures. Shooting efficiency, per-possession ratings, +/- figures - these could all be dramatically improved if they could be stripped off influence by teammates, opponents, and pace. (Not unlike ERA+ and OPS+ in baseball.) Of course, adjusting these is the (often) mind-numbing quarry.

On the topic of this whole debate, I'm thoroughly convinced no magic bullet number will ever exist for basketball. It's been said before, but it bears repeating - basketball is a game of roles. These statistics would probably do better by finding out how to measure players' ability in those roles - scorer, rebounder, passer, ball-handler, defender, etc. And that will aid team-building.

(After a quick re-read, I don't think any of that makes sense. I'm hitting 'submit' with that disclaimer, nonetheless.)
_________________
SactownRoyalty.com
tziller@gmail.com

deepak



Joined: 26 Apr 2006
Posts: 664


PostPosted: Sat Jun 10, 2006 2:25 pm Post subject: Berri: NBA's Secret Superstars (NYT) Reply with quote
http://www.nytimes.com/2006/06/10/opini ... ref=slogin

Thoughts?

Also, could someone who's knowledgable about his methods also explain how/why his results differ from individual Net Wins via Offensive and Defensive Ratings. Are they trying to answer the same question, or are they actually describing two different concepts?

Thanks.
Back to top
View user's profile Send private message
Dan Rosenbaum



Joined: 03 Jan 2005
Posts: 541
Location: Greensboro, North Carolina

PostPosted: Sat Jun 10, 2006 11:16 pm Post subject: Reply with quote
[Let me rewrite some of this to add clarity.]

I have not yet picked up the book, but their "simple" formula is the following.

Win Score = PTS + REB + STL - FGA - TO + 0.5*(AST + BLK - FTA - PF)

They also have a simpler formula that they use in several of their academic papers.

Old Win Score = PTS + REB + STL - FGA - TO - 0.5*FTA

As far as I can tell, all of the stats are per-minute (or per-48 minute or per-40 minute) measures. I don't know if they are pace-adjusted.

But Win Score is a simplified version of Wins Produced. And here is the difference between the two.

http://dberri.wordpress.com/2006/05/21/ ... erformance

Quote:
To get at Wins Produced you have to use the exact values, and make a few adjustments – such as adjusting for position played – which we note in the book. Still, Win Score is sufficient to give you a quick snapshot of a player’s performance. And it is especially useful if you wish to know if a player is playing better or worse than he did before.

By the "exact values" they mean they are using multipliers different from 1, -1, 0.5, and -0.5. I would imagine this helps, but in their academic papers with their Old Win Score metric, they suggested that using the exact regression parameters did not matter much. Also they write the following.

Quote:
To keep it simple, one can argue that each of these latter stats is worth ½ a point, rebound, etc… Now the ½ value is not exact, but using ½ keeps it simple and we find one gains very little using the exact relative value. In other words, player rankings do not change very much when you use the exact values.

Making adjustments by position could be more helpful, but I would be suprised if it helped a lot. And by arguing that Win Score does a pretty good job, they in essence argue something similar.

So what have I done with these two metrics?

I have related these two metrics to my adjusted plus/minus ratings, which in theory capture practically everything of value that a player does while he is on the floor. It may not be a perfect metric to measure future productivity, but to gauge the usefulness of metrics like these, it is probably ideal.

So I run a regression with my adjusted plus/minus measure as the dependent variable and various metrics as the independent variable. I report the R-squared values from these regressions. R-squared measures the fraction of the total variation explained by the regression. A value of 0 means that the regression explained nothing and a value of 1 means the regression explained everything. Here are my results.

Win Score - R-squared = 0.2009
Old Win Score - R-squared = 0.0843
NBA Efficiency - R-squared = 0.2209
Putting in each of the variables in the Berri1 formula in the regression model separately - Rsquared = 0.3740
My statistical plus/minus rating - R-squared > 0.45

This may not be the best way to gauge which metric is best, but I cannot think of a better way. And so my results suggest that the Win Score is a little worse than the NBA efficiency metric. (It is a big improvement, however, over the metric they use in their academic papers. That metric explains almost nothing.)

So, according to these results, there is not a lot to recommend the Win Score as a big improvement over the metrics that folks here use.

Now again Wins Produced is a different animal but from what I can gather, I suspect it might be a little better than NBA efficiency but that's about it. If it was A LOT better than Win Score, I would think the authors would be hesitant to ever use Win Score.

But I suspect it is still worthwhile to read the book. Many of the points they make are still valid - even if they are not using an ideal metric. Also, they touch on a lot of topics that get little play here (such as how salaries relate to stats), and I suspect folks might find that interesting.

Last edited by Dan Rosenbaum on Sun Jun 11, 2006 5:49 pm; edited 4 times in total
Back to top
View user's profile Send private message Send e-mail Visit poster's website Yahoo Messenger
edijorj



Joined: 21 Apr 2006
Posts: 30


PostPosted: Sun Jun 11, 2006 12:47 am Post subject: Reply with quote
Dan Rosenbaum,

I havn't read the book either (and I probably won't), but I think Berri1 is what they call Win Score, not Wins Produced. Win Score is supposedly similar to Wins Produced if player position is accounted for. (Berri apparently realizes that his metric is biased against guards.) So it would make more sense if you added some dummy variables for player position.

It seems like the problem with Berri's analysis is that he believes that his metric is good if it explain team wins. To give an extreme example, the equation Y=1.78*FG-1.49*FGA+1.78*Th-0.57*ThA+0.51*FT-0.41*FTA+1.46*Orb+1.41*Drb+0.06*ast+1.76*stl+0.13*blk-1.52*to+0.13*pf explains team wins very well, but is clearly wrong for evaluating a player.
Back to top
View user's profile Send private message
deepak



Joined: 26 Apr 2006
Posts: 664


PostPosted: Sun Jun 11, 2006 10:22 am Post subject: Reply with quote
Dan Rosenbaum wrote:
I have not yet picked up the book, but I think their formula is the following.

Berri1 = PTS + REB + STL - FGA - TO + 0.5*(AST + BLK - FTA - PF)

They also have a simpler formula that they use in several of their academic papers.

Berri2 = PTS + REB + STL - FGA - TO - 0.5*FTA

As far as I can tell, all of the stats are per-minute (or per-48 minute or per-40 minute) measures. I don't know if they are pace-adjusted.


Is their metric really that simple? For all the press the book is getting, somehow I thought it would be a tad more sophisticated than that.

They describe their approach in this way:
Quote:

The approach we took in The Wages of Wins is simply to utilize regression analysis – a common technique in economics – to determine the relative impact of each statistics on team wins. We had three objectives in constructing our model. Ultimately we wanted a measure that was simple, complete, and accurate. In the end, we think each of these objectives was met. The Wins Produced model is not hard to understand, it incorporates each of the statistics tracked for individual players, and it connects accurately to team wins.



From here. they claim:
Quote:

In particular, they argue, traditional talent evaluation over-rates the importance of points scored, and under-rates the importance of turnovers, rebounds and scoring percentage. Wages of Wins also obliterates the so-called NBA Efficiency rating, which is the official algorithm used by the league and many basketball experts to rank the statistical performance of players. The Efficiency rating, they argue, makes the same error. It dramatically over-rewards players who take lots and lots of shots.”

...

“Okay: part two. Is the Wages of Wins algorithm an improvement over the things like the NBA Efficiency system? To make the case for their system, the authors “fit” their algorithm to the real world. For the 2003-04 season, they add up the number of wins predicted by their algorithm for every player on every team, and compare that number to the team’s actual win total. Their average error? 1.67 wins. In other words, if you give them the statistics for every player on a given team, they can tell you how many wins that team got that season, with a margin of error under two wins. That’s pretty good.”


So, apparently, they believe their metric is much more accurate than NBA efficiency.

Considering that PER has been become much more publicly well known of late (through ESPN, 82games, here, etc.), I wonder why they don't make any mention of it?
Back to top
View user's profile Send private message
Kevin Pelton
Site Admin


Joined: 30 Dec 2004
Posts: 978
Location: Seattle

PostPosted: Sun Jun 11, 2006 11:48 am Post subject: Reply with quote
deepak_e wrote:
Is their metric really that simple? For all the press the book is getting, somehow I thought it would be a tad more sophisticated than that.

No. As edijorj said in the previous post, that is Win Score, as opposed to the more detailed Wins Produced.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Mark



Joined: 20 Aug 2005
Posts: 807


PostPosted: Sun Jun 11, 2006 11:51 am Post subject: Reply with quote
We can't compute error on individual player level but it seems like there could be higher levels of player error that cancel each other out. Error reported at team level really just measures error for formula for combined team level stats doesnt it? Several other make this same case in this thread http://gladwell.typepad.com/gladwellcom ... _by_t.html
This would apply to the simple version of Berri, would it to the regression-based version too?

I noticed that 3 of the 10 most underrated were Memphis Grizzlies. http://gladwell.typepad.com/gladwellcom ... _by_t.html
Only one brought in by Jerry West but the presence of the other two may have been some attraction to him and he did give them 6 year extensions. But they can't get out of the first round even with Gasol . Even win a playoff game over last three years. I thought the bench was supposed to have been a strength. Is it the coaching? Is it the level of "clutchness"? (Dan gave clutch time extra weight and drew an apparent criticism for doing so from edijori earlier http://sonicscentral.com/apbrmetrics/vi ... ight=#8403, I tend to support a clutchtime weight but not sure how much. I mention in case either or others wish to expand of whether clutch weights should or should not be in best practice methods.) Just the toughness of the opponents, not getting a higher playoff seed? But why not higher seeding or better playoff performance? Is it that they are strong on other things that make them appear underrated but not high enough as a group on the "overrated" category of scoring?

I wonder how different the ratings would be if the weights were selected based on team playoff performance instead of regular season.


This year's Knicks had for a time 4 of the bottom 10 "difference makers" on this 2004 list from Dan http://www.82games.com/comm29.htm

Single examples don't validate or demonish a formula, I just noticed these cases. I believe the two differnet systems produce ratings for Kobe Bryant and Ray Allen that are close to each other. But that is surface level result similarity only and folks will write in different explanations of it.

Last edited by Mark on Mon Jun 12, 2006 7:13 am; edited 8 times in total
Back to top
View user's profile Send private message
deepak



Joined: 26 Apr 2006
Posts: 664


PostPosted: Sun Jun 11, 2006 12:17 pm Post subject: Reply with quote
More explanation on Win Score and Wins Produced

Win Score is a simple, approximate formula that can be used to quickly assess how good a game a player had. I supposed it's analogous to John Hollinger's game scores -- though even simpler to calculate.

Also, I wonder what their rationale is in adjusting for positions, beyond making their list more credible.

Here are the top 20 players, in terms of Win Score per 48 minutes:

Code:

Rank Player WinScore/48min
1 garnett,kevin 19.7
2 marion,shawn 18.8
3 camby,marcus 18.5
4 wallace,ben 17.5
5 hayes,chuck 16.9
6 foster,jeff 16.3
7 brand,elton 15.8
8 ming,yao 15.5
9 nowitzki,dirk 15.4
10 howard,dwight 15.3
11 duncan,tim 15.2
12 wallace,gerald 14.6
13 boozer,carlos 14.5
14 o'neal,shaq 14.2
15 mutombo,dikembe 14.2
16 gooden,drew 14.0
17 mourning,alonzo 14.0
18 chandler,tyson 13.9
19 evans,reggie 13.7
20 kidd,jason 13.7


I love Chuck Hayes, but number 5? And no LeBron anywhere.

BTW, I just noticed that this book was being discussed in a thread down below. Sorry about making a duplicate.
Back to top
View user's profile Send private message
Mark



Joined: 20 Aug 2005
Posts: 807


PostPosted: Sun Jun 11, 2006 1:08 pm Post subject: Reply with quote
18 big men and 2 perimeter by this method? I'm all for giving full deserved credit for big men but is that the main discovery?

Last edited by Mark on Wed Jun 28, 2006 12:42 am; edited 4 times in total
Back to top
View user's profile Send private message
bchaikin



Joined: 27 Jan 2005
Posts: 685
Location: cleveland, ohio

PostPosted: Sun Jun 11, 2006 1:23 pm Post subject: Reply with quote
I love Chuck Hayes, but number 5?

what chuck hayes produced statistically in just 535 minutes in 05-06 was excellent. i can easily see why a rating system (not just this one) would have him rated very high....

just looking at his numbers (and not noting that he may have accrued them solely against other teams' reserves and not their starters, which may or may not be true but if true could make a world of difference), if you multiplied his stats by 5 so instead of 535 minutes he played 2675 minutes he'd have scored 18.5 pts/g on a ScFG% of over 56% with 11 rebs/g playing 33 min/g - with 130 steals, 70 bs, and just 60 turnovers (uh... and also 380 fouls committed)....

would he do this if giving the playing time? who knows? but his stats in those 500+ minutes were outrageously good, kind of like anderson varejao in 04-05...
Back to top
View user's profile Send private message Send e-mail Visit poster's website
HoopStudies



Joined: 30 Dec 2004
Posts: 705
Location: Near Philadelphia, PA

PostPosted: Sun Jun 11, 2006 4:23 pm Post subject: Reply with quote
Dan Rosenbaum wrote:

But I suspect it is still worthwhile to read the book. Many of the points they make are still valid - even if they are not using an ideal metric. Also, they touch on a lot of topics that get little play here (such as how salaries relate to stats), and I suspect folks might find that interesting.


I just got the book and read some. Being pretty familiar with sports economics literature, I knew of many of these things, but I have been finding them interesting for years, which is why I'm familiar. If you do like issues on salaries vs winning, competitive balance, impacts of labor strikes on attendance, the value of having a star on your team -- you'll find it interesting.

Or if you want to hear him call me a Marxist, that's also in there.

The NBA GMs do seem to be reading it, as well, as I heard mention of it from a few people down in Orlando this week.

(It's not really my job to do so, but I will say that the method in this book is definitely NOT the one that he used in the 1999 paper. I walked him through a few things that caused him to change it pretty significantly. As to whether it's "good" or "bad" -- it's a rating system, there are a million of them, and the distinctive aspect of this one is that it penalizes poor shooters more. Dan says it's not necessarily better or worse than the NBA efficiency -- well, that's what most of them are.)
_________________
Dean Oliver
Author, Basketball on Paper
The postings are my own & don't necess represent positions, strategies or opinions of employers.
Back to top
View user's profile Send private message Visit poster's website
Mark



Joined: 20 Aug 2005
Posts: 807


PostPosted: Sun Jun 11, 2006 5:52 pm Post subject: Reply with quote
HoopStudies wrote:


Or if you want to hear him call me a Marxist, that's also in there.


Since you brought it up, I am curious on what grounds or in what context does he say this and what is your response?
Back to top
View user's profile Send private message
Mark



Joined: 20 Aug 2005
Posts: 807


PostPosted: Sun Jun 11, 2006 5:57 pm Post subject: Another question Reply with quote
edijorj, what is the source of the linear equation in your last post? How good is the fit?
Back to top
View user's profile Send private message
HoopStudies



Joined: 30 Dec 2004
Posts: 705
Location: Near Philadelphia, PA

PostPosted: Sun Jun 11, 2006 6:42 pm Post subject: Reply with quote
Mark wrote:
HoopStudies wrote:


Or if you want to hear him call me a Marxist, that's also in there.


Since you brought it up, I am curious on what grounds or in what context does he say this and what is your response?



Apparently, the concept of assigning credit based on difficulty of accomplishing a portion of a task is Marxian.
_________________
Dean Oliver
Author, Basketball on Paper
The postings are my own & don't necess represent positions, strategies or opinions of employers.
Back to top
View user's profile Send private message Visit poster's website
Ed Küpfer



Joined: 30 Dec 2004
Posts: 785
Location: Toronto

PostPosted: Sun Jun 11, 2006 6:51 pm Post subject: Reply with quote
HoopStudies wrote:
Apparently, the concept of assigning credit based on difficulty of accomplishing a portion of a task is Marxian.


It had nothing to do with your balaclava and Che Guevara underoos?
_________________
ed
Back to top
View user's profile Send private message Send e-mail
Mark



Joined: 20 Aug 2005
Posts: 807


PostPosted: Sun Jun 11, 2006 7:40 pm Post subject: Reply with quote
HoopStudies wrote:
"Apparently, the concept of assigning credit based on difficulty of accomplishing a portion of a task is Marxian."

I'd have to hear more of that explanation to understand what was meant. Maybe you could say that at a level of technical Marxian economic analysis, of value accounting- stressing the importance of the labor input.

But at the simplest philosophical level communism is "from each according to their abilities, to each according to their needs" and that seems contrary to the practice in the quote. I would think a classical political Marxist would likely have a VOP attitude of equal worth of different activities that produce equivalent results to me. And would probably be against moneyball and a skewed freemarket bidding war distribution of salary heavily to the top performers. The quote sounds like it has more in common with a capitalist attitude than a political Marxist one to me.

But I guess economic Marxian economic analysis can be considered somewhat distinct from that and it was initially applied to understanding capitalism, how the system actually works according to the existing rules and market values of different inputs in producing outputs-wins.

Berri is giving all rebounds defensive and offensive equal weight (at least in the simple version)? And you dont view the world that way. Is that part of what makes you Marxian to him? Hmmm... Not sure the label is particularly useful if that is the issue, though this debate is important. I guess if difficult, perhaps especially physical labor (like off. rebounding, defense, blocked shots) was seen as deserving more credit than shooters (and in some systems it is) that might perhaps be called Maoist Gang of 4 peasants vs the elite rhetoric.

Team +/- ratings for players could certainly seem pure communist but for the opposite rationale than the one asserted in the quote. Everybody gets the same score for that time on court together! Does he pull away from that by just using individual boxscore data, not give stress to contextual data, opposed that line of thinking? Extra weight for crunch time is that opposed too?

The general idea of value being added by high level managers, traders, consultants thru their intellectual capital contributions still seems patently capitalist, though I guess you could call it central planning and get away with the same thing as in historical communist state practice. The workers/players continue to be steered to some degree. But compared to football, basketball is a player's game.

page 2 missing

Author Message
Mark



Joined: 20 Aug 2005
Posts: 807


PostPosted: Mon Jun 18, 2007 7:46 pm Post subject: Scoring efficiency and shot creation together in a ranking Reply with quote
David Locke has revised his shooting rating significantly and has rankings by position here for big minute players:
http://lockedonsonics.blogspot.com
Seems useful in addition to looking at all the pieces separately.
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3604
Location: Hendersonville, NC

PostPosted: Tue Jun 19, 2007 5:58 am Post subject: Reply with quote
Kid starting out playing with stats?
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Mark



Joined: 20 Aug 2005
Posts: 807


PostPosted: Tue Jun 19, 2007 1:57 pm Post subject: Reply with quote
More like long-term mainstream media guy (radio) who understands the main modern stats and has introduced them to the broader audience and somebody who has looked at / messed with stats as a fan/student of the game like many here.

I thought it was worth a mention. It gets at total scoring edge produced by the player (compared to replacement player, not teammates though).

Last edited by Mark on Tue Jun 19, 2007 6:11 pm; edited 1 time in total
Back to top
View user's profile Send private message
Harold Almonte



Joined: 04 Aug 2006
Posts: 616


PostPosted: Tue Jun 19, 2007 3:27 pm Post subject: Reply with quote
I really don't understand why two skills only, and let off the whole scoring usage (he call it scoring oportunities). Maybe he is trying to say these are the players that really deserve "shooting oportunities". What it does mean a scoring oportunity for one player is not the same for another one. A player could have the ability and "oportunity for him" to score with an opponent on him; another player, on the other hand, could (or should) better opt to pass, or to drive and loose the ball. But, it's good to see a scoring exclusive metric, without defense and rebounding noises, wich put them at scoring order.
Back to top
View user's profile Send private message
mtamada



Joined: 28 Jan 2005
Posts: 377


PostPosted: Tue Jun 19, 2007 8:36 pm Post subject: Reply with quote
Mark wrote:
More like long-term mainstream media guy (radio) who understands the main modern stats and has introduced them to the broader audience and somebody who has looked at / messed with stats as a fan/student of the game like many here.


Yeah, Locke's about the most stats-savvy of all the mainstream announcers. For Sonics radio broadcasts, in the seasons before he became the main announcer, he was given little 10-second time slots at station breaks to chime in with some stats, and he used some pretty advanced ones, way more advanced that anything I'd ever heard on the radio or seen on TV (or in newspapers). For awhile there, with DeanO in the back office and KevinP and David Locke doing media work, the Sonics were at the cutting edge apbrmetrically.

But, I didn't like his scoring opportunities analysis. Didn't look useful or valid.
Back to top
View user's profile Send private message
Mark



Joined: 20 Aug 2005
Posts: 807


PostPosted: Tue Jun 19, 2007 10:25 pm Post subject: Reply with quote
Locke's approach to me measures value of scoring performance with the one leap/disconnect of reference to generic replacement player instead of keeping the context of team style/pace environment something that Mike G's scoring stat emphasizes. It is not a measure of scoring ability in a pure sense though for big minute guys it might be in many cases. Clearly some players are maximized or optimized and some are not. I would be interested in hearing your critique further if you are willing.
Back to top
View user's profile Send private message
Kevin Pelton
Site Admin


Joined: 30 Dec 2004
Posts: 979
Location: Seattle

PostPosted: Tue Jun 19, 2007 11:34 pm Post subject: Reply with quote
Locke's work isn't entirely original; it's actually an extension of Hollinger's Brick Index. I don't know that John ever intended that method to be used to rate players' scoring -- he was more interested in finding the really bad scorers than good ones.

Basically, all this stat does is adjust True Shooting Percentage for usage rate. I don't see the harm in that.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Mike G



Joined: 14 Jan 2005
Posts: 3604
Location: Hendersonville, NC

PostPosted: Wed Jun 20, 2007 5:43 am Post subject: Reply with quote
Radio guy, my bad. It just looked like he was reinventing the wheel, along the way misspelling lots of players' names, disregarding punctuation and other conventions that make reading easier; eventually concluding Chucky Atkins is a better shooter than Ray Allen?

There are many homemade measures that elevate players from all-O/no-D teams, and this just looked like another one. Maybe he won't like the results and will change it again. Almost certainly he'll learn some stuff and get better at it.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Harold Almonte



Joined: 04 Aug 2006
Posts: 616


PostPosted: Thu Jun 21, 2007 8:55 am Post subject: Reply with quote
mtamada said:
Quote:
But, I didn't like his scoring opportunities analysis. Didn't look useful or valid.


Just because it shows half scoring reality doesn't mean it can't be useful. If he does a parallel metric including A/TO, or better than that, Hollinger's PPR, then he can have two metrics: a "ball touches deserving metric" and his "final attempts (TS% oportunities) deserving metric", the whole usage package, and that would be very useful for a coach. But, however, scoring inmediately after an OReb. can overrate a bit some players's shooting at "oportunities" metric, although that may mean they deserve minutes.

Last edited by Harold Almonte on Thu Jun 21, 2007 6:25 pm; edited 2 times in total
Back to top
View user's profile Send private message
Mark



Joined: 20 Aug 2005
Posts: 807


PostPosted: Thu Jun 21, 2007 1:19 pm Post subject: Reply with quote
Harold, I agree you'd need to look at other metrics to get good sense of player's game, but this metric does what it intends and nothing else and as such is useful to me.

Mike, the early version stumbled on Atkins/Allen but the second version is better on formula and laugh test results check, elevating Allen ahead of Atkins and Duncan ahead of Nene. It might not be completely finished or the best for all but it is another tool available.

With it on top of the shooting %s alone and usage alone you can quickly line up similars on one of these or this product and see how the players array on the other two. Fodder for who's better discussions or impact discussions if you looked at similar usage guys and took leap of faith that FG% would transfer to some moderately reliable and high level.

Italian Stallion



Joined: 04 Mar 2009
Posts: 112


PostPosted: Thu Jul 15, 2010 9:28 am Post subject: Calculating the value of high usage inefficient scorers? Reply with quote
We've all seen examples of high usage scorers whose TS% is at or below the league average of (54%-55%). In fact quite a few top NBA scorers fall into that category.

If a high usage player's TS% is only average, how much value is he actually adding by taking a lot of extra shots and scoring a lot more points?

It seems to me that in most cases most of those those shots could easily be taken by other players on the team with similar results.

So little or no value is added.

If a high usage player's TS% is below average, how much value is he adding or actually destroying by taking a lot of extra shots and scoring a lot more points?

It seems to me he is destroying value.

I suppose it's possible that by taking some extra shots, a player's teammates might get to be more selective and "might" increase their own efficiency mildly. But intuitively it seems to me that the majority of value a scorer adds comes from contributing the first 15 or so points at average TS% or better so he doesn't weaken the overall offense and all the remaining value only comes from scoring at above average TS%. Any time you are scoring at below average efficiency you weaken the team (unless of course your teammates are even worse).
Back to top
View user's profile Send private message
BobboFitos



Joined: 21 Feb 2009
Posts: 201
Location: Cambridge, MA

PostPosted: Thu Jul 15, 2010 10:48 am Post subject: Reply with quote
Quote:

It seems to me that in most cases most of those those shots could easily be taken by other players on the team with similar results.


this is where you're mistaken. that pretty much is the root of it.
_________________
http://pointsperpossession.com/

@PPPBasketball
Back to top
View user's profile Send private message
DSMok1



Joined: 05 Aug 2009
Posts: 611
Location: Where the wind comes sweeping down the plains

PostPosted: Thu Jul 15, 2010 12:39 pm Post subject: Re: Calculating the value of high usage inefficient scorers? Reply with quote
Italian Stallion wrote:
We've all seen examples of high usage scorers whose TS% is at or below the league average of (54%-55%). In fact quite a few top NBA scorers fall into that category.

If a high usage player's TS% is only average, how much value is he actually adding by taking a lot of extra shots and scoring a lot more points?

It seems to me that in most cases most of those those shots could easily be taken by other players on the team with similar results.

So little or no value is added.

If a high usage player's TS% is below average, how much value is he adding or actually destroying by taking a lot of extra shots and scoring a lot more points?

It seems to me he is destroying value.

I suppose it's possible that by taking some extra shots, a player's teammates might get to be more selective and "might" increase their own efficiency mildly. But intuitively it seems to me that the majority of value a scorer adds comes from contributing the first 15 or so points at average TS% or better so he doesn't weaken the overall offense and all the remaining value only comes from scoring at above average TS%. Any time you are scoring at below average efficiency you weaken the team (unless of course your teammates are even worse).


I just ran a regression on this as part of my new Advanced SPM. Some samples of players near 0 in offensive worth, including assists and turnovers: Dwight Howard, Josh Smith, Gerald Wallace, Zach Randolph, Andrew Bogut, Lamar Odom, Rudy Gay, Jeff Green, Al Jefferson.

Players with TS% below 54% that have above +2 offense: Joe Johnson, Tyreke Evans, Russel Westbrook, Derrick Rose. Basically lots of assists or, in Joe Johnson's case, very few turnovers.

Basically, it varies. It depends on the turnover percentage, the assist percentage, and both true shooting and usage rates.

Here's a chart using league-average turnover% and league-average AST rate:


Back to top
View user's profile Send private message Send e-mail Visit poster's website
Italian Stallion



Joined: 04 Mar 2009
Posts: 112


PostPosted: Thu Jul 15, 2010 8:35 pm Post subject: Reply with quote
BobboFitos wrote:
Quote:

It seems to me that in most cases most of those those shots could easily be taken by other players on the team with similar results.


this is where you're mistaken. that pretty much is the root of it.


I understand that there may be some relationship between usage and efficiency, but let's take an example.

Player "X" has a TS% of 52% and takes close to 20 shots per night. That would make him a high scorer with slightly below average efficiency.

At any give time he's on the court with other 4 players.

1 has above average efficiency, 2 are average, and 1 is similar to him (net of average but better than him).

I find it hard to believe that player "X" couldn't reduce his total shot count by 2-4 shot per night, distribute them between the other players, and produce a better value for the team.
Back to top
View user's profile Send private message
DSMok1



Joined: 05 Aug 2009
Posts: 611
Location: Where the wind comes sweeping down the plains

PostPosted: Thu Jul 15, 2010 9:52 pm Post subject: Reply with quote
I believe this is a "Nash Equilibrium"/ game theory problem? If each player had linear usage vs. efficiency curves, then the optimum would have every player with the same TS%. The problem is that they aren't linear. You can construct parabolic skill curves for each player and then use Excel's solver to predict the best usage pattern for that team.... Only we don't know the skill curves.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
bchaikin



Joined: 27 Jan 2005
Posts: 690
Location: cleveland, ohio

PostPosted: Thu Jul 15, 2010 10:56 pm Post subject: Reply with quote
If a high usage player's TS% is only average, how much value is he actually adding by taking a lot of extra shots and scoring a lot more points? It seems to me that in most cases most of those those shots could easily be taken by other players on the team with similar results. So little or no value is added. If a high usage player's TS% is below average, how much value is he adding or actually destroying by taking a lot of extra shots and scoring a lot more points? It seems to me he is destroying value.

this is where you're mistaken. that pretty much is the root of it.

in cases where a really high scorer shoots overall worse than the league average, i certainly agree with the first statement...

for example, from 01-02 to 03-04 allen iverson scored 28.5 pts/g playing 43 min/g (2nd best scoring in the league during that time). he shot just a 48.1 ScFG% (43% on 2s, 28% on 3s - 250+ 3pt FGAs/yr - 78% on FTs) when the league average ScFG% was 51.1%, or 3% better. the rest of the 76ers shot a 50.9% ScFG% these 3 seasons, close to the league average...

the worst shooting team during this 3 year stretch was denver at a 48.6% ScFG% (the best dallas at a 53.6% ScFG%). so iverson all by his lonesome self shot worse than what every other single team did...

i'd like to hear the logic for why the rest of the 76ers would shoot worse as a team than what iverson shot (had he not played) when he himself shot worse than every other team, should anyone believe so...

in 05-06 the 76ers outside of iverson (who lead the team in scoring with 33 pts/g and minutes played with 3103) shot a 52.7% ScFG%. in 06-07 he was traded to denver (after 15 games and just 640 minutes) and the 76ers without him shot a 52.3% ScFG%...

in 07-08 denver outside of iverson (who lead the team in scoring with 26.4 pts/g and in minutes played with 3424) shot a 54.5% ScFG%. in 08-09 he was traded after 3 games and 123 minutes played, yet the nuggets shot a 55.2% ScFG%...

any idea that iverson shot the ball a ton simply because his teammates wouldn't or couldn't is nonsense...

iverson was the ages of 26-28 from 01-02 to 03-04 (looking at b-days as of 11/1). looking at all players since 1977-78 ages 26-28 that averaged 20+ pts/g and that played at least 3000 total minutes (a total of 78 different players) over the 3 years, iverson had - by far - the lowest/worst overall shooting...

as a matter of fact just the average ScFG% of these 78 players was a 55.0% ScFG%. so iverson shot almost 7% less than just the average of this group of high scorers at a similar age. you have to go back all the way to the early 1970s to find a player (elvin hayes) who shot worse overall from the ages of 26-28 while scoring 20+ pts/g and playing 3000+ minutes...
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Kevin Pelton
Site Admin


Joined: 30 Dec 2004
Posts: 979
Location: Seattle

PostPosted: Fri Jul 16, 2010 10:28 am Post subject: Reply with quote
DSMok1 wrote:
Only we don't know the skill curves.

Further, everything we do know about them suggests that they are much flatter for star players than for role players. It's perfectly reasonable to concoct a model that shows very different average True Shooting Percentages but identical marginal True Shooting Percentages for the scenario described above. And the latter is what is really relevant to the question of whether a player should shoot less.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Italian Stallion



Joined: 04 Mar 2009
Posts: 112


PostPosted: Fri Jul 16, 2010 10:41 pm Post subject: Reply with quote
Kevin Pelton wrote:
DSMok1 wrote:
Only we don't know the skill curves.

Further, everything we do know about them suggests that they are much flatter for star players than for role players. It's perfectly reasonable to concoct a model that shows very different average True Shooting Percentages but identical marginal True Shooting Percentages for the scenario described above. And the latter is what is really relevant to the question of whether a player should shoot less.


Does anyone think the Nuggets and Lakers wouldn't be better off if Melo and Kobe didn't shoot as much given that both teams have a few much higher efficiency options available?

I'm not sure how many extra shots you could get for those players and keep the efficiency on the incremental shots higher than Kobe/Melo, but the gap in efficiency is so large it almost has to be a few.
Back to top
View user's profile Send private message
DSMok1



Joined: 05 Aug 2009
Posts: 611
Location: Where the wind comes sweeping down the plains

PostPosted: Sat Jul 17, 2010 7:54 am Post subject: Reply with quote
Italian Stallion wrote:
Kevin Pelton wrote:
DSMok1 wrote:
Only we don't know the skill curves.

Further, everything we do know about them suggests that they are much flatter for star players than for role players. It's perfectly reasonable to concoct a model that shows very different average True Shooting Percentages but identical marginal True Shooting Percentages for the scenario described above. And the latter is what is really relevant to the question of whether a player should shoot less.


Does anyone think the Nuggets and Lakers wouldn't be better off if Melo and Kobe didn't shoot as much given that both teams have a few much higher efficiency options available?

I'm not sure how many extra shots you could get for those players and keep the efficiency on the incremental shots higher than Kobe/Melo, but the gap in efficiency is so large it almost has to be a few.


Here's the deal. Melo and Kobe take a lot of "unassisted" shots. If a shot is assisted, its TS% is somewhere around 7% higher, not to mention a lot of close shots wouldn't even happen without an assist. If, say, Chris Anderson was going to increase his number of shots, well, he can't without jacking bad shots. His shots are nearly all assisted. I would hazard that his "unassisted" shots are much, much lower TS%. I calculated once what an assist contributed. Most of the value of a shot at the rim has to do with the passer, not the shooter, because the shooter would not get the shot without the passer.

In fact, in another SPM regression I was experimenting with using shot location and assisted%. Basically, a player like Chris Anderson who shoots a very high TS% but most shots are assisted really isn't the one to credit at all. He can't take more shots unless someone assists him. He has the high TS% purely because of Billups and, to a lesser extent, Melo.

Incidentally, that was the only regression that could explain the high APM split between Nash and Amare. Amare's high TS% is mostly attributable to Nash's ridiculous number of assists to "At Rim".
_________________
GodismyJudgeOK.com/DStats
Twitter.com/DSMok1
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Italian Stallion



Joined: 04 Mar 2009
Posts: 112


PostPosted: Sun Jul 18, 2010 11:06 am Post subject: Reply with quote
DSMok1 wrote:
Italian Stallion wrote:
Kevin Pelton wrote:
DSMok1 wrote:
Only we don't know the skill curves.

Further, everything we do know about them suggests that they are much flatter for star players than for role players. It's perfectly reasonable to concoct a model that shows very different average True Shooting Percentages but identical marginal True Shooting Percentages for the scenario described above. And the latter is what is really relevant to the question of whether a player should shoot less.


Does anyone think the Nuggets and Lakers wouldn't be better off if Melo and Kobe didn't shoot as much given that both teams have a few much higher efficiency options available?

I'm not sure how many extra shots you could get for those players and keep the efficiency on the incremental shots higher than Kobe/Melo, but the gap in efficiency is so large it almost has to be a few.


Here's the deal. Melo and Kobe take a lot of "unassisted" shots. If a shot is assisted, its TS% is somewhere around 7% higher, not to mention a lot of close shots wouldn't even happen without an assist. If, say, Chris Anderson was going to increase his number of shots, well, he can't without jacking bad shots. His shots are nearly all assisted. I would hazard that his "unassisted" shots are much, much lower TS%. I calculated once what an assist contributed. Most of the value of a shot at the rim has to do with the passer, not the shooter, because the shooter would not get the shot without the passer.

In fact, in another SPM regression I was experimenting with using shot location and assisted%. Basically, a player like Chris Anderson who shoots a very high TS% but most shots are assisted really isn't the one to credit at all. He can't take more shots unless someone assists him. He has the high TS% purely because of Billups and, to a lesser extent, Melo.

Incidentally, that was the only regression that could explain the high APM split between Nash and Amare. Amare's high TS% is mostly attributable to Nash's ridiculous number of assists to "At Rim".


I don't disagree with anything you are saying, but IMO if both Melo and Kobe looked to pass more often instead of jacking up the occasional idiotic shot, their TS% would probably rise and the team would be more productive because both have high TS% alternatives to get the ball to and both have enough passing skills to accomplish the task.
Back to top
View user's profile Send private message
thref23



Joined: 13 Aug 2007
Posts: 90


PostPosted: Sun Jul 18, 2010 5:44 pm Post subject: Reply with quote
Not in disagreement with discussion thus far, I think it's vital to compare a player's efficiency with that of his teammates on the floor with him...usage (or the equivalent) + efficiency should invite defensive attention, invited defensive attention should improve teammates' efficiency, shot selection should maximize distributory effects (and plus/minus stats should help with analysis of such effects).

From a conceptual standpoint I think this is interesting:

http://sloansportsconference.com/upload ... etball.pdf
Back to top
View user's profile Send private message Send e-mail
Italian Stallion



Joined: 04 Mar 2009
Posts: 112


PostPosted: Mon Jul 19, 2010 1:41 pm Post subject: Reply with quote
DSMok1,

I am really looking forward to seeing how Amare does with the Knicks. I think he's a pretty good test case for the value of Nash and whether his box score numbers or his +/- numbers are telling a more accurate story about his contribution.
Back to top
View user's profile Send private message
gabefarkas



Joined: 31 Dec 2004
Posts: 1313
Location: Durham, NC

PostPosted: Tue Jul 20, 2010 7:29 am Post subject: Reply with quote
bchaikin wrote:
for example, from 01-02 to 03-04 allen iverson scored 28.5 pts/g playing 43 min/g (2nd best scoring in the league during that time). he shot just a 48.1 ScFG% (43% on 2s, 28% on 3s - 250+ 3pt FGAs/yr - 78% on FTs) when the league average ScFG% was 51.1%, or 3% better. the rest of the 76ers shot a 50.9% ScFG% these 3 seasons, close to the league average...
Bob - can you remind me how ScFG% is calculated?
Back to top
View user's profile Send private message Send e-mail AIM Address
Prash



Joined: 30 Jul 2010
Posts: 1
Location: California

PostPosted: Sat Jul 31, 2010 12:07 pm Post subject: Reply with quote
Just wanted to chime in here on the TS% vs usage rate and the difference between role players and star players

let's assume we are talking about the Lakers and we suspect that if Kobe reduces his usage rate his TS% will go up

Then you would distribute this 'extra' usage rate that Kobe just sacrificed to the players on the court who had the highest marginal TS% gains with increases in usage rate (most likely Gasol/Bynum)

I am just wondering if it is possible to really know the marginal TS% gains ahead of time?

The reason I am skeptical is that defenses play the Lakers a certain way - they expect Kobe to have a certain usage rate. Once his usage rate changes wouldn't defenses adapt to these changes eventually?

And once defenses do adapt wouldn't this affect the marginal TS% gains of all other Lakers on the court because the defense now pays more attention to them since Kobe has a lower usage rate?
Back to top
View user's profile Send private message
Crow



Joined: 20 Jan 2009
Posts: 825


PostPosted: Mon Aug 02, 2010 2:51 am Post subject: Reply with quote
By DSMok1's method and chart, it appears to me that a player with usage of 25% and a TS% of about 51% (and 1 to 1 assist% / turnover%) is about neutral and if he can raise either of these parts he becomes positive on offensive SPM because of the "scoring term".

And for the usage band between 20-25% (and still assuming 1 to 1 assist% / turnover%) the TS% needed to break into positive territory is pretty flat and about that same 51% TS%.

At 27.5% usage the TS% threshold for positive offensive SPM impact from the scoring term is right at 50% TS% and at 30% usage it is slightly below 50% TS%.


Looking at players who played over 2000 minutes last season with usage at or above 25% I see 26 of 30 over 51% TS%. I didn't stop to check the Assist % vs TO% for all but I do quickly see just 7 with less than an raw count A/TO ratio of 1 but only 2-3 significantly lower.

Limit to those at or above 27.5% usage and all 10 are over 50% TS% safely and are all near or above 52% TS%. 2 have less than a raw count A/TO of 1 (Durant and Bosh) but not that badly.

Lower the usage to 20% and there 74 qualifiers of which 66 are over 51% TS%. 18 less than a raw count A/TO of 1 but not many badly, depending on where it becomes significantly bad.


Based on these findings and my current understanding of them, it appears that the source of any negative impact for the offensive or overall SPM of big minute, over 20% usage players who have one is usually not these parts of their "scoring term" but rather the other parts of the scoring term or non-scoring related elements.

How different is the chart and positive breakpoint if the A%/TO% is less or more than 1? A chart of the break-even points for usage levels of 15, 20, 25% across these various ratios would be helpful to see.


If the impact of changes in Assist% is not static and equivalent to equal unit changes in Turnover% at other levels besides 13% on each as I assumed above, how does "the break-even impact ratio" of them change across the distributions for Assist% and Turnover%? Just how important is this ratio in shifting the TS% break-even threshold for the scoring term of lead play or pass makers vs non-elevated passers?


For a) the most recent previous version of overall SPM (which uses TR%) and b) the newer offensive and defensive splits (which use the respective OR% and DR%s) how often is a shortfall from league average rebounding rate for all players at all positions and levels of responsibility and opportunity the leading reason why a player with 20+% or 25+% usage has a negative overall SPM or split SPM? If it is very often I'd think that would a good reason to position or role adjust the SPM. If you don't want to do it for 5 traditional positions, maybe PG, wing, big would be enough or even perimeter-interior.

Last edited by Crow on Tue Aug 03, 2010 2:55 pm; edited 2 times in total

Scott S



Joined: 10 Feb 2008
Posts: 46
Location: East Rutherford, NJ

PostPosted: Mon Aug 02, 2010 12:58 pm Post subject: Reply with quote
I'd like to second DSMok1's point about higher usage players taking a lower percentage of assisted shots. If Chris Andersen shoots 55% on all close assisted shot and these assist type passes have an 80% chance of reaching him without creating a turnover, another player who creates all of his own shots will only need to effectively shoot 44% to be as effective an option for his team, even though the entry passing turnovers might not be Andersen's fault. I mention this in my website's most recent post and will go into this in more detail in the near future.
As a side point, in regards to skill curves, do they assume that if a player takes more shots, each additional shot created is more difficult or that if the player takes a larger role in the teams offense, all of his shots are equally more difficult? I've always assumed the former, but think a hint of the latter can be included. It seemed that the guy who gave the presentation about traffic patterns and high scorers taking too many shots at the Sloan Conference this past year assumed the second perspective (or at least some people I was conversing with after his presentation).
Back to top
View user's profile Send private message Visit poster's website
Crow



Joined: 20 Jan 2009
Posts: 798


PostPosted: Mon Aug 02, 2010 2:50 pm Post subject: Reply with quote
For the whether to pass and at least try to get Andersen an assisted shot vs just go with another player unassisted offensive efficiency choice, it would be helpful to know his actual turnover rate per pass received, when using the possession and when he doesn't (in addition to his overall turnover% per possession used). That way you would get a better estimate of the specific cost of the "look" / pass. It is probably quite a bit less than the overall turnover% per possession he actually used.

In general it would be helpful to know rough estimates for the average number of touches per possession or play and the league average turnover rate per perimeter and interior pass made / received. I looked briefly but didn't find it.

How much more risky / costly are interior passes than perimeter passes? It matters for optimization of team activity at the discrete action level. Sounds like you might be going somewhere close to this. If somebody looked at a sufficient sample of tape it might be possible to estimate by more detailed floor zone.

Last edited by Crow on Tue Aug 03, 2010 2:44 pm; edited 1 time in total
Back to top
View user's profile Send private message
Scott S



Joined: 10 Feb 2008
Posts: 46
Location: East Rutherford, NJ

PostPosted: Mon Aug 02, 2010 4:15 pm Post subject: Reply with quote
I was going to use this in my next post, but my tentative results using regression analysis to predict turnovers are as follows:

Code:

TOs/100 Poss Coefficient
astdclose 0.260
createdclose 0.203
astd2j 0.051
created2j 0.066
astd3p 0.062
created3p 0.045
3pasts 0.233
2Jasts 0.122
Closeasts 0.283


Code:

TOs/100 Poss Coefficient
AstdClose 0.322
AstdClose^2 -0.012
CreatedClose 0.245
CreatedClose^2 -0.007
Astd2J 0.032
Created2J 0.113
Created2J^2 -0.005
Astd3p 0.042
Created3p 0.059
3pAsts 0.109
2JAsts 0.112
CloseAsts 0.410
CloseAsts^1.5 -0.047


These indicate that the success rate in creating an additional assist on a close basket is between 70% and 80% in most instances. 2 point assists correspond with 90% "success rate" (this could mean the player passes the ball to a teammate in position to score a 2 point jump shot 300 times and they pass it up 100 times, make 90 shots, miss 100 shots and there is a passing turnover 10 times). Assists correlate with players who handle the ball, so some of the fluctuation in assists could be an indicator of ballhandling turnovers. I am more likely to believe the second model's indication that 3 point assists are more than 90% successful opposed to the first model's conclusion that success rates are only 81%.

Of course, you should keep in mind my study was done to determine the efficiency of a player who attempts to pass. This includes players who commits a ball-handling turnover while looking for a teammate to pass to, but not a player looking to create a close shot but throws the ball away in desperation. I don't know if passing turnovers would be more applicable or not in this situation.

Another side note, it would be helpful in terms of value if passes were tracked, or at least passes from a scoring position. If a player has a 75% chance of making a pass for a layup without turning it over and the shooter would make the shot 70% of the time, but they kick it out, that action hurts his team and he might even get credit for the assist using the best measurements I've seen at the moment.

Eli W



Joined: 01 Feb 2005
Posts: 401


PostPosted: Thu Mar 06, 2008 5:05 pm Post subject: Usage vs. efficiency yet again Reply with quote
I took the method from my last study on diminishing returns in rebounding and applied it to scoring, which is basically just another way of investigating the usage vs. efficiency tradeoff. I used some of Dean's stats from Basketball on Paper and lineup data from BasketballValue. I think the big advantage to the method I used is that it narrows things down to situations in which players were forced to increase or decrease their usage, which removes most of the usual confounds from players upping their usage in games/matchups when they were playing better (more efficiently) due to having a weak defender on them or just being "hot."

http://www.countthebasket.com/blog/2008 ... fficiency/
_________________
Eli W. (formerly John Quincy)
CountTheBasket.com
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Neil Paine



Joined: 13 Oct 2005
Posts: 774
Location: Atlanta, GA

PostPosted: Thu Mar 06, 2008 6:28 pm Post subject: Reply with quote
That's some phenomenal work there, JQ. I'm not a huge fan of the low R^2, but I like the methodology a lot... In fact, changing the WARP formula to use the 1.25 tradeoff instead of 0.6 creates results that "feel" more believable (Calderon is no longer side-by-side with, say, Kobe -- JC dropped to 17th-most-valuable, and Kobe jumped up to 3rd).

Also, I'd love to hear Dean's thoughts on this new research, and whether it means we can now assume the typical player's usage-efficiency tradeoff is +/- 1.25, not +/- 0.6.
Back to top
View user's profile Send private message Visit poster's website
Harold Almonte



Joined: 04 Aug 2006
Posts: 616


PostPosted: Thu Mar 06, 2008 7:29 pm Post subject: Reply with quote
I'm not sure, but I feel that this +/-1.25ORtg. each -/+1% usage is not linearly proportional all the way at the player level. I also feel that the way a player's usage is built has something to do. A player like Calderon with an A/TO higher than 3, and a lot of his usage built by that, and a player like Kevin Martin with a lot of his usage built by FTA, won't change eff. vs. usage like an average player.

Last edited by Harold Almonte on Thu Mar 06, 2008 7:51 pm; edited 1 time in total
Back to top
View user's profile Send private message
HoopStudies



Joined: 30 Dec 2004
Posts: 705
Location: Near Philadelphia, PA

PostPosted: Thu Mar 06, 2008 7:38 pm Post subject: Reply with quote
I need to read what Eli did, but I will say that the 0.6 number I generated was a quick and dirty thing. I never use that since I generate individual values and use those. If Eli finds a different number, it's not a big deal. I hope he puts some range on it so that we can be confident it's not 0....

I'll try to look at it soon.
_________________
Dean Oliver
Author, Basketball on Paper
The postings are my own & don't necess represent positions, strategies or opinions of employers.
Back to top
View user's profile Send private message Visit poster's website
Neil Paine



Joined: 13 Oct 2005
Posts: 774
Location: Atlanta, GA

PostPosted: Thu Mar 06, 2008 7:45 pm Post subject: Reply with quote
So, I'm guessing you're not at liberty to disclose how you generate individual values... or are you?

(Sorry, I had to try! Smile)
Back to top
View user's profile Send private message Visit poster's website
HoopStudies



Joined: 30 Dec 2004
Posts: 705
Location: Near Philadelphia, PA

PostPosted: Thu Mar 06, 2008 7:48 pm Post subject: Reply with quote
Harold Almonte wrote:
I'm not sure, but I feel that this +/-1.25ORtg. each -/+1% usage is not linearly proportional all the way at the player level. I also feel that the way a player's usage is built has something to do.


There is no doubt that usage vs efficiency is a simplification for something more complex. All linear weights methods assume a constant value of stats regardless of how many you rack up. Hell, my own stuff is only a little better by having weights depending on context (which is reflected in quantity). But dramatically changing contexts are hard. Usg v efficiency - skill curves - was the first step towards dealing with that. Going further opens up a can of worms.
_________________
Dean Oliver
Author, Basketball on Paper
The postings are my own & don't necess represent positions, strategies or opinions of employers.
Back to top
View user's profile Send private message Visit poster's website
Ben F.



Joined: 07 Mar 2005
Posts: 391


PostPosted: Thu Mar 06, 2008 10:09 pm Post subject: Reply with quote
Really great job, Eli, as always.

One thing I wonder about the graph - how big is the sample size for the farthest right point on the graph? Because without that point the graph looks like it's saying adding usage to a lineup gets you very little, and it doesn't matter how much you add. Whether it's +10% usage or +2% usage, you only get around +1 ORTG. Yet +12% usage jumps you way up, to close to +3 ORTG. That seems odd, so I wonder if there's any sample issues with that. If so, what does it mean that adding a lot of usage above league average doesn't get you much? And if not, why the sudden jump?
Back to top
View user's profile Send private message
Eli W



Joined: 01 Feb 2005
Posts: 401


PostPosted: Thu Mar 06, 2008 10:42 pm Post subject: Reply with quote
Ben F. wrote:
One thing I wonder about the graph - how big is the sample size for the farthest right point on the graph? Because without that point the graph looks like it's saying adding usage to a lineup gets you very little, and it doesn't matter how much you add. Whether it's +10% usage or +2% usage, you only get around +1 ORTG. Yet +12% usage jumps you way up, to close to +3 ORTG. That seems odd, so I wonder if there's any sample issues with that. If so, what does it mean that adding a lot of usage above league average doesn't get you much? And if not, why the sudden jump?


The last point represents just 97 lineups with 2784 total possessions played. Generally I don't think the sample sizes of the bins are large enough to draw any specific conclusions about the ups and downs of the chart. I was mainly just trying to show the general trend. Another issue with the chart is that I weighted the diffORtg's of the lineups in each bin by the number of possessions each lineup played - I think that's an improvement over taking just a plain average, but it also has some disadvantages (lineups that played together a lot can have a large effect). With more years of data it would be easier to draw conclusions on smaller segments, but my guess is more data would just serve to smooth things out more. Though the suggestion that low-usage lineups lose more efficiency in ramping up their usage than high-usage lineups gain when decreasing their usage does fit with Dean's findings in his study.
_________________
Eli W. (formerly John Quincy)
CountTheBasket.com
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Guy



Joined: 02 May 2007
Posts: 109


PostPosted: Thu Mar 06, 2008 10:44 pm Post subject: Reply with quote
Nice work, Eli. A few random thoughts:

Is the sample big enough to look separately at lineups with high vs low projected efficiency? It would be especially interesting to know what happens to low-usage/high-projected-efficiency lineups (since these are the players who should be shooting more, in absence of diminishing returns).

I think your projORtg should probably reflect actual usage in that lineup (if you have it), not the season average. For example, if a .90 usage lineup gets to 1.00 by giving all the 'extra' shots to its most efficient shooter, then the no-diminishing-returns efficiency expectation is a bit higher. (However, don't think this would change your results materially.)

It seems to me it's the low usage part of your graph that matters more, which is where the tradeoff is even larger (looks like a coefficient of approx .35). The important question is not how much more efficient, if at all, Kobe would be at average usage, but how much better he makes his teammates by employing 31% of the possessions. Looking at your graph, it looks like reducing the usage of four players from .80 to .69 (.0275 per player) increases efficiency by about .05. That means an extra 2.5 to 3 points a game, if I've done the math right. Pretty impressive. The fact that reducing the usage of high-usage players seems to pay only a small dividend is further disincentive to shift more shooting to low-usage players.
Back to top
View user's profile Send private message
Eli W



Joined: 01 Feb 2005
Posts: 401


PostPosted: Thu Mar 06, 2008 10:49 pm Post subject: Reply with quote
Harold Almonte wrote:
I'm not sure, but I feel that this +/-1.25ORtg. each -/+1% usage is not linearly proportional all the way at the player level. I also feel that the way a player's usage is built has something to do. A player like Calderon with an A/TO higher than 3, and a lot of his usage built by that, and a player like Kevin Martin with a lot of his usage built by FTA, won't change eff. vs. usage like an average player.


Definitely. All I was looking at was a general average, which obviously lumps a lot of things together and obscures differences between players and lineups. And you're right that individual possessions is something of a blunt instrument for these purposes. It's worth investigating how the tradeoff affects different parts of usage, from shooting to foul drawing to assists to turnovers.
_________________
Eli W. (formerly John Quincy)
CountTheBasket.com
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Eli W



Joined: 01 Feb 2005
Posts: 401


PostPosted: Thu Mar 06, 2008 10:55 pm Post subject: Reply with quote
Guy wrote:
Is the sample big enough to look separately at lineups with high vs low projected efficiency? It would be especially interesting to know what happens to low-usage/high-projected-efficiency lineups (since these are the players who should be shooting more, in absence of diminishing returns).


I tried two separate regressions - one with high-usage lineups and one with low-usage lineups, and unfortunately the sample size was a problem. I couldn't get a statistically significant coefficient for high-usage lineups. I think I did get one for low-usage lineups though. I'll go back and try to find that.

Guy wrote:
I think your projORtg should probably reflect actual usage in that lineup (if you have it), not the season average. For example, if a .90 usage lineup gets to 1.00 by giving all the 'extra' shots to its most efficient shooter, then the no-diminishing-returns efficiency expectation is a bit higher. (However, don't think this would change your results materially.)


Unfortunately I don't have that information. It would be nice to have and investigate though.
_________________
Eli W. (formerly John Quincy)
CountTheBasket.com
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Guy



Joined: 02 May 2007
Posts: 109


PostPosted: Thu Mar 06, 2008 11:01 pm Post subject: Reply with quote
Quote:
I tried two separate regressions - one with high-usage lineups and one with low-usage lineups, and unfortunately the sample size was a problem.

Starting with 4 simple bins, usage X projected efficiency, might be instructive:
low-usage/low-projected-efficiency
low-usage/high-projected-efficiency
high-usage/low-projected-efficiency
high-usage/high-projected-efficiency
Back to top
View user's profile Send private message
Eli W



Joined: 01 Feb 2005
Posts: 401


PostPosted: Thu Mar 06, 2008 11:19 pm Post subject: Reply with quote
Ok, I found the separated regressions that I ran. I'll try the breakdown Guy suggested later.

For all 5387 low-usage lineups (csum%TmPoss < 1), the OLS coefficient was 0.56 with a SE of 0.12 and an R-squared of 0.004. Weighting by possessions, the coefficient was 0.32 with a SE of 0.08, R-squared of 0.003. Using only the 240 low-usage lineups that played at least 50 possessions together, the coefficient was -0.23 with a SE of 0.19 and an R-squared of 0.006.

All 2729 high-usage lineups: coefficient of 0.24, SE 0.24, R-squared 0.0004. Weighted by possessions: coefficient of 0.096, SE 0.103, R-squared 0.0003. The 315 high-usage lineups with at least 50 possessions: coefficient of 0.40, SE 0.17, R-squared 0.02.
_________________
Eli W. (formerly John Quincy)
CountTheBasket.com
Back to top
View user's profile Send private message Send e-mail Visit poster's website
DLew



Joined: 13 Nov 2006
Posts: 222


PostPosted: Thu Mar 06, 2008 11:41 pm Post subject: Reply with quote
This is really great work Eli. I quite enjoyed it. I hope that Henry links it on Truehoop.

Could you do the same thing with multiple seasons worth of data to determine how robust the findings are?
Back to top
View user's profile Send private message
Eli W



Joined: 01 Feb 2005
Posts: 401


PostPosted: Thu Mar 06, 2008 11:55 pm Post subject: Reply with quote
DLew wrote:
Could you do the same thing with multiple seasons worth of data to determine how robust the findings are?


Yeah, I forgot that BasketballValue had lineup data from the last two seasons that could work for this study. For the rebounding study only this year's data worked (because that's when Aaron added lineup rebounding numbers), so when I started this study I unthinkingly again just used this season's data. I'll try to re-run things with the larger data set in the next few days.
_________________
Eli W. (formerly John Quincy)
CountTheBasket.com

cherokee_ACB



Joined: 22 Mar 2006
Posts: 157


PostPosted: Fri Mar 07, 2008 4:01 am Post subject: Reply with quote
One quick calculation. If the usage coefficient is 0.25, then the marginal efficiency is roughly 106-0.25*100 = 81, which corresponds to 33% eFG.
Back to top
View user's profile Send private message
Guy



Joined: 02 May 2007
Posts: 128


PostPosted: Fri Mar 07, 2008 6:55 am Post subject: Reply with quote
Eli: Could you say more about why you think weighting lineups by possessions is the better approach? I see the advantage in terms of sample size and reliability. But this will tend to give much more weight to lineups with usage rates close to 1, and minimize impact of very high/low usage lineups. If we only cared about the lineup overall, that would probably be fine. But you're using this to try to understand players, whose usage rates vary far more. So getting a good read on extreme lineups is arguably where you'll get important insights.

It's also interesting that the coefficients are so different depending on how you weight the sample. Suggests the possibility that you have a non-linear relationship here, in which the impact on efficiency grows the more a player departs from his normal usage level (at least for low-usage players, less clear re: high-usage players). That certainly seems plausible. When you have the expanded dataset, you may want to look at non-linear models.
Back to top
View user's profile Send private message
HoopStudies



Joined: 30 Dec 2004
Posts: 705
Location: Near Philadelphia, PA

PostPosted: Fri Mar 07, 2008 10:47 am Post subject: Reply with quote
Good piece.

Obviously, I come from the bias that this effect must exist. I think your work supports this and illustrates that it is statistically significant. There are some ways to poke small holes in this, but my sense is that the burden of proof goes further towards those saying that this effect doesn't exist.

If you do go to multiple years, you have to consider Ortg/usage being from career, multiyear, or single year data.

As to the magnitude, that ends up being an argument for the linear weights people to deal with. I haven't spent a lot of time trying to figure out the assumed importance those methods make on shot creation or how that relates to the baseline shooting percentage.
_________________
Dean Oliver
Author, Basketball on Paper
The postings are my own & don't necess represent positions, strategies or opinions of employers.
Back to top
View user's profile Send private message Visit poster's website
Eli W



Joined: 01 Feb 2005
Posts: 402


PostPosted: Fri Mar 07, 2008 3:15 pm Post subject: Reply with quote
Guy wrote:
Eli: Could you say more about why you think weighting lineups by possessions is the better approach? I see the advantage in terms of sample size and reliability. But this will tend to give much more weight to lineups with usage rates close to 1, and minimize impact of very high/low usage lineups. If we only cared about the lineup overall, that would probably be fine. But you're using this to try to understand players, whose usage rates vary far more. So getting a good read on extreme lineups is arguably where you'll get important insights.


Yeah, I'm open to any suggestions in this area. With no weights, the high uncertainty actORtg's of lineups that played together for only a few possessions count equally with the more reliable actORtg's of lineups that played together a lot. Weighting by possessions controls for that but it may go too far in the other direction in terms of over-weighting the lineups that played together a lot. A compromise is to use a minimum possession cutoff, but that makes the sample sizes for each bin pretty small. Using more data should help things.

That's all just for the binned chart values. For running regressions on the data as a whole it's not as big an issue because the sample is larger.
_________________
Eli W. (formerly John Quincy)
CountTheBasket.com
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Mountain



Joined: 13 Mar 2007
Posts: 1527


PostPosted: Sat Mar 08, 2008 1:46 am Post subject: Reply with quote
Trying to get back to the challenge of a coach I wonder if some data separation could be done to look at relationship between offensive usage/offensive efficiency and defensive efficency in 5 man lineups?

Does that help explain use of low usage and / or efficency lineups better than the offensive data does alone?

Do high defensive efficency lineups within the other 3 groups besides high usage-high efficiency make up for their offensive shortfall on the defensive side of the ball?

Do coaches do better or worse on this net efficiency challenge with lineups used 50+ possessions compared to those below 50?

Would game to game recalculation of this show on average NBA coaches learning the relative strength or weakness of a lineup and either increasing or decreasing use or effectively administering instruction to improve the net results? Or are they guessing in the dark and / or not demonstratig learning or improvement on average? How many coaches/ teams show good results on this?

Strength of opponent defense matters as does strength of defense of your lineup.
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3601
Location: Hendersonville, NC

PostPosted: Sun Mar 09, 2008 8:02 am Post subject: Reply with quote
Guy wrote:
Quote:
I tried two separate regressions - one with high-usage lineups and one with low-usage lineups, and unfortunately the sample size was a problem.

Starting with 4 simple bins, usage X projected efficiency, might be instructive:
low-usage/low-projected-efficiency
low-usage/high-projected-efficiency
high-usage/low-projected-efficiency
high-usage/high-projected-efficiency

I like Guy's idea; of course, then you have even smaller sample sizes. But don't we all wonder how high-eff/low-usage lineups fare? Again, opponent response to such unbalanced lineups becomes an issue. Efficiency, in scoring or rebounding, is really about gains vs the opponent.

HoopStudies wrote:
... I come from the bias that this effect must exist.
...
As to the magnitude, that ends up being an argument for the linear weights people to deal with. I haven't spent a lot of time trying to figure out the assumed importance those methods make on shot creation or how that relates to the baseline shooting percentage.

One could assume that a high-% shooter can shoot more, and his % can drop a ways before it's not good enough to shoot more. Conversely, a low-% shooter generally has an inflated scoring rate; and he might be a 'better' scorer by shooting less.

The low-% shooter's scoring rate can be deflated by multiplying by (his TS%, divided by a standard TS%), raised to some power. Playing with that exponent, and supposing a TS% of comparison is .525, then we can standardize the 'effective' scoring of guys who score 15 pts/36min with disparate TS%:
Sco = Pts/36 * (TS%/.525)^E

Code:
15pts/36 'equivalent' pts/36
TS% E= .00 .50 1.00 2.00
.625 15.0 16.4 17.9 21.3
.525 15.0 15.0 15.0 15.0
.425 15.0 13.5 12.1 9.8

The 'average' shooter, .525, is a 'true' 15 pts/36 scorer. The terrible one, .425, may be 'really' a 12 or a 10 scorer.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Eli W



Joined: 01 Feb 2005
Posts: 402


PostPosted: Tue Mar 11, 2008 3:10 pm Post subject: Reply with quote
Eli W wrote:
DLew wrote:
Could you do the same thing with multiple seasons worth of data to determine how robust the findings are?


Yeah, I forgot that BasketballValue had lineup data from the last two seasons that could work for this study. For the rebounding study only this year's data worked (because that's when Aaron added lineup rebounding numbers), so when I started this study I unthinkingly again just used this season's data. I'll try to re-run things with the larger data set in the next few days.


I spoke too soon. The BasketballValue data from 05-06 and 06-07 overestimated possessions (there's been some discussion of it on the forum), and that's causing a lot difficulties. Team possession figures being too high makes player %TmPoss and lineup actual ORtg's too low. I can make some estimated adjustments, but I'm afraid that will throw things off too much for the kind of sensitive analysis I'd like to do on the larger data set. I'm going to keep working on it, but at least for now I'm not going to be able to supplement the current season regressions that I ran.
_________________
Eli W. (formerly John Quincy)
CountTheBasket.com
Back to top
View user's profile Send private message

Italian Stallion



Joined: 04 Mar 2009
Posts: 112


PostPosted: Mon Jun 01, 2009 10:35 pm Post subject: Scoring, Usage, Efficiency, and Shot Distribution Reply with quote
Has anyone done any detailed studies on how variations in player usage impact scoring efficiency?

Before anyone comments along these lines, I already understand that simply lowering a player's shot count won't change a thing. What has to change is the distribution of his shot attempts.

In other words if he take 20 shots made up of 4 dunks, 4 layups, 4 mid range jumpers, and 8 shots with tough coverage or from deep in two point territory, nothing is going to change if it's 2, 2, 2, and 4 for a total of 10 instead other than his PPG.

To become more efficient, there has to be a reduction in the 8 tough shots or an increase in the easy ones.

I would assume it's fairly easy to reduce tough shots, but not always in the best interests of the team if the alternatives are sometimes worse.

I would assume it's fairly difficult to increase the easy ones without an actual change in skill otherwise the player/team would already be doing it .

I would also assume it's very difficult for lower skilled, lower usage, but highly efficient players to increase their usage much without adding more difficult shots and decreasing their efficiency.

How to measure this tradeoff seems to be a key to understanding PPG vs. Scoring Effeciency. Am I off base?
Back to top
View user's profile Send private message
kjb



Joined: 03 Jan 2005
Posts: 865
Location: Washington, DC

PostPosted: Tue Jun 02, 2009 11:32 am Post subject: Reply with quote
Dean Oliver examines this issue in his book "Basketball On Paper." The relationship between efficiency and usage is an important one. Somewhere on these boards, Dean posted a rough rule of thumb showing the average relationship (I don't remember what it is off the top of my head). But, there's variation. Some guys get more efficient when they start using more possessions. Other guys get worse.
Back to top
View user's profile Send private message AIM Address Yahoo Messenger
Mike G



Joined: 14 Jan 2005
Posts: 3613
Location: Hendersonville, NC

PostPosted: Tue Jun 02, 2009 12:25 pm Post subject: Reply with quote
kjb wrote:
... Some guys get more efficient when they start using more possessions. Other guys get worse.

Right-O. Another RoT is that most guys do better off the bench than starting. Or 'vs subs' better than 'vs starters'.

Also, youngsters and old guys fade in the playoffs.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Italian Stallion



Joined: 04 Mar 2009
Posts: 112


PostPosted: Tue Jun 02, 2009 7:38 pm Post subject: Reply with quote
Interesting that some get more efficient. I hadn't thought of that angle, but it makes perfect sense. As a player gets into the flow of the game and take more shots, it probably increases their feel and accuracy the same as getting to the free throw line a lot can sometimes give a player a good feel there.

There must be some kind of curve where you improve for awhile and then start deteriorating as the quality of the shots start deteriorating.

What I'm really interested in what a guy like Kobe could do if his role wasn't to be the primary scorer.

Let's suppose he was paid to be efficient instead of trying to maximize his scoring/efficiency relationship. He could immediately eliminate all the marginal shots he now takes and let less skilled players on the team take them instead. His PPG would fall but his efficiency would rise. The net would probably be a negative for the team because in most cases the team is better off with Kobe taking the occasional and unavoidable tough shots, but I'd still like to know what that relationship was.

On the flip side, suppose the Knicks offered David Lee a 10 million dollar bonus to score over 25 PPG for the season. There's no way he could get enough layups, dunks, and other relatively easy shots around the basket to accomplish that. He would have to increase his outside shooting, take tougher and more contested shots around the basket etc... His FG% and TS% would certainly both fall. But I'd love to know by how much.
Back to top
View user's profile Send private message
Harold Almonte



Joined: 04 Aug 2006
Posts: 616


PostPosted: Tue Jun 02, 2009 9:44 pm Post subject: Reply with quote
There are people who would kill to refute your argument. The "diminishing return" argument could probably be applied to scorers the same as it is for rebounding teammates. Some people think everybody is able to score anytime into the flow of the offensive, and that a relative easy shot is always able to be created for and from anybody, and that every field goal attempt (forced or not, contested or not) have the same scoring probabilities and fall under the "shot created" nomenclature. Under this parameter, efficiency and productivity, like in a line of manufacture, are both the same.

Last edited by Harold Almonte on Tue Jun 02, 2009 10:26 pm; edited 2 times in total
Back to top
View user's profile Send private message
Mike G



Joined: 14 Jan 2005
Posts: 3613
Location: Hendersonville, NC

PostPosted: Tue Jun 02, 2009 9:47 pm Post subject: Reply with quote
Quote:
... what a guy like Kobe could do if his role wasn't to be the primary scorer.

Kobe does this for whiles, when he's in 'team' mode.

On the Olympics team, he might have had more leeway to do it; but that's not vs NBA competition.

There's no NBA team of super-scorers that Kobe could join and be this way all year, though. Facilitator, garbage man ... I can't see it.
_________________
`
36% of all statistics are wrong
Back to top
View user's profile Send private message Send e-mail
Chicago76



Joined: 06 Nov 2005
Posts: 98


PostPosted: Wed Jun 03, 2009 1:06 am Post subject: Reply with quote
Italian Stallion wrote:
Interesting that some get more efficient. I hadn't thought of that angle, but it makes perfect sense. As a player gets into the flow of the game and take more shots, it probably increases their feel and accuracy the same as getting to the free throw line a lot can sometimes give a player a good feel there.

There must be some kind of curve where you improve for awhile and then start deteriorating as the quality of the shots start deteriorating.

What I'm really interested in what a guy like Kobe could do if his role wasn't to be the primary scorer.

Let's suppose he was paid to be efficient instead of trying to maximize his scoring/efficiency relationship. He could immediately eliminate all the marginal shots he now takes and let less skilled players on the team take them instead. His PPG would fall but his efficiency would rise. The net would probably be a negative for the team because in most cases the team is better off with Kobe taking the occasional and unavoidable tough shots, but I'd still like to know what that relationship was.


Other than the "flow of the game" issue, there is an additional reason why you may see a player's usage decline and his efficiency decrease as well.

Using Bryant from your hypothetical: he could eliminate his sub-optimal shots, but that is easier to do earlier in the clock than later when he may be called upon to be the 1v1 creator to beat the clock. It might be suboptimal for his efficiency, but optimal for his team.

He could eliminate some poor shots, but the % of shots he takes at the end of th clock may actually go up and reduce his overall eFG/FG/TS%.

A good real life example of this occuring from a given space on the floor involves Larry Bird. His rookie year he took 143 3s--a lot in those days and hit 40.6% of them. The next four years, he took them only when he was wide open or when required due to score/shot clock. His attempts ranged from 52 to 77 a year and he never shot better than 29%. All those extra threes he didn't shoot? To equal his rookie season 3%, he would need to hit over 54% of those. His sub-optimal shots were actually pretty optimal. You might think his rookie season was a fluke, but he was given the green light to shoot more often from beyond the arc in his sixth season again, and for the next 4 years, his 3% was never lower than 40%.
Back to top
View user's profile Send private message
gabefarkas



Joined: 31 Dec 2004
Posts: 1313
Location: Durham, NC

PostPosted: Wed Jun 03, 2009 8:30 am Post subject: Re: Scoring, Usage, Efficiency, and Shot Distribution Reply with quote
Italian Stallion wrote:
Before anyone comments along these lines, I already understand that simply lowering a player's shot count won't change a thing. What has to change is the distribution of his shot attempts.

In other words if he take 20 shots made up of 4 dunks, 4 layups, 4 mid range jumpers, and 8 shots with tough coverage or from deep in two point territory, nothing is going to change if it's 2, 2, 2, and 4 for a total of 10 instead other than his PPG.

To become more efficient, there has to be a reduction in the 8 tough shots or an increase in the easy ones.

I would assume it's fairly easy to reduce tough shots, but not always in the best interests of the team if the alternatives are sometimes worse.

I would assume it's fairly difficult to increase the easy ones without an actual change in skill otherwise the player/team would already be doing it .

I would also assume it's very difficult for lower skilled, lower usage, but highly efficient players to increase their usage much without adding more difficult shots and decreasing their efficiency.

How to measure this tradeoff seems to be a key to understanding PPG vs. Scoring Effeciency. Am I off base?

Are you sure about that? I don't know if someone like Keith Bogans or Matt Maloney would think getting into the lane for those dunks and layups is going to be easier than shooting the 3's.
Back to top
View user's profile Send private message Send e-mail AIM Address
Harold Almonte



Joined: 04 Aug 2006
Posts: 616


PostPosted: Wed Jun 03, 2009 10:18 am Post subject: Reply with quote
That's why a players's scoring skill curve(s) could be more complicated than a fingerprint. There are even zones of those curves that would need to be assumed with very few data sample. Some players could produced narrow data samples on certain shot quality, and others not.
Back to top
View user's profile Send private message
Italian Stallion



Joined: 04 Mar 2009
Posts: 112


PostPosted: Wed Jun 03, 2009 4:52 pm Post subject: Reply with quote
"Using Bryant from your hypothetical: he could eliminate his sub-optimal shots, but that is easier to do earlier in the clock than later when he may be called upon to be the 1v1 creator to beat the clock. It might be suboptimal for his efficiency, but optimal for his team."

I understand the point you are making, but this is exactly the kind of thing I would like to eliminate from the results.

He is often called on in these situations, but suppose he flat out refused to be "the guy" and made a pass every time?

Then you'd have guys that should not be taking difficult outside shots and guys that should not be trying to create for themselves being forced to do so more often under difficult circumstances. Their efficiency would have to fall as a result.
Back to top
View user's profile Send private message
Italian Stallion



Joined: 04 Mar 2009
Posts: 112


PostPosted: Wed Jun 03, 2009 4:55 pm Post subject: Re: Scoring, Usage, Efficiency, and Shot Distribution Reply with quote
gabefarkas wrote:
Italian Stallion wrote:
Before anyone comments along these lines, I already understand that simply lowering a player's shot count won't change a thing. What has to change is the distribution of his shot attempts.

In other words if he take 20 shots made up of 4 dunks, 4 layups, 4 mid range jumpers, and 8 shots with tough coverage or from deep in two point territory, nothing is going to change if it's 2, 2, 2, and 4 for a total of 10 instead other than his PPG.

To become more efficient, there has to be a reduction in the 8 tough shots or an increase in the easy ones.

I would assume it's fairly easy to reduce tough shots, but not always in the best interests of the team if the alternatives are sometimes worse.

I would assume it's fairly difficult to increase the easy ones without an actual change in skill otherwise the player/team would already be doing it .

I would also assume it's very difficult for lower skilled, lower usage, but highly efficient players to increase their usage much without adding more difficult shots and decreasing their efficiency.

How to measure this tradeoff seems to be a key to understanding PPG vs. Scoring Effeciency. Am I off base?

Are you sure about that? I don't know if someone like Keith Bogans or Matt Maloney would think getting into the lane for those dunks and layups is going to be easier than shooting the 3's.


Understood.

Obviously, each player has his own unique skill set. What is easy or difficult for them "to get" will vary, but there is no doubt that dunks and layups are easier for everyone when the are available than contested jump shots from just inside the arc.
Back to top
View user's profile Send private message
Italian Stallion



Joined: 04 Mar 2009
Posts: 112


PostPosted: Wed Jun 03, 2009 5:19 pm Post subject: Reply with quote
Harold Almonte wrote:
There are people who would kill to refute your argument. The "diminishing return" argument could probably be applied to scorers the same as it is for rebounding teammates. Some people think everybody is able to score anytime into the flow of the offensive, and that a relative easy shot is always able to be created for and from anybody, and that every field goal attempt (forced or not, contested or not) have the same scoring probabilities and fall under the "shot created" nomenclature. Under this parameter, efficiency and productivity, like in a line of manufacture, are both the same.


I am mathematically inclined and love stats, but am not knowledgeable enough about stats to participate at the highest levels of these debates.

I was also a basketball player when I was a kid. I was limited athletically because of my size, so I never got really good, but I think I understood the game fairly well and probably would have made a good coach if my life went in that direction.

In these various discussions and debates, I often run into people that I believe fall strongly into one camp or the other and don't really blend the two well.

There is a quote from Charlie Munger (famous investor) about this kind of phenomenon.

"To a man with only a hammer, everything looks like a nail".

What that means is that people want to believe things that fit into their neat little model of the world even when experience, common sense, some of the evidence etc.. says they are wrong. They will ignore what refutes them and focus on what supports them and remain delusional no matter how bright they are.

To me it's obvious that the relationships I am talking about exist. I just have no way of measuring all the factors or the results etc...

When I was a kid, I averaged 6 PPG in CYO. I did so by being very selective and only taking shots I knew I could make a decent % of the time. The best player on the team averaged about 15 PPG doing exactly the same thing. There is no freaking way on earth I could have averaged 15 PPG simply by being more aggressive and taking more shots. I didn't have the skill set to get more good shots and would have had to add in bad shots. If I added enough, I might have eventually got to 15 PPG, but my teammates would have beaten the shit out of me before that became a realistic possibility because my efficiency would have been a joke. If the best player suddenly stopped shooting jumpers because he wanted his TS% to be higher, we all would have screamed at him because idiots like me would have had to take more shots that were bad for us.

Even as children we understood good shots, bad shots, how they varied from player to player depending on their skill set, how the number of good ones they could get varied depending on their skill set etc...

I think some stats guys are like a man with a hammer, but I don't see much evidence of that on this board. This seems like a great place to find the balance and intellect I enjoy. Very Happy

mathom



Joined: 16 Feb 2005
Posts: 19


PostPosted: Fri Mar 04, 2005 3:58 pm Post subject: free throws, efficiency, possession usage, random thoughts Reply with quote
In looking over some player stats, I began wondering about the value of 2p and 3p shots or if anyone has studied the affect of 3p/2p FGs on FTs. It seems to me as though players are moving more and more toward the 3pt line and many players (and coaches) might be doing themselves a disservice because it lowers the number of opportunities they have to get to the line. I don't doubt there are some players that are costing themselves efficiency in this way even though their 3pFG% may be quite respectable. It seems also that getting to the free throw line is indirectly more valuable as well, by putting players in foul trouble - not unlike the value of pitches taken per at bat in baseball. In a simplistic comparison, the efficiency of a 3p shot would be 3pFG% *1.5 vs 2pFG%. But it seems to me a better comparison then is 3pFG%*1.5*P1 vs 2pFG%*P2, where P1/P2 is some constant that measures a player's ability to get to the line when taking a FG of whatever type. To determine P1/P2, I think some research will need to be done regarding the rate at which players are fouled on 3p shots first. Once you have that piece of knowledge, you could use the ratio of (FTA-FTA from fouls on 3pers)/2pFGA, and determine the efficiency of each type.

However, I think a better method than FTA/2pFGA would be to introduce another constant, one that is universal and relates the number of FTs to the length of time a player possesses the ball. I would say the large majority of non-shooting fouls that result in FTs are done against the player holding the ball. While you could probably argue that some players do have an ability to draw non shooting fouls more than others, I dont think the variance will be that large, and that using a league wide constant and applying it to all players (or at the least, players of the same position) would give a very good approximation. In order to determine that sort of constant however, some data collection would need to be done on non shooting fouls that resulted in FTs.

I have a feeling that if someone were to go through all of the work and calculations, it would turn out that 2pFGs are actually more efficient than 3pFGs in many cases, and that a lot of players would be better served driving to the basket than taking 3pFGs, even when on the surface it appears their 1.5*3pFG% is higher than their 2pFG%. (And it might help explain why the championship teams always seem to be teams that have a quality bigman and get to the line).

One last parting thought: should technical free throws be subtracted from a player's PPG average when considering efficiency? If a player takes 1 tech FT every other game, and sinks say 85%, it would increase their PPG avg by more than .4 a game, which depending on the number of possessions the player uses otherwise could swing their efficiency substantially. I tend to think maybe we should, but I can see an argument being made both ways.
Back to top
View user's profile Send private message
kjb



Joined: 03 Jan 2005
Posts: 861
Location: Washington, DC

PostPosted: Fri Mar 04, 2005 5:17 pm Post subject: Reply with quote
Here's a pretty straighforward graph of where points have come from in the NBA since 73-74. Each category is expressed as a percentage of total points. Note that the percentage of pts from 2pt fgs has declined, has gone up from 3pt fgs, and remained relatively constant from FT. Pts from the FT line may be declining a little since 2000, but it's not a big move.


Back to top
View user's profile Send private message AIM Address Yahoo Messenger
mathom



Joined: 16 Feb 2005
Posts: 19


PostPosted: Fri Mar 04, 2005 8:01 pm Post subject: Reply with quote
Actually, I'd say this graph is rather deceptive Wink

Even though it looks like FT have remained stable, FTA and FTA/FGA have both decreased rather significantly since 1980 (by about 15% and 10% respectively).

http://www.geocities.com/torch772/ssnbastats.htm

If you click the historical shooting tab at the bottom, it has some stats on shooting over the years.
Back to top
View user's profile Send private message
kjb



Joined: 03 Jan 2005
Posts: 861
Location: Washington, DC

PostPosted: Fri Mar 04, 2005 10:20 pm Post subject: Reply with quote
I don't think it's deceptive at all. Smile

It's a measure of where pts are coming from -- as a percentage of total points. FTA/FGA may be down (I haven't looked at that measure specifically), but pts from FT has remained pretty stable through the years, as has FT%.
Back to top
View user's profile Send private message AIM Address Yahoo Messenger
mathom



Joined: 16 Feb 2005
Posts: 19


PostPosted: Fri Mar 04, 2005 10:53 pm Post subject: Reply with quote
Right, but that doesn't really tell me much.. if FTA/FGA is going down, but my efficiency is also going down and I score less points, then my FTA/PPG can be relatively stable, and it masks what's really going on.
Back to top
View user's profile Send private message
kjb



Joined: 03 Jan 2005
Posts: 861
Location: Washington, DC

PostPosted: Fri Mar 04, 2005 11:06 pm Post subject: Reply with quote
mathom wrote:
Right, but that doesn't really tell me much.. if FTA/FGA is going down, but my efficiency is also going down and I score less points, then my FTA/PPG can be relatively stable, and it masks what's really going on.


It's FTM/PTS, but I think I understand what you're saying.

The study you're talking about conducting could probably be done using the play-by-play data. Figure out how many times players get fouled attempting 3pt shots, how many times they shoot free throws for non-shooting fouls (loose ball, reaching in, etc.) and how many times they get fouled attempting 2pt shots.

I think if we're going to look at whether the rise of the 3pt shot is reducing FTA, we ought to be looking at FTA/POS -- what share of possessions come from free throws through the years. Comparing it to FGA is of lesser value (in my opinion) because not every FTA is related to an FGA.
Back to top
View user's profile Send private message AIM Address Yahoo Messenger
gabefarkas



Joined: 31 Dec 2004
Posts: 1312
Location: Durham, NC

PostPosted: Mon Mar 07, 2005 2:57 pm Post subject: Reply with quote
WizardsKev wrote:
I don't think it's deceptive at all. Smile

It's a measure of where pts are coming from -- as a percentage of total points. FTA/FGA may be down (I haven't looked at that measure specifically), but pts from FT has remained pretty stable through the years, as has FT%.


well, 3 point shots are worth more than 2 point shots. so, if scoring remains constant, a .5% uptick in 3 pointers doesn't correspond directly to a .5% downtick in 2 pointers.

same with FT's.

i'd be curious to see a graph with 3 lines: X/(3PM+2PM+(FTM*0.44)) where X is each of the terms in the denominator.
Back to top
View user's profile Send private message Send e-mail AIM Address
kjb



Joined: 03 Jan 2005
Posts: 861
Location: Washington, DC

PostPosted: Mon Mar 07, 2005 3:40 pm Post subject: Reply with quote
gabefarkas wrote:
WizardsKev wrote:
I don't think it's deceptive at all. Smile

It's a measure of where pts are coming from -- as a percentage of total points. FTA/FGA may be down (I haven't looked at that measure specifically), but pts from FT has remained pretty stable through the years, as has FT%.


well, 3 point shots are worth more than 2 point shots. so, if scoring remains constant, a .5% uptick in 3 pointers doesn't correspond directly to a .5% downtick in 2 pointers.

same with FT's.

i'd be curious to see a graph with 3 lines: X/(3PM+2PM+(FTM*0.44)) where X is each of the terms in the denominator.


I don't understand what you're asking. What's X ?
Back to top
View user's profile Send private message AIM Address Yahoo Messenger
gabefarkas



Joined: 31 Dec 2004
Posts: 1312
Location: Durham, NC

PostPosted: Mon Mar 07, 2005 8:34 pm Post subject: Reply with quote
right now your 3 lines represent:

Points from 2FGA
Points from 3FGA
Points from FTA

i'd be curious to see:

2FGM
3FGM
0.44*FTM
Back to top
View user's profile Send private message Send e-mail AIM Address
kjb



Joined: 03 Jan 2005
Posts: 861
Location: Washington, DC

PostPosted: Tue Mar 08, 2005 4:59 am Post subject: Reply with quote
gabefarkas wrote:
right now your 3 lines represent:

Points from 2FGA
Points from 3FGA
Points from FTA

i'd be curious to see:

2FGM
3FGM
0.44*FTM


I understand now. I'll post something when I get back to my spreadsheet.
Back to top
View user's profile Send private message AIM Address Yahoo Messenger
gabefarkas



Joined: 31 Dec 2004
Posts: 1312
Location: Durham, NC

PostPosted: Tue Mar 08, 2005 9:34 am Post subject: Reply with quote
awesome. i'm looking forward to it. if you want to share the raw data at all, i'd be happy to take a crack at it.

my thinking was this: since 2FGM, 3FGM and FTM all affect Total Points differently, a change in each affects Total Points (and thus each's % of "Source of Points") in a different magnitude.
Back to top
View user's profile Send private message Send e-mail AIM Address
kjb



Joined: 03 Jan 2005
Posts: 861
Location: Washington, DC

PostPosted: Tue Mar 08, 2005 11:41 am Post subject: Reply with quote
Okay -- here's the graph you requested:



Very similar to the pts source graph I posted above. Free throw makes are slightly higher in recent years than they were in the 70s even as 2pt makes have given way to 3pt shooting. It's also interesting that there was a slight uptick in FTs when 3pt shooting surged abruptly in the 90s. I don't remember precisely, but that was probably when the league went to the shorter 3pt line.

I'd be happy to share the raw data. Just email me at kevinbroom@realgm.com and I'll send back the spreadsheet.
Back to top
View user's profile Send private message AIM Address Yahoo Messenger
mathom



Joined: 16 Feb 2005
Posts: 19


PostPosted: Tue Mar 08, 2005 1:50 pm Post subject: Reply with quote
Using FTM/2pFGM/3pFGM etc clouds the issue because FGM is a function of both FGA and FG%. If FG% over the given period were constant, it would not be an issue. Let me give a quick example to demonstrate:

Say 10000 total shot attempts were taken in the league in 1970 (all 2pers since its 1970). Suppose also that a player gets to the line once for every 4 FGA, ie 2500 FTA were taken in 1970. If you say that 50% of FGA are made and 75% of FTA are made, then FTM/FGM+FTM = .1579. Now, flash forward to present, where tempo has died down considerably, where say 8000 total shots are taken. Suppose 7000 are 2pFGs and 1000 are 3pFGs. You still get 1 FTA per 4 FGA, but now also suppose that you get 1 FTA per 10 3pFGA. Suppose that 2pFG% has dropped to 45%, 3pFG% is 35%, and FT% is still 75%. That means there were 1387.5 FTM, and 7000*.9 + 1000*1.05 + 1850*.75 points scored, so FTM/points = .1588. FTM are now a higher portion of total points scored, but the fact that FTA/FGA has decreased is entirely masked.

While I did make up the numbers, the effect is the same. If you want to find a correlation between 2pFGA, 3pFGA and FTA, using FTM/2pFGM/3pFGM is deceiving because 2pFG%/3pFG% have also changed over time and will give you an incorrect view of things.

Edit:
Quote:
It's FTM/PTS, but I think I understand what you're saying.

The first one I did mean FTA/FGA, and yeah, the second I meant FTM/PPG, and not FTA/PPG.
Back to top
View user's profile Send private message
kjb



Joined: 03 Jan 2005
Posts: 861
Location: Washington, DC

PostPosted: Tue Mar 08, 2005 2:39 pm Post subject: Reply with quote
mathom: I just took a look at (.44*FTM)/oting has increased and 2pt shooting has decreased.

I also just took a look at FTA/FGA, which is dopossessions as one more way of looking at what you're talking about in the first post of this thread. The data looks basically the same as the other graphis I posted -- that free throws have remained fairly constant even while 3pt shown over the past few years, but which also went up when 3pt shooting spiked to its highest level in the 90s. You may be right that teams & players may be costing themselves efficiency with 3pt attempts. I'm asking Roland if he has data on foul rates by shot type. Of course, this data wouldn't be historical so we're still kinda in the same place we're in now.

You could well be right about this, but I think we're going to need more information to prove it.
Back to top
View user's profile Send private message AIM Address Yahoo Messenger
Roland_Beech



Joined: 14 Jan 2005
Posts: 43


PostPosted: Wed Mar 09, 2005 5:05 pm Post subject: Reply with quote
Well...

2 pt shots draw fouls on 14.7% of tries
3 pt shots draw fouls on 0.9% of tries

where 'tries' is FGA + shooting fouls drawn - 'and 1' shooting fouls

eg for 2's there are 143,528 FGA and 24,119 fouls that led to free throws, but 3806 of them were 'and one' so 163,841 tries...

looks like it all adds up pretty well with official NBA stats when you add in some technical free throws...

unfortunately when a player is fouled and misses the shot, I don't know the distance of the shot at this time (for all games), other than it being a 2 versus 3...

that distance of the foul is something along with distance of a turnover that we are charting and looking to chart to better understand pts/poss type stuff from various distances/location. I do have offensive rebound rates by shooting distance too, so it all kind of goes together if you get the foul/turnover distance detail...
Back to top
View user's profile Send private message Visit poster's website

mathom



Joined: 16 Feb 2005
Posts: 19


PostPosted: Wed Mar 09, 2005 7:14 pm Post subject: Reply with quote
hmm, (24119-3806)/143528 = .142, how are you getting .147? and I assume 163,841 is a typo, but even off by a factor of 10 it seems to be incorrect, so I'm not sure what's going on there.
Back to top
View user's profile Send private message
Ben F.



Joined: 07 Mar 2005
Posts: 391


PostPosted: Wed Mar 09, 2005 8:03 pm Post subject: Reply with quote
He's not subtracting 3,806 from the numerator, only from the overall number of tries. Although I think your way is more right.
Back to top
View user's profile Send private message
mathom



Joined: 16 Feb 2005
Posts: 19


PostPosted: Thu Mar 10, 2005 1:37 am Post subject: Reply with quote
I dont think so.. unless Im missing something overall tries = 163.8k - 3.8k = 160k, which doesn't match with any of his numbers, and obviously there are not more FTAs than FGAs.
Back to top
View user's profile Send private message
Roland_Beech



Joined: 14 Jan 2005
Posts: 43


PostPosted: Thu Mar 10, 2005 2:54 am Post subject: Reply with quote
...

2-pt shots

fga 143,528
shooting fouls 24,119...of which 3806 were of the 'and one' variety

since 'and one' fouls already include a made field goal, they are already being counted as a try under fga, so to avoid double counting you subtract them from the total...[missed field goals that draw a foul are not charged as an FGA...]

so, tries = 143,528 + 24,119 - 3,806 = 163,841
so foul drawing % of fouls / tries = 24119/163841 = 14.7%
Back to top
View user's profile Send private message Visit poster's website
mathom



Joined: 16 Feb 2005
Posts: 19


PostPosted: Thu Mar 10, 2005 3:05 am Post subject: Reply with quote
Ah, ok. I totally missed the entire thing about tries. I was taking everything to be FGA. Thanks for clarifying.