APBRmetrics

The discussion of the analysis of basketball through objective evidence, especially basketball statistics.
It is currently Tue Jul 29, 2014 10:52 pm

All times are UTC




Post new topic Reply to topic  [ 2 posts ] 
Author Message
PostPosted: Fri Apr 15, 2011 10:32 am 
Offline

Joined: Thu Apr 14, 2011 11:10 pm
Posts: 2253
Author Message
mateo82



Joined: 06 Aug 2005
Posts: 211


PostPosted: Fri Oct 14, 2005 3:32 pm Post subject: Opponents Points Per 100 Possessions = definitive team stat Reply with quote
I am convinced. Living in a defensive era, there is no stat more important. Since people don't talk a lot about team statistics here, I thought I'd bring up my favorite. Using basketball-reference.com, here are the rankings of every team that has made an NBA finals since 1989 in opponents points per 100 possessions:

2005 San Antonio Spurs (1st) over Detroit Pistons (3rd)
2004 Detroit Pistosn (2nd) over Los Angeles Lakers (8th)
2003 San Antonio Spurs (3rd) over New Jersey Nets (1st)
2002 Los Angeles Lakers (7th)* over New Jersey Nets (2nd)
2001 Los Angeles Lakers (19th)* over Philadelphia 76ers (5th)
2000 Los Angeles Lakers (1st) over Indiana Pacers (13th)
1999 San Antonio Spurs (1st) over New York Knicks (3rd)
1998 Chicago Bulls (3rd)* over Utah Jazz (15th)
1997 Chicago Bulls (4th)* over Utah Jazz (11th)
1996 Chicago Bulls (1st) over Seattle Sonics (2nd)
1995 Houston Rockets (12th)* over Orlando Magic (13th)
1994 Houston Rockets (4th) over New York Knicks (1st)
1993 Chicago Bulls (3rd)* over Phoenix Suns (9th)
1992 Chicago Bulls (4th)* over Portland Trailblazers (1st)
1991 Chicago Bulls (6th) over Los Angeles Lakers (4th)
1990 Detroit Pistons (1st)* over Portland Trailblazers (3rd)
1989 Detroit Pistons (3rd) over Los Angeles Lakers (7th)

* an asterisks for repeat winners

Note that repeat champions often are low. Why this is, is anyone's guess. Maybe the correlation means nothing, maybe it does. Maybe repeat winners, like Shaq used to say, take the regualr season off. I don't know. Just thought i'd throw out this wonderful stat for discussion.

Last edited by mateo82 on Fri Oct 14, 2005 6:32 pm; edited 1 time in total
Back to top
View user's profile Send private message
jambalaya



Joined: 30 Jan 2005
Posts: 282


PostPosted: Fri Oct 14, 2005 3:54 pm Post subject: I agree Reply with quote
I agree it is a very important stat. 22 of the 34 finalists in 17 years had a top5 defense on this measure or 65% of all finalists, and 82% were top 10. The average rank was just about 5th.

Last edited by jambalaya on Fri Oct 14, 2005 4:24 pm; edited 2 times in total
Back to top
View user's profile Send private message
mateo82



Joined: 06 Aug 2005
Posts: 211


PostPosted: Fri Oct 14, 2005 3:59 pm Post subject: Reply with quote
All winners since 1994 (except repeats) have been top 3. (1994 Houston is listed as #4 but are actually tied for #3 with Seattle).
Back to top
View user's profile Send private message
jambalaya



Joined: 30 Jan 2005
Posts: 282


PostPosted: Fri Oct 14, 2005 4:08 pm Post subject: Reply with quote
"Except repeats" is an important caveat. The only case with a defense below the median for all teams that year was the defending Lakers who picked it up as Shaq said and repeated in 2001.

It would be interesting to see the data on playoff defensive performance on this measure and see how it matches up. That would seem a more valuable measure since the regular season is different and ultimately not as important.

On the regular season data, the winners averaged 4.7th ranked and the losers 5.5th ranked. The better ranked team won 10 of 17 of the match ups, nice but not a large margin, so it may say more about getting there than winning the championship series. I suspect the team with the higher offense - defense point differential per 100 possession would be the winner more often and you can rank higher on that several ways.

Last edited by jambalaya on Fri Oct 14, 2005 4:29 pm; edited 3 times in total
Back to top
View user's profile Send private message
mateo82



Joined: 06 Aug 2005
Posts: 211


PostPosted: Fri Oct 14, 2005 4:26 pm Post subject: Reply with quote
jambalaya wrote:
"Except repeats" is an important caveat.


It is, but because all repeat winners had a lower ranking than their first year (without exception), maybe this means that the correlation is not only coincidence. but maybe it is.

Quote:
It would be interesting to see the data on playoff defensive performance on this measure and see how it matches up. That would seem a more valuable measure since the regular season is different and ultimately not as important.


I would love to see this as well. I also wonder if this stat is compiled at the end of seasons or as the season goes on. It seems like it would be meticulous to do it myself.

Quote:
On the regular season data, the winners averaged 4.7th ranked and the losers 5.5th ranked. The better ranked team won 10 of 17 of the match ups, nice but not a large margin, so it may say more about getting there than winning the championship series. I suspect the team with offense - defense point differential per 100 possession would be the winner more often and you can rank high on that several ways.


Well, I wouldn't go that far. Just getting there isn't necessarily enough. If you get without a very elite defensive team then you stand no chance, good offense or not.

I checked the median, thinking the very bad 2001 lakers might have dragged down the winners down in mean but the median is 3 for both.
Back to top
View user's profile Send private message
mateo82



Joined: 06 Aug 2005
Posts: 211


PostPosted: Fri Oct 14, 2005 4:30 pm Post subject: Reply with quote
As a side issue, when compiling this, I was shocked by how overwhelmingly dominant the 1994 Knicks were on defense. 97.6, which is ridiculously good. Almost 4 points better than the Rockets. And, if I recall correctly, no one else was even close in the years I covered. I think 2003 Nets were something like 98.2... but that's the closest it ever came (except 1999, when about 10 teams were as good or better.... so i'm rejecting that year altogether, that just doesn't make sense).

It kind of reinforces the idea that NY choked hard that year.
Back to top
View user's profile Send private message
jambalaya



Joined: 30 Jan 2005
Posts: 282


PostPosted: Fri Oct 14, 2005 4:33 pm Post subject: Reply with quote
A look at basketball reference's playoffs index suggests you could manually compute team offensive performance from that but not defense. Perhaps it could be enhanced to allow easier calculations. You probably could assemble from espn.com or nba.com playoff data but maybe not all the way back. You could get three years from 82games.com and just one from dougsstats.

I agree being top5 defense is very very important, top 10 close to an absolute requirement to win a championship (or have a Shaq or Oluwajon in take care of the paint in the clutch). But still you can be a lot of different good combinations of top 5-10 defense and top 5-10 offenses and get it done. And your offense might be somewhat better than your defense (you don't have to be defensive-biased, netting the two measures). But it is true that in 13 of 17 most recent cases the winner was top5.

If I was planning on a championship, I'd plan on being top 5 on defense if at all possible, using every means available. Phoenix's off season moves and the players/coaches greater efforts could move them up from 16th on defensive efficency at http://www.knickerblogger.net/stats/d_de.htm
to perhaps top 10 but can they move into top5? I'd be somewhat surprised. Both Brian Grant and Kurt thomas give up more than 50% FG to opposing centers. And no meaningful improvement on that over Amare and the departed Hunter at that spot. And Raja Bell was actually a little worse than the departed Q Richardson and Joe Johnson. So I guess it better get done on chemistry and greater effort because I am not sure they really upgraded defensive talent that much- at least by this one measure, which admittedly isnt sufficient by itself to answer the whole defensive question- even if they upgraded defensive "reputation".
Back to top
View user's profile Send private message
mateo82



Joined: 06 Aug 2005
Posts: 211


PostPosted: Fri Oct 14, 2005 6:35 pm Post subject: Reply with quote
Here are the stats of a team's own points per 100 possessions so we can compare to their opponents:

2005 San Antonio Spurs (8th) over Detroit Pistons (16th)
2004 Detroit Pistons (18th) over Los Angeles Lakers (6th)
2003 San Antonio Spurs (6th) over New Jersey Nets (19th)
2002 Los Angeles Lakers (3rd)* over New Jersey Nets (16th)
2001 Los Angeles Lakers (2nd)* over Philadelphia 76ers (13th)
2000 Los Angeles Lakers (6th) over Indiana Pacers (1st)
1999 San Antonio Spurs (10th) over New York Knicks (26th)
1998 Chicago Bulls (9th)* over Utah Jazz (1st)
1997 Chicago Bulls (1st)* over Utah Jazz (2nd)
1996 Chicago Bulls (1st) over Seattle Supersonics (6th)
1995 Houston Rockets (8th)* over Orlando Magic (2nd)
1994 Houston Rockets (16th) over New York Knicks (14th)
1993 Chicago Bulls (3rd)* over Phoenix Suns (1st)
1992 Chicago Bulls (1st)* over Portland Trailblazers (7th)
1991 Chicago Bulls (1st) over Los Angeles Lakers (5th)
1990 Detroit Pistons (11th)* over Portland Trailblazers (8th)
1989 Detroit Pistons (8th) over Los Angeles Lakers (1st)


* an asterisks for repeat winners

Winners are average of 6.5 and losers are 8.5

Obviously defense is much more important. Aside from the averages: just look at a few examples. The 2004 Lakers, for example, who were quite balanced. 8th in defense, 6th in offense. So if being a top10 in both categories is the strategy... it doesn't always work. The Pistons were a miserable 18th in offense but 2nd in defense.

I think focusing on that top 3 or 4 defense is the way to go.

Even among teams that were top 5 in defense and top 10 in offense, there are quite a few examples of losers that fit that description: Sonics, 92 Blazers, 91 Lakers, 90 Blazers all were top 5 defensive and top 10 offense and lost. That's how unimportant "balance" seems to be. Dominant defense is what works.
Back to top
View user's profile Send private message
jambalaya



Joined: 30 Jan 2005
Posts: 282


PostPosted: Fri Oct 14, 2005 7:00 pm Post subject: Reply Reply with quote
Top 5 offense, top 10 defense isnt always enough, I agree.

But balance does still matter. In 12 of the 17 cases the most balanced team, computed as absolute value of rank on defense - rank on offense, won the championship match up (with one tie):

differential = defense rank - offensive rank

winner loser
2005 7 13
2004 16 -2
2003 3 18
2002 -4 14
2001 -17 8
2000 5 -12
1999 9 23
1998 6 -14
1997 -3 -9
1996 0 4
1995 -4 -11
1994 12 13
1993 0 0
1992 -3 6
1991 -5 1
1990 10 5
1989 5 -6

avg 2.176470588 3

It is noteworthy that both are indeed defensively biased, in line with your argument that defense is more important (and I agree with that to a point, but add the data about balance instead of rejecting it),

but the winner, on average was less defensively biased than the loser, slightly [/u]

and "balance" is a slightly better predictor of the winner year to year (12 of 17) compared to defensive ranking alone (10 of 17 called correctly).

In 2002 the lakers were #3 O / #7 D and won the championship, so a top5 offense, top ten but not top 5 defense can win. And in another case it also happened 1991 Bulls #1 O / #6 D. In two other cases an offensively biased team also won 1997 bulls #1 0 / #4 D 1992 Bulls #1 / #4 D and came just short a little short of meeting this description. It is fairly rare, 2 cases and 2 close calls in 17 years, but these four examples help counterbalance the ones you selected were they failed.

Still it appears true that a top5 defense / top 10 offense is a better strategy with 8 winners (if you count one case where the offense was even worse the 94 Rockets). And I agree would seem the better strategy than the reverse.

Nonetheless, overall 7 of the 17 winners were "offensively biased" by this simple net rank measure. So good offense rank is usually also important. The amount of importance varies.

It can be outweighed sometimes by great D as with the Pistons. In three of the 4 cases where a team was 10 or higher on offense yet won the championship they had a #1 or #2 defense with the fourth case being #4.

But there are two cases of a team 10 or higher on defense still winning, in one case they were #2 on offense, in the other #8.

As for the top4 in defense model, such teams appeared 22 times, impressive. They won 13 times, also impressive. Certainly top5 defense is a great strategy, maybe the best.

But 8 times top4 defenses met each other, in 5 cases the lesser defense won (even in two cases where their offense was actually worse than the opponent), in 3 cases the better defense won. A small margin, but another piece of the record.

By comparison top 4 offenses appeared half as often, 11 times but won 7 times. Less often, but a slightly better winning percentage 64% to 59%, still on balance I'll stay with the top 4 defense over top 4 offense as a better strategy. A top 4 "balance" team also appeared 11 times and won 7 of them.

If I could acheive any of the three top 4 ranks I'd run with it. In fact the three different top 4 strategies account for 16 of the 17 examples.

There were five cases where defense and offense top4 status overlapped, and four of them were the strongest dynasty of the era Chicago Bulls and the other was the breakthru, start of the modern defensive era, 89 Pistons. There are no cases of a team with a top4 on offense and defense not winning the championship in this period though there are 4 cases of failure that are close, with a top5 -7 rank in one field while top4 in the other.

There are 5 cases where teams where top4 on offense and balance on won a championship and no examples of losses; but in four of the cases they were also top4 in defense to complete the trifecta (the only exception being the 2002 Lakers with a good but still #7 defense in a repeat situation so it was probably still better for the playoffs).

There are 6 cases on top4 in defense and balance that won the championship, but 2 cases where they failed to win a championship- the 96 Sonics, 91 Lakers, and two more other cases where they failed to win a championship with a top 4 in defense but with a top5-6 finish in balance.

I think it would be fairly safe to say your strategy ought to be on having at least one of the three top4 ranks. I dont think any lesser rank strategy on offense, defense or balance is likely to cut it very often. And it wouldnt hurt to try to have two or at least to improve your relatively weak area as much as you can so that it has less chance of being what keeps you from winning with your strength.

In one case a top 4 defense meet a top 4 balance team who was not also a top4 defense, the top 4 defense won - 2004 Pistons. In another case though, a top 3 balance team beat a top4 defense team - 1992 Bulls (But they did have a just outside the top4 standard, top 5 defense to add to their #1 offense and effective balance; while the Blazers had a top7offense to go with thier slightly superior #4 D . Both team good cases for balance working well.)

In three cases a top 4 defense team that was not also top4 on offense meet a top4 offense that was not also top4 on defense and the better defense won 2-1.

My overall point, besides trying to give a fair look to as many angles as I could think of, is that the record has enough diversity that it should be noted / considered along with "the defensive wins a lot" evidence. There is some flexibility in how defense, offense and balance play out and who wins. Defense seems to have the edge but by itself it can not claim all the credit or all the championships.

Last edited by jambalaya on Fri Oct 14, 2005 9:40 pm; edited 7 times in total
Back to top
View user's profile Send private message
mateo82



Joined: 06 Aug 2005
Posts: 211


PostPosted: Fri Oct 14, 2005 8:18 pm Post subject: Reply with quote
good post, i'll respond to it later.

just to let you know, in my original post I had mistakenly listed the 1992 Suns as 1st in defense when it fact they were 9th. I editted in the correct value, but if that messed up any of your data you might consider that.
Back to top
View user's profile Send private message
jambalaya



Joined: 30 Jan 2005
Posts: 282


PostPosted: Fri Oct 14, 2005 8:29 pm Post subject: Reply Reply with quote
I think it was a good topic, good exchange, pretty full consideration of the data and alternate models. Maybe others will add views later.

The error might affect some of my statements by one case, but not others, and I probably won't go back thru and revise everything because most of the points will remain close to the same in meaning.
Back to top
View user's profile Send private message
jambalaya



Joined: 30 Jan 2005
Posts: 282


PostPosted: Fri Oct 14, 2005 11:12 pm Post subject: Leading contenders for this coming year Reply with quote
Based on this concept that you need to be a top 4 offense, defense or balance team, I turned to last year's playoff data as a relevant (play at end of year / when it is really on the line, stats not influenced by performance against nonplayoff teams), not yet examined database to start a consideration of top contenders for this year (despite uneven playoff competition which diostorts). Changes to this year's roster to be briefly considered next.

2005 Playoff data

Team Scoring Team Defense
Team Pts/Game Team Pts/Game
1 PhoenixSuns 112.0 1 DetroitPistons 85.6
2 DallasMavericks 105.2 2 IndianaPacers 86.9
3 MemphisGrizzlies 102.8 3 BostonCeltics 88.3
4 ChicagoBulls 102.5 4 SanAntonioSpurs 92.6

Team Scoring Differential
Team Pts/Game
1 DetroitPistons 4.7
2 MiamiHeat 4.3
3 SanAntonioSpurs 4.3
4 PhoenixSuns 4.2

San Antonio, Detroit and Phoenix appear on two lists, a good starting point for being elite. San Antonio got better. Detroit probably hasnt aged or changed enough to slip, if Flip handles the challenge. Phoenix changed, maybe not better on talent, but more experienced with each other and being a contender so I think it safe to include them too.

The number 2's in each of the categories Dallas on offense, Indiana on defense and Miami on balance seem like good starters for at least being second tier elite if not equal to the first tier. Miami changed the most but probably a net improvement. Indiana got Artest and another guard. Dallas may have slipped a bit.

A third tier seems to include Chicago, Boston, Memphis who made appearances in the top4s. You could argue each slipped down, but with Chicago it is debatable and with more experience for the young guys offsetting they are most likely to still be competitive for the elite 8.

Seattle and Washington were elite 8 last year but not on these specific playoff measures. I'd put them third tier to start. Both had some changes, probably mild net losses.

Any teams likely to move up close enough to consider them as well? Denver, Houston, Sacramento, Cleveland, New Jersey.

Thats 16 teams mentioned. Using top 4 in something as the standard for weeding, I'd say it is San Antonio, Detroit and Phoenix for sure, probably Dallas, Indiana and Miami. I'd give Denver and Chicago as more likely final wildcards for top4 on defense or balance. Most or all of the top 4 next playoffs are likely to come from these elite 8. Washington and Houston seems like the next most likely bubble teams to me.

I won't try to pick the top4 or top 2 yet but by all-star game the odds of guessing right then will improve quite a bit.
Back to top
View user's profile Send private message
Johnny Slick



Joined: 15 Oct 2005
Posts: 45
Location: Seattle

PostPosted: Sun Oct 16, 2005 3:39 am Post subject: Reply with quote
First off, I kept meaning to get around to coming here, and I'm glad to finally get over. FWIW I posted on the APBR listserv under the pseudonym John Craven (okay, maybe that was more of a "nym"). Anyway...

The problem I have with a stat like this is that really to me basketball is composed of six basic components, not two:

- Scoring points efficiently (measured in terms of points divided by shot attempts, counting a free throw attempt as .4 of one per what everyone else is doing WRT possessions)
- Avoiding turnovers
- Grabbing offensive rebounds
- Grabbing defensive rebounds
- Forcing turnovers
- Forcing your opponent into inefficient shooting.

We're lumping the first 3 into offense and the last 3 into defense. I don't think that's right. It's kind of like baseball sabermetricians grouping pitching together with defense. Do rebounds feed off of good D? Sure, but defensive turnovers should also lead to a better scoring-efficiency rating and I don't see anybody trying to look at *that*.

Anyway, just for fun I'm going to list teams' ratings by each of these metrics, year by year, then come up with a horribly flawed statistic that attempts to quantify what "percentage" of the game is offense, defense, and so on. Anyway, here goes. Stats go team name (scoring-TOs-ORs / defensive scoring-TOs-DRs).

1989: Detroit (1-17-7/3-19-4) over LA Lakers (8-10-12/6-21-6)
1990: Detroit (13-8-5/1-15-4) over Portland (15-10-2/11-6-3)
1991: Chicago (4-3-4/14-4-9) over LA Lakers (8-6-15/1-23-13)
1992: Chicago (6-2-5/8-7-4) over Portland (8-12-4/10-5-2)
1993: Chicago (16-1-1/18-3-12) over Phoenix (1-18-10/10-14-6)
1994: Houston (5-20-27/1-22-5) over New York (13-22-8/2-2-1)
1995: Houston (4-19-26/4-20-19) over Orlando (3-7-3/5-24-15)
1996: Chicago (10-1-1/2-4-8 ) over Seattle (1-28-16/6-1-14)
1997: Chicago (9-1-2/1-15-12) over Utah (1-15-18/15-5-5)
1998: Chicago (19-4-2/2-13-7) over Utah (1-18-9/16-22-5)
1999: San Antonio (10-14-16/1-21-19) over New York (16-27-24/4-12-5)
2000: LA Lakers (14-3-5/1-23-5) over Indiana (1-7-28/7-27-15)
2001: LA Lakers (7-6-3/13-28-10) over Philadelphia (13-21-2/3-8-9)
2002: LA Lakers (7-2-16/2-22-8 ) over New Jersey (17-18-13/4-3-13)
2003: San Antonio (3-24-13/2-12-17) over New Jersey (18-19-14/3-3-6)
2004: Detroit (19-20-9/2-7-13) over LA Lakers (8-5-16/11-14-6)
2005: San Antonio (6-1-15/1-6-3) over Detroit (23-14-4/2-18-5)

Analysis to follow...
Back to top
View user's profile Send private message Yahoo Messenger
Johnny Slick



Joined: 15 Oct 2005
Posts: 45
Location: Seattle

PostPosted: Sun Oct 16, 2005 3:59 am Post subject: Reply with quote
Here are the average ranks for winners and losers:

Winners:
O-Eff: 9.000
O-TO: 8.588
O-Reb: 9.353
D-Eff: 4.47
D-TO: 14.176
D-Reb: 9.353

Losers:
O-Eff: 9.117
O-TO: 14.647
O-Reb: 11.118
D-Eff: 6.824
D-TO: 12.236
D-Reb: 7.588

STATHEADS PLEASE LOOK AWAY. THE FOLLOWING METRIC WILL MAKE YOU DRY HEAVE. There were, on average, just over 28 teams in the league per year. That means that any stat where the average is around 14th or lower doesn't really matter, as teams, on average, won their conference despite being in the middle of the pack. Theoretically, if teams were *lower* than 14th on average, that'd be a sign that that stat somehow detracted from a team's winning. Fortunately, that didn't happen. Here is how championship basketball breaks down in that "baseball is 50% pitching" way that doesn't really make a lot of sense but nonetheless is fun to talk about:

Offensive efficiency is 19.1% of the game.
Avoiding turnovers makes up 9.4%.
Offensive rebounds? Try 14.7% on for size.
Defensive efficiency makes up an amazing 32.1%.
Creating turnovers, strangely enough, only contributes 3.3%.
And defensive rebounds add up to 21.4%.

The old school definitions of O and D would have it at about 43% offense and 57% defense. However, I prefer to look at it as 28.5% offense, 35.5% defense, and 35% rebounding.
Back to top
View user's profile Send private message Yahoo Messenger
jambalaya



Joined: 30 Jan 2005
Posts: 282


PostPosted: Sun Oct 16, 2005 10:31 am Post subject: Interesting but please explain conversion method Reply with quote
How did you go from the 6 stat ranks to the % of the game stats? That is key to undestanding your model, so please explain step by step how you made this conversion. Thanks.

We've been using ranks in this thread because it is easy to look at but if some version of this method were to be used in the future I think it would be preferable to go back to basing the comparison in the actual values because on some stats teams are bunched closely in general, or in certain parts of the distribution more so than others, so rank doesnt give a fully accurate display of the underlining comparison of raw data.

If you go to a % of the game method, using raw data is easy, though you would have to explain why their "wining value" might vary up and down from their weights in commonly used balanced statistical weight systems. And you are well aware, given your long participaion at the old board, there would likely be some discussion or resistance to any specific winning value weights depending how good the analysis and reasoning was behind their assignment. But lets see it and see where it leads and what is learned along the way.


Last edited by Crow on Thu May 12, 2011 3:16 am, edited 1 time in total.

Top
 Profile  
 
PostPosted: Fri Apr 15, 2011 10:48 am 
Offline

Joined: Thu Apr 14, 2011 11:10 pm
Posts: 2253
Author Message mateo82



Joined: 06 Aug 2005
Posts: 211
Posted: Fri Oct 14, 2005 3:32 pm Post subject: Opponents Points Per 100 Possessions = definitive team stat

I am convinced. Living in a defensive era, there is no stat more important. Since people don't talk a lot about team statistics here, I thought I'd bring up my favorite. Using basketball-reference.com, here are the rankings of every team that has made an NBA finals since 1989 in opponents points per 100 possessions: 2005 San Antonio Spurs (1st) over Detroit Pistons (3rd) 2004 Detroit Pistosn (2nd) over Los Angeles Lakers (8th) 2003 San Antonio Spurs (3rd) over New Jersey Nets (1st) 2002 Los Angeles Lakers (7th)* over New Jersey Nets (2nd) 2001 Los Angeles Lakers (19th)* over Philadelphia 76ers (5th) 2000 Los Angeles Lakers (1st) over Indiana Pacers (13th) 1999 San Antonio Spurs (1st) over New York Knicks (3rd) 1998 Chicago Bulls (3rd)* over Utah Jazz (15th) 1997 Chicago Bulls (4th)* over Utah Jazz (11th) 1996 Chicago Bulls (1st) over Seattle Sonics (2nd) 1995 Houston Rockets (12th)* over Orlando Magic (13th) 1994 Houston Rockets (4th) over New York Knicks (1st) 1993 Chicago Bulls (3rd)* over Phoenix Suns (9th) 1992 Chicago Bulls (4th)* over Portland Trailblazers (1st) 1991 Chicago Bulls (6th) over Los Angeles Lakers (4th) 1990 Detroit Pistons (1st)* over Portland Trailblazers (3rd) 1989 Detroit Pistons (3rd) over Los Angeles Lakers (7th) * an asterisks for repeat winners Note that repeat champions often are low. Why this is, is anyone's guess. Maybe the correlation means nothing, maybe it does. Maybe repeat winners, like Shaq used to say, take the regualr season off. I don't know. Just thought i'd throw out this wonderful stat for discussion.Last edited by mateo82 on Fri Oct 14, 2005 6:32 pm; edited 1 time in total
Back to top

jambalaya



Joined: 30 Jan 2005
Posts: 282
Posted: Fri Oct 14, 2005 3:54 pm Post subject: I agree

I agree it is a very important stat. 22 of the 34 finalists in 17 years had a top5 defense on this measure or 65% of all finalists, and 82% were top 10. The average rank was just about 5th.Last edited by jambalaya on Fri Oct 14, 2005 4:24 pm; edited 2 times in total
Back to top

mateo82



Joined: 06 Aug 2005
Posts: 211
Posted: Fri Oct 14, 2005 3:59 pm Post subject:

All winners since 1994 (except repeats) have been top 3. (1994 Houston is listed as #4 but are actually tied for #3 with Seattle).
Back to top

jambalaya



Joined: 30 Jan 2005
Posts: 282
Posted: Fri Oct 14, 2005 4:08 pm Post subject:

"Except repeats" is an important caveat. The only case with a defense below the median for all teams that year was the defending Lakers who picked it up as Shaq said and repeated in 2001. It would be interesting to see the data on playoff defensive performance on this measure and see how it matches up. That would seem a more valuable measure since the regular season is different and ultimately not as important. On the regular season data, the winners averaged 4.7th ranked and the losers 5.5th ranked. The better ranked team won 10 of 17 of the match ups, nice but not a large margin, so it may say more about getting there than winning the championship series. I suspect the team with the higher offense - defense point differential per 100 possession would be the winner more often and you can rank higher on that several ways.Last edited by jambalaya on Fri Oct 14, 2005 4:29 pm; edited 3 times in total
Back to top

mateo82



Joined: 06 Aug 2005
Posts: 211
Posted: Fri Oct 14, 2005 4:26 pm Post subject:

jambalaya wrote:
"Except repeats" is an important caveat.
It is, but because all repeat winners had a lower ranking than their first year (without exception), maybe this means that the correlation is not only coincidence. but maybe it is. Quote:
It would be interesting to see the data on playoff defensive performance on this measure and see how it matches up. That would seem a more valuable measure since the regular season is different and ultimately not as important.
I would love to see this as well. I also wonder if this stat is compiled at the end of seasons or as the season goes on. It seems like it would be meticulous to do it myself. Quote:
On the regular season data, the winners averaged 4.7th ranked and the losers 5.5th ranked. The better ranked team won 10 of 17 of the match ups, nice but not a large margin, so it may say more about getting there than winning the championship series. I suspect the team with offense - defense point differential per 100 possession would be the winner more often and you can rank high on that several ways.
Well, I wouldn't go that far. Just getting there isn't necessarily enough. If you get without a very elite defensive team then you stand no chance, good offense or not. I checked the median, thinking the very bad 2001 lakers might have dragged down the winners down in mean but the median is 3 for both.
Back to top

mateo82



Joined: 06 Aug 2005
Posts: 211
Posted: Fri Oct 14, 2005 4:30 pm Post subject:

As a side issue, when compiling this, I was shocked by how overwhelmingly dominant the 1994 Knicks were on defense. 97.6, which is ridiculously good. Almost 4 points better than the Rockets. And, if I recall correctly, no one else was even close in the years I covered. I think 2003 Nets were something like 98.2... but that's the closest it ever came (except 1999, when about 10 teams were as good or better.... so i'm rejecting that year altogether, that just doesn't make sense). It kind of reinforces the idea that NY choked hard that year.
Back to top

jambalaya



Joined: 30 Jan 2005
Posts: 282
Posted: Fri Oct 14, 2005 4:33 pm Post subject:

A look at basketball reference's playoffs index suggests you could manually compute team offensive performance from that but not defense. Perhaps it could be enhanced to allow easier calculations. You probably could assemble from espn.com or nba.com playoff data but maybe not all the way back. You could get three years from 82games.com and just one from dougsstats. I agree being top5 defense is very very important, top 10 close to an absolute requirement to win a championship (or have a Shaq or Oluwajon in take care of the paint in the clutch). But still you can be a lot of different good combinations of top 5-10 defense and top 5-10 offenses and get it done. And your offense might be somewhat better than your defense (you don't have to be defensive-biased, netting the two measures). But it is true that in 13 of 17 most recent cases the winner was top5. If I was planning on a championship, I'd plan on being top 5 on defense if at all possible, using every means available. Phoenix's off season moves and the players/coaches greater efforts could move them up from 16th on defensive efficency at http://www.knickerblogger.net/stats/d_de.htm to perhaps top 10 but can they move into top5? I'd be somewhat surprised. Both Brian Grant and Kurt thomas give up more than 50% FG to opposing centers. And no meaningful improvement on that over Amare and the departed Hunter at that spot. And Raja Bell was actually a little worse than the departed Q Richardson and Joe Johnson. So I guess it better get done on chemistry and greater effort because I am not sure they really upgraded defensive talent that much- at least by this one measure, which admittedly isnt sufficient by itself to answer the whole defensive question- even if they upgraded defensive "reputation".
Back to top

mateo82



Joined: 06 Aug 2005
Posts: 211
Posted: Fri Oct 14, 2005 6:35 pm Post subject:

Here are the stats of a team's own points per 100 possessions so we can compare to their opponents: 2005 San Antonio Spurs (8th) over Detroit Pistons (16th) 2004 Detroit Pistons (18th) over Los Angeles Lakers (6th) 2003 San Antonio Spurs (6th) over New Jersey Nets (19th) 2002 Los Angeles Lakers (3rd)* over New Jersey Nets (16th) 2001 Los Angeles Lakers (2nd)* over Philadelphia 76ers (13th) 2000 Los Angeles Lakers (6th) over Indiana Pacers (1st) 1999 San Antonio Spurs (10th) over New York Knicks (26th) 1998 Chicago Bulls (9th)* over Utah Jazz (1st) 1997 Chicago Bulls (1st)* over Utah Jazz (2nd) 1996 Chicago Bulls (1st) over Seattle Supersonics (6th) 1995 Houston Rockets (8th)* over Orlando Magic (2nd) 1994 Houston Rockets (16th) over New York Knicks (14th) 1993 Chicago Bulls (3rd)* over Phoenix Suns (1st) 1992 Chicago Bulls (1st)* over Portland Trailblazers (7th) 1991 Chicago Bulls (1st) over Los Angeles Lakers (5th) 1990 Detroit Pistons (11th)* over Portland Trailblazers (8th) 1989 Detroit Pistons (8th) over Los Angeles Lakers (1st) * an asterisks for repeat winners Winners are average of 6.5 and losers are 8.5 Obviously defense is much more important. Aside from the averages: just look at a few examples. The 2004 Lakers, for example, who were quite balanced. 8th in defense, 6th in offense. So if being a top10 in both categories is the strategy... it doesn't always work. The Pistons were a miserable 18th in offense but 2nd in defense. I think focusing on that top 3 or 4 defense is the way to go. Even among teams that were top 5 in defense and top 10 in offense, there are quite a few examples of losers that fit that description: Sonics, 92 Blazers, 91 Lakers, 90 Blazers all were top 5 defensive and top 10 offense and lost. That's how unimportant "balance" seems to be. Dominant defense is what works.
Back to top

jambalaya



Joined: 30 Jan 2005
Posts: 282
Posted: Fri Oct 14, 2005 7:00 pm Post subject: Reply

Top 5 offense, top 10 defense isnt always enough, I agree. But balance does still matter. In 12 of the 17 cases the most balanced team, computed as absolute value of rank on defense - rank on offense, won the championship match up (with one tie): differential = defense rank - offensive rank winner loser 2005 7 13 2004 16 -2 2003 3 18 2002 -4 14 2001 -17 8 2000 5 -12 1999 9 23 1998 6 -14 1997 -3 -9 1996 0 4 1995 -4 -11 1994 12 13 1993 0 0 1992 -3 6 1991 -5 1 1990 10 5 1989 5 -6 avg 2.176470588 3 It is noteworthy that both are indeed defensively biased, in line with your argument that defense is more important (and I agree with that to a point, but add the data about balance instead of rejecting it), but the winner, on average was less defensively biased than the loser, slightly [/u] and "balance" is a slightly better predictor of the winner year to year (12 of 17) compared to defensive ranking alone (10 of 17 called correctly). In 2002 the lakers were #3 O / #7 D and won the championship, so a top5 offense, top ten but not top 5 defense can win. And in another case it also happened 1991 Bulls #1 O / #6 D. In two other cases an offensively biased team also won 1997 bulls #1 0 / #4 D 1992 Bulls #1 / #4 D and came just short a little short of meeting this description. It is fairly rare, 2 cases and 2 close calls in 17 years, but these four examples help counterbalance the ones you selected were they failed. Still it appears true that a top5 defense / top 10 offense is a better strategy with 8 winners (if you count one case where the offense was even worse the 94 Rockets). And I agree would seem the better strategy than the reverse. Nonetheless, overall 7 of the 17 winners were "offensively biased" by this simple net rank measure. So good offense rank is usually also important. The amount of importance varies. It can be outweighed sometimes by great D as with the Pistons. In three of the 4 cases where a team was 10 or higher on offense yet won the championship they had a #1 or #2 defense with the fourth case being #4. But there are two cases of a team 10 or higher on defense still winning, in one case they were #2 on offense, in the other #8. As for the top4 in defense model, such teams appeared 22 times, impressive. They won 13 times, also impressive. Certainly top5 defense is a great strategy, maybe the best. But 8 times top4 defenses met each other, in 5 cases the lesser defense won (even in two cases where their offense was actually worse than the opponent), in 3 cases the better defense won. A small margin, but another piece of the record. By comparison top 4 offenses appeared half as often, 11 times but won 7 times. Less often, but a slightly better winning percentage 64% to 59%, still on balance I'll stay with the top 4 defense over top 4 offense as a better strategy. A top 4 "balance" team also appeared 11 times and won 7 of them. If I could acheive any of the three top 4 ranks I'd run with it. In fact the three different top 4 strategies account for 16 of the 17 examples. There were five cases where defense and offense top4 status overlapped, and four of them were the strongest dynasty of the era Chicago Bulls and the other was the breakthru, start of the modern defensive era, 89 Pistons. There are no cases of a team with a top4 on offense and defense not winning the championship in this period though there are 4 cases of failure that are close, with a top5 -7 rank in one field while top4 in the other. There are 5 cases where teams where top4 on offense and balance on won a championship and no examples of losses; but in four of the cases they were also top4 in defense to complete the trifecta (the only exception being the 2002 Lakers with a good but still #7 defense in a repeat situation so it was probably still better for the playoffs). There are 6 cases on top4 in defense and balance that won the championship, but 2 cases where they failed to win a championship- the 96 Sonics, 91 Lakers, and two more other cases where they failed to win a championship with a top 4 in defense but with a top5-6 finish in balance. I think it would be fairly safe to say your strategy ought to be on having at least one of the three top4 ranks. I dont think any lesser rank strategy on offense, defense or balance is likely to cut it very often. And it wouldnt hurt to try to have two or at least to improve your relatively weak area as much as you can so that it has less chance of being what keeps you from winning with your strength. In one case a top 4 defense meet a top 4 balance team who was not also a top4 defense, the top 4 defense won - 2004 Pistons. In another case though, a top 3 balance team beat a top4 defense team - 1992 Bulls (But they did have a just outside the top4 standard, top 5 defense to add to their #1 offense and effective balance; while the Blazers had a top7offense to go with thier slightly superior #4 D . Both team good cases for balance working well.) In three cases a top 4 defense team that was not also top4 on offense meet a top4 offense that was not also top4 on defense and the better defense won 2-1. My overall point, besides trying to give a fair look to as many angles as I could think of, is that the record has enough diversity that it should be noted / considered along with "the defensive wins a lot" evidence. There is some flexibility in how defense, offense and balance play out and who wins. Defense seems to have the edge but by itself it can not claim all the credit or all the championships.Last edited by jambalaya on Fri Oct 14, 2005 9:40 pm; edited 7 times in total
Back to top

mateo82



Joined: 06 Aug 2005
Posts: 211
Posted: Fri Oct 14, 2005 8:18 pm Post subject:

good post, i'll respond to it later. just to let you know, in my original post I had mistakenly listed the 1992 Suns as 1st in defense when it fact they were 9th. I editted in the correct value, but if that messed up any of your data you might consider that.
Back to top

jambalaya



Joined: 30 Jan 2005
Posts: 282
Posted: Fri Oct 14, 2005 8:29 pm Post subject: Reply

I think it was a good topic, good exchange, pretty full consideration of the data and alternate models. Maybe others will add views later. The error might affect some of my statements by one case, but not others, and I probably won't go back thru and revise everything because most of the points will remain close to the same in meaning.
Back to top

jambalaya



Joined: 30 Jan 2005
Posts: 282
Posted: Fri Oct 14, 2005 11:12 pm Post subject: Leading contenders for this coming year

Based on this concept that you need to be a top 4 offense, defense or balance team, I turned to last year's playoff data as a relevant (play at end of year / when it is really on the line, stats not influenced by performance against nonplayoff teams), not yet examined database to start a consideration of top contenders for this year (despite uneven playoff competition which diostorts). Changes to this year's roster to be briefly considered next. 2005 Playoff data Team Scoring Team Defense Team Pts/Game Team Pts/Game 1 PhoenixSuns 112.0 1 DetroitPistons 85.6 2 DallasMavericks 105.2 2 IndianaPacers 86.9 3 MemphisGrizzlies 102.8 3 BostonCeltics 88.3 4 ChicagoBulls 102.5 4 SanAntonioSpurs 92.6 Team Scoring Differential Team Pts/Game 1 DetroitPistons 4.7 2 MiamiHeat 4.3 3 SanAntonioSpurs 4.3 4 PhoenixSuns 4.2 San Antonio, Detroit and Phoenix appear on two lists, a good starting point for being elite. San Antonio got better. Detroit probably hasnt aged or changed enough to slip, if Flip handles the challenge. Phoenix changed, maybe not better on talent, but more experienced with each other and being a contender so I think it safe to include them too. The number 2's in each of the categories Dallas on offense, Indiana on defense and Miami on balance seem like good starters for at least being second tier elite if not equal to the first tier. Miami changed the most but probably a net improvement. Indiana got Artest and another guard. Dallas may have slipped a bit. A third tier seems to include Chicago, Boston, Memphis who made appearances in the top4s. You could argue each slipped down, but with Chicago it is debatable and with more experience for the young guys offsetting they are most likely to still be competitive for the elite 8. Seattle and Washington were elite 8 last year but not on these specific playoff measures. I'd put them third tier to start. Both had some changes, probably mild net losses. Any teams likely to move up close enough to consider them as well? Denver, Houston, Sacramento, Cleveland, New Jersey. Thats 16 teams mentioned. Using top 4 in something as the standard for weeding, I'd say it is San Antonio, Detroit and Phoenix for sure, probably Dallas, Indiana and Miami. I'd give Denver and Chicago as more likely final wildcards for top4 on defense or balance. Most or all of the top 4 next playoffs are likely to come from these elite 8. Washington and Houston seems like the next most likely bubble teams to me. I won't try to pick the top4 or top 2 yet but by all-star game the odds of guessing right then will improve quite a bit.
Back to top

Johnny Slick



Joined: 15 Oct 2005
Posts: 45
Location: Seattle
Posted: Sun Oct 16, 2005 3:39 am Post subject:

First off, I kept meaning to get around to coming here, and I'm glad to finally get over. FWIW I posted on the APBR listserv under the pseudonym John Craven (okay, maybe that was more of a "nym"). Anyway... The problem I have with a stat like this is that really to me basketball is composed of six basic components, not two: - Scoring points efficiently (measured in terms of points divided by shot attempts, counting a free throw attempt as .4 of one per what everyone else is doing WRT possessions) - Avoiding turnovers - Grabbing offensive rebounds - Grabbing defensive rebounds - Forcing turnovers - Forcing your opponent into inefficient shooting. We're lumping the first 3 into offense and the last 3 into defense. I don't think that's right. It's kind of like baseball sabermetricians grouping pitching together with defense. Do rebounds feed off of good D? Sure, but defensive turnovers should also lead to a better scoring-efficiency rating and I don't see anybody trying to look at *that*. Anyway, just for fun I'm going to list teams' ratings by each of these metrics, year by year, then come up with a horribly flawed statistic that attempts to quantify what "percentage" of the game is offense, defense, and so on. Anyway, here goes. Stats go team name (scoring-TOs-ORs / defensive scoring-TOs-DRs). 1989: Detroit (1-17-7/3-19-4) over LA Lakers (8-10-12/6-21-6) 1990: Detroit (13-8-5/1-15-4) over Portland (15-10-2/11-6-3) 1991: Chicago (4-3-4/14-4-9) over LA Lakers (8-6-15/1-23-13) 1992: Chicago (6-2-5/8-7-4) over Portland (8-12-4/10-5-2) 1993: Chicago (16-1-1/18-3-12) over Phoenix (1-18-10/10-14-6) 1994: Houston (5-20-27/1-22-5) over New York (13-22-8/2-2-1) 1995: Houston (4-19-26/4-20-19) over Orlando (3-7-3/5-24-15) 1996: Chicago (10-1-1/2-4-8 ) over Seattle (1-28-16/6-1-14) 1997: Chicago (9-1-2/1-15-12) over Utah (1-15-18/15-5-5) 1998: Chicago (19-4-2/2-13-7) over Utah (1-18-9/16-22-5) 1999: San Antonio (10-14-16/1-21-19) over New York (16-27-24/4-12-5) 2000: LA Lakers (14-3-5/1-23-5) over Indiana (1-7-28/7-27-15) 2001: LA Lakers (7-6-3/13-28-10) over Philadelphia (13-21-2/3-8-9) 2002: LA Lakers (7-2-16/2-22-8 ) over New Jersey (17-18-13/4-3-13) 2003: San Antonio (3-24-13/2-12-17) over New Jersey (18-19-14/3-3-6) 2004: Detroit (19-20-9/2-7-13) over LA Lakers (8-5-16/11-14-6) 2005: San Antonio (6-1-15/1-6-3) over Detroit (23-14-4/2-18-5) Analysis to follow...
Back to top

Johnny Slick



Joined: 15 Oct 2005
Posts: 45
Location: Seattle
Posted: Sun Oct 16, 2005 3:59 am Post subject:

Here are the average ranks for winners and losers: Winners: O-Eff: 9.000 O-TO: 8.588 O-Reb: 9.353 D-Eff: 4.47 D-TO: 14.176 D-Reb: 9.353 Losers: O-Eff: 9.117 O-TO: 14.647 O-Reb: 11.118 D-Eff: 6.824 D-TO: 12.236 D-Reb: 7.588 STATHEADS PLEASE LOOK AWAY. THE FOLLOWING METRIC WILL MAKE YOU DRY HEAVE. There were, on average, just over 28 teams in the league per year. That means that any stat where the average is around 14th or lower doesn't really matter, as teams, on average, won their conference despite being in the middle of the pack. Theoretically, if teams were *lower* than 14th on average, that'd be a sign that that stat somehow detracted from a team's winning. Fortunately, that didn't happen. Here is how championship basketball breaks down in that "baseball is 50% pitching" way that doesn't really make a lot of sense but nonetheless is fun to talk about: Offensive efficiency is 19.1% of the game. Avoiding turnovers makes up 9.4%. Offensive rebounds? Try 14.7% on for size. Defensive efficiency makes up an amazing 32.1%. Creating turnovers, strangely enough, only contributes 3.3%. And defensive rebounds add up to 21.4%. The old school definitions of O and D would have it at about 43% offense and 57% defense. However, I prefer to look at it as 28.5% offense, 35.5% defense, and 35% rebounding.
Back to top

jambalaya



Joined: 30 Jan 2005
Posts: 282
Posted: Sun Oct 16, 2005 10:31 am Post subject: Interesting but please explain conversion method

How did you go from the 6 stat ranks to the % of the game stats? That is key to undestanding your model, so please explain step by step how you made this conversion. Thanks. We've been using ranks in this thread because it is easy to look at but if some version of this method were to be used in the future I think it would be preferable to go back to basing the comparison in the actual values because on some stats teams are bunched closely in general, or in certain parts of the distribution more so than others, so rank doesnt give a fully accurate display of the underlining comparison of raw data. If you go to a % of the game method, using raw data is easy, though you would have to explain why their "wining value" might vary up and down from their weights in commonly used balanced statistical weight systems. And you are well aware, given your long participaion at the old board, there would likely be some discussion or resistance to any specific winning value weights depending how good the analysis and reasoning was behind their assignment. But lets see it and see where it leads and what is learned along the way.


Author Message HoopStudies



Joined: 30 Dec 2004
Posts: 705
Location: Near Philadelphia, PA
Posted: Sun Oct 16, 2005 11:56 am Post subject:

Johnny Slick wrote:
First off, I kept meaning to get around to coming here, and I'm glad to finally get over. FWIW I posted on the APBR listserv under the pseudonym John Craven (okay, maybe that was more of a "nym"). Anyway... The problem I have with a stat like this is that really to me basketball is composed of six basic components, not two: - Scoring points efficiently (measured in terms of points divided by shot attempts, counting a free throw attempt as .4 of one per what everyone else is doing WRT possessions) - Avoiding turnovers - Grabbing offensive rebounds - Grabbing defensive rebounds - Forcing turnovers - Forcing your opponent into inefficient shooting. We're lumping the first 3 into offense and the last 3 into defense. I don't think that's right. It's kind of like baseball sabermetricians grouping pitching together with defense.
Whoa! Offensive and defensive points per 100 possessions are great summaries. It's not as though the other stuff is dismissed at all. In fact, that's what the 4 Factors are about. Pts scored (or allowed) per 100 possessions are composed of - shooting percentage - turnover rate - offensive boards - getting to the line Combining shooting percentage with getting to the line is what may give you less, but I choose to keep them separate. At times I combine them. At times, I don't. Depends on what I'm looking at and how big a scale you're looking at it. And that is the point. Points per possession is a structure that allows analysis at an infinite number of levels. It is the biggest of pictures, details a little murky. But you can break it down however you want. Not sure whether your numbers for responsibility of defense, efficiency, etc. are right. But it is fair to look at things that way._________________Dean Oliver Author, Basketball on Paper The postings are my own & don't necess represent positions, strategies or opinions of employers.
Back to top

Johnny Slick



Joined: 15 Oct 2005
Posts: 45
Location: Seattle
Posted: Sun Oct 16, 2005 12:12 pm Post subject: Re: Interesting but please explain conversion method

jambalaya wrote:
How did you go from the 6 stat ranks to the % of the game stats? That is key to undestanding your model, so please explain step by step how you made this conversion. Thanks. We've been using ranks in this thread because it is easy to look at but if some version of this method were to be used in the future I think it would be preferable to go back to basing the comparison in the actual values because on some stats teams are bunched closely in general, or in certain parts of the distribution more so than others, so rank doesnt give a fully accurate display of the underlining comparison of raw data.
I just subtracted the average rank from 14 and went from there. Absolutely, that's not a really great way of doing things; when I re-do this, I'll probably want to do it by standard deviations from the mean. I just used rank because a. it's a quick and dirty way of seeing what mattered and what did not, and b. we'd been using ranks in the thread. No, I wouldn't take too much from this, although it does seem to show that, come playoff time, adjusting your opponents' shot selection is really, really important, forcing turnovers isn't important at all, and rebounding in general is pretty darn important. That jibes pretty well with CW, I think: the George Karl Sonics, for example, were consistently derided as a team built for the regular season in part because one of their most consistent attributes was the ability to force turnovers out of their opponents. To Dean: I agree with what you're saying, but here's my general point: to me, rebounding isn't really defense or offense, it's its own thing. The "four factors" is nice and all, don't get me wrong. I think it's more accurate to look at things in terms of the three spheres, to actually separate offense from offensive rebounding and so on. You "can" separate all that any time you want to, sure, but I'm not talking in terms of "cans". I'm talking in terms of "shoulds."
Back to top

Kevin Pelton
Site Admin


Joined: 30 Dec 2004
Posts: 976
Location: Seattle
Posted: Sun Oct 16, 2005 12:31 pm Post subject:

Johnny Slick wrote:
First off, I kept meaning to get around to coming here, and I'm glad to finally get over. FWIW I posted on the APBR listserv under the pseudonym John Craven (okay, maybe that was more of a "nym"). Anyway...
And we add again to our percentage of Sonics fans. Good to see you, and I hope the difference between here and the P-I forum isn't too disconcerting. I tend to agree with you about rebounding, though I'd argue that - at the team level - even offensive and defensive rebounding are separate skills, giving you a base of four. I don't know that it's that surprising that CW would be reflected by the performance of championship teams; not only are they creating CW with their performance, they're responding to it. Teams trying to win the championship are going to build like they think is necessary to win a championship. Because of that, what I think is more interesting is grouping teams and then looking at how they've performed in the playoffs. When Detroit plays San Antonio, for example, it doesn't really tell us much, because their philosophies are basically the same. When Detroit plays Phoenix, however, that's much more revealing.
Back to top

jambalaya



Joined: 30 Jan 2005
Posts: 282
Posted: Sun Oct 16, 2005 4:04 pm Post subject: Reply of translation method

Ok, I took the stat part ranks and subtracted them from 14 and added the numbers for the winning and losing teams together and then took each stats rank difference as a pct. of the total rank differences and got basically what you've shown and see how the conversion is done. Using rank is an imperfect simplification but useful. Using your method and numbers I see how you derive and say "The old school definitions of O and D would have it at about 43% offense and 57% defense. However, I prefer to look at it as 28.5% offense, 35.5% defense, and 35% rebounding." I think given the earlier discussion of offensive and defensive shares of importance it is also worth saying that for the winners and the losers the sum of all three parts of your data (instead of just two left in "offensive effeciency as rebounds is removed into a separate category) would show that full "offense" and "defense" scores this way: Winners total offense 51.8% total defense 48.2% Losers total off 31.7% total def 68.3% (my numbers are slightly differnt from yours but seem close enough) That would seem to suggest that total offense is more important than total defense to the championship winner? And Total rebounding Winners32.0% Losers 41.4% Rebounding is less important to the winners? Because it is less important, or because they scored more and that scoring was also important and relatively more so as a result? There is useful information here but reason for caution and not over interpreting from the data too. You use the argument that stat rank difference from the mean (28 ) as a justification for weighting some stat categories more than others in terms of winning value. But if you accept that a rebound is a possession and worth about a point and so is a turnover, the box score and traditional weight sets would also say that offensive and deffensive efficiency are about 2.3 times bigger in impact than rebounding (say an average of 92 points in finals vs 40 rebounds) and turnovers about one-third of what rebounding was (about 14 per team). So you could possibly adjust your weights based on that too? Or, using these boxscore-based weights by themselves, How many % importance off eff 92 36.5% offturnovr 14 5.6% 42.1% offensive efficency off reb 14 5.6% def eff 92 36.5% defturnovr 14 5.6% 42.1% defensive efficency deff reb 26 10.3% 15.9% total rebounding total 252 Rebounding would be less important with these weights than your method and on average offense and defense would be equally important. Looking at both importance estimates and others, the discussion will go on. But good stuff.
Back to top

Johnny Slick



Joined: 15 Oct 2005
Posts: 45
Location: Seattle
Posted: Mon Oct 17, 2005 4:14 am Post subject:

I think that part of the point of why defense in particular seems to be rated more highly in terms of Finals teams is that idea that the playoffs are a different game than the regular season. Good defense gets better, everybody stops letting you score on back door cuts and so on, there's very little careless, screw-it-we're 20-62 point guarding going on and therefore fewer steal opportunities, and so on and so forth. In that respect, possessions are specifically not meaningful except to compare those stats with what "really are" important when it comes to actually getting to the Finals. In advance of the Finals deal, here are the average standard deviations +1 for 55+ win teams (because we're seeing, in this case, how good teams tend to define themselves above mediocrity in the regular season): Shooting: 1.17 TOs: 1.06 ORs: 0.97 D-Shooting: 1.64 D-TOs: 0.93 D-Rebs: 2.43 (that's no typo; of the 85 teams who won 55 or more games since 1988, only 23 were below average, and just 3 1 standard deviation below. Conversely, 27 teams were at least 1 SD above average. 5 scored at 1.9 SDs or higher) So, the "percentages" (do with these what you will, and it's probably not accurate to compare them with the earlier percentages because those were calculated in a completely different fashion): Shooting: 14.3% TOs: 13.0% ORs: 11.9% Defense: 19.9% D-TOs: 11.3% DRs: 29.7% Offense: 39.1% Defense: 60.9% or O: 27.2% D: 31.2% R: 41.5%
Back to top

jambalaya



Joined: 30 Jan 2005
Posts: 282
Posted: Mon Oct 17, 2005 12:06 pm Post subject:

The standard deviation data is more accurate and therefore powerful than rank alone. Thanks for the work that went into it. I am guessing that you are not moved to weight the stats by relative boxscore importance. I remain an advocate for that in some fashion. I also would prefer to see the discussion based in playoff data but believe it is not as readily available. Does anyone have a modern history, playoff, team statistics with opponents stats already compiled database? It appears only 5 years of playoff data is available at nba.com. Good but is there more elsewhere? I might compile it, but if somebody already has that would save time. It is broken down by playoff round, so if you argued as some might, that the really important basketball is the top 4 teams against each other, then you could restrict the study to just the conference finals and championship.Last edited by jambalaya on Mon Oct 17, 2005 12:44 pm; edited 1 time in total
Back to top

kjb



Joined: 03 Jan 2005
Posts: 861
Location: Washington, DC
Posted: Mon Oct 17, 2005 12:36 pm Post subject:

I appreciate the work that's going into this, but I find it very difficult to believe that rebounding the basketball is more important than offense or defense. It just doesn't make sense, especially in light of plenty of other research showing that shooting efficiency is of first importance, followed by turnovers (at least in the NBA). I fundamentally disagree that rebounding is its own thing. Dean's point in BoP (which I agree with) is that there are offensive rebounds, which help a team's offense by getting it more shots at the basket; and there are defensive rebounds, which help a team's defense by preventing the opponent from getting additional shots at the basket. How much work went into making sure these stat categories are independent? In other words, could the rebounding numbers be "picking up" stuff related to shooting efficiency or turnovers? Along the same lines, what numbers are you using for rebounds? (Season totals or per game averages, or rebounding percentages?)
Back to top

Johnny Slick



Joined: 15 Oct 2005
Posts: 45
Location: Seattle
Posted: Tue Oct 18, 2005 1:09 am Post subject:

For rebounding, I'm using defensive boards divided by opponents' offensive boards and vice versa. That eliminates the direct effect of solid defense creating more defensive rebounding opportunities in the first place. There is, I am sure, a secondary or tertiary effect on those two data points on each other (for example, cleaning the glass means fewer putbacks which means fewer easy shots in the paint), but I'm not exactly sure how to get rid of that. As for Dean's point, I do understand it but I fundamentally disagree. On an individual level, players can do all kinds of things that keep defensive scoring efficiency low that either don't help or have a deleterious effect on defensive rebounding. For example, shotblocking specialists often put themselves out of position to pull down rebounds when they do their shot-altering. On the other hand, giving the opposing 4 the area 16 feet away from the basket and out may put your man in a better position to corral boards, but probably will hurt you when you play against teams with good shooting forwards. John Hollinger pointed out in his book that one reason Tim Duncan is so amazing is that, unlike so many bigs, he can both alter shots and get rebounds. The way he talks about how Duncan does it demonstrates that the Spurs power forward is a master of expressing one skill set without hampering another one. It seems like John H. understands that these are two different skills, even though he may not have explicitly made the point. So defensive rebounding and defense are obviously their own skills on an individual level... why not assume such on a team level? I think that a large part of this has little to do with the way the game is actually played and much to do with peoples' perception of the way all games are played.* All the team-based sports I can think of are divided into the two poles of offense and defense. Baseball, football... even hockey has its defensemen and wings. It doesn't make sense in the conventional way to divide basketball three ways. But I think it makes good statistical sense. As for what's more important, I'm not really asking anyone to believe anything. I'm just pointing out here that, when it comes to 55+ win teams (fwiw, the standard deviation on wins since 88-89 is just over 14, so a 55-win team is about 1 SD above the mean), some shoot well, lots play good defense... but nearly all of them rebound the basketball. Take from that what you will. *I want to point out that this is behavior endemic to all sports fans, and am not singling Dean out or anyone else who chooses to define basketball as O v D as stupid or anything. Conventional wisdom is very, very hard to overthrow, and Dean's done incredible work toppling a lot of other CW data.
Back to top

kjb



Joined: 03 Jan 2005
Posts: 861
Location: Washington, DC
Posted: Tue Oct 18, 2005 9:06 am Post subject:

Difference in perspective and views, I guess. I think it's more useful analytically to look at rebounding as parts of offense and defense -- you haven't stopped anyone until you get that defensive board, nor have you been stopped as long as you can get that offensive rebound. I understand your point -- I just don't agree.
Back to top

HoopStudies



Joined: 30 Dec 2004
Posts: 705
Location: Near Philadelphia, PA
Posted: Tue Oct 18, 2005 11:19 am Post subject:

Johnny Slick wrote:
As for Dean's point, I do understand it but I fundamentally disagree. On an individual level, players can do all kinds of things that keep defensive scoring efficiency low that either don't help or have a deleterious effect on defensive rebounding. For example, shotblocking specialists often put themselves out of position to pull down rebounds when they do their shot-altering. On the other hand, giving the opposing 4 the area 16 feet away from the basket and out may put your man in a better position to corral boards, but probably will hurt you when you play against teams with good shooting forwards. John Hollinger pointed out in his book that one reason Tim Duncan is so amazing is that, unlike so many bigs, he can both alter shots and get rebounds. The way he talks about how Duncan does it demonstrates that the Spurs power forward is a master of expressing one skill set without hampering another one. It seems like John H. understands that these are two different skills, even though he may not have explicitly made the point. So defensive rebounding and defense are obviously their own skills on an individual level... why not assume such on a team level? I think that a large part of this has little to do with the way the game is actually played and much to do with peoples' perception of the way all games are played.* All the team-based sports I can think of are divided into the two poles of offense and defense. Baseball, football... even hockey has its defensemen and wings. It doesn't make sense in the conventional way to divide basketball three ways. But I think it makes good statistical sense. As for what's more important, I'm not really asking anyone to believe anything. I'm just pointing out here that, when it comes to 55+ win teams (fwiw, the standard deviation on wins since 88-89 is just over 14, so a 55-win team is about 1 SD above the mean), some shoot well, lots play good defense... but nearly all of them rebound the basketball. Take from that what you will.
Let me re-emphasize, it's not a matter of right or wrong here. I don't know how you can "strongly disagree" since I can generate importance of rebounding, offensive shooting, defensive shooting, etc. using the 4 Factor framework. I don't know how anything you're proposing can be, as you say, "more accurate." What is more accurate about what you do? Further, regarding "skills" or roles, there are lots of them and you can classify them into many different categories (KevinP published a list of mine last year). I almost always end up with at least 6, among which rebounding is always one (not split up between O and D because players who do one tend do the other, though I personally think the techniques are different). EdK pointed out a paper years ago that spurred my thought on this, though it was about identifying how many positions there are -- not just PG, SG, etc. Technical Qs: - Using rebounding percentages rather than ratios may make a little bit of a difference. Not sure since I never do it your way. Kevin's point about independence hasn't been fully addressed. There is correlation among the different factors. - Is it turnovers per possession? Turnovers per something else? (I edited out the more broad questions I originally posted since one was answered and the other was too general.) - "Rebounding is its own thing" is the phrase used by those who analyze the game not using possessions. If you have offense, defense, and rebounding, you have an entirely different list of what teams are good and bad offensively and defensively than the rest of us. And you then need some formula that relates offense, defense, AND rebounding to winning. I'd love to see such a formula. It's definitely possible, but how accurate is it?_________________Dean Oliver Author, Basketball on Paper The postings are my own & don't necess represent positions, strategies or opinions of employers.
Back to top

Kevin Pelton
Site Admin


Joined: 30 Dec 2004
Posts: 976
Location: Seattle
Posted: Tue Oct 18, 2005 12:05 pm Post subject:

HoopStudies wrote:
And you then need some formula that relates offense, defense, AND rebounding to winning. I'd love to see such a formula. It's definitely possible, but how accurate is it?
I did that in this column. That regression is the basis of my individual player ratings tying together offense, defense, offensive rebounding and defensive rebounding. http://www.hoopsworld.com/cgi-bin/news/ ... 9&num=6237 But I was specifically trying not to look at what makes a championship club, but instead a winning regular-season team.
Back to top

Johnny Slick



Joined: 15 Oct 2005
Posts: 45
Location: Seattle
Posted: Tue Oct 18, 2005 12:30 pm Post subject:

HoopStudies wrote:
Let me re-emphasize, it's not a matter of right or wrong here. I don't know how you can "strongly disagree" since I can generate importance of rebounding, offensive shooting, defensive shooting, etc. using the 4 Factor framework. I don't know how anything you're proposing can be, as you say, "more accurate." What is more accurate about what you do?
I think it's more a philosophical difference than anything else. Personally, I'd no sooner lump rebounding in with defense or offense than I'd combine prevention of offensive turnovers with defense. Your mileage, as the old Internet saying goes, may vary. Quote:
Further, regarding "skills" or roles, there are lots of them and you can classify them into many different categories (KevinP published a list of mine last year). I almost always end up with at least 6, among which rebounding is always one (not split up between O and D because players who do one tend do the other, though I personally think the techniques are different).
Strangely enough, *teams* that do one have no tendency whatsoever to do the other. Yeah, didn't make any sense to me either, but that's what the pro data from 1988 to now says. Quote:
EdK pointed out a paper years ago that spurred my thought on this, though it was about identifying how many positions there are -- not just PG, SG, etc.
That's an interesting point in and of itself. Man, I miss the old apbr forum! Quote:
Technical Qs: - Using rebounding percentages rather than ratios may make a little bit of a difference. Not sure since I never do it your way. Kevin's point about independence hasn't been fully addressed. There is correlation among the different factors.
Oh yeah, of course there's correlation. Everything in basketball feeds off of everything else to some extent; it's one of the reasons why the game is so intriguing on a statistical level. When I figure out how to do regression analysis between defensive rebounds and defensive scoring efficiency, I'll do it. As for doing the rebounding my way... the whole point of rebounding in and of itself is to get them at the expense of your opponents, right? Not at the expense of turnovers, shots, etc. In fact, I think that if I *did* do rebounds per possession, it'd be less accurate because of all that noise. Bad shooting, low turnover teams would show up as better offensive rebounders than good shooting, high turnover squads and so on. Quote:
- Is it turnovers per possession? Turnovers per something else?
I did turnovers by shots attempted in order to keep them and shooting on the same, erm, axis. Since I the first quick-and-dirty look used rank and the second will use standard deviations, making it per shot and per a re-defined possession that doesn't count offensive rebounds would yield the same results. Quote:
(I edited out the more broad questions I originally posted since one was answered and the other was too general.) - "Rebounding is its own thing" is the phrase used by those who analyze the game not using possessions.
Hmm. It's not that I'm removing possessions; it's more that I consider a possession to be the space between one turnover, made shot, or rebound of either kind, whereas you count them as each time a team goes to the other end of the court. I appreciate that side of things, and ideally I'd like to find a way to separate putbacks, but really... is there that much difference between a situation where a point guard walks the ball upcourt and when a rebounder passes the ball out to the top of the key and resets the offense? Quote:
If you have offense, defense, and rebounding, you have an entirely different list of what teams are good and bad offensively and defensively than the rest of us. And you then need some formula that relates offense, defense, AND rebounding to winning. I'd love to see such a formula. It's definitely possible, but how accurate is it?
Well, that's what I'm working on...
Back to top

jambalaya



Joined: 30 Jan 2005
Posts: 282
Posted: Tue Oct 18, 2005 1:30 pm Post subject: Team definitions

Reading Kevin's article leads me to think it may be wise (especially in the current discussion about high quality teams, different from his more general survey) to split teams into 3 categories instead of two (i.e. offensively and defensively biased), reserving offensive bias for teams above the league average in offense with an offense better compared to the league than their defense compared to the league, the same for the term defensive biased, and all teams below the league average in both to simply be called below average teams. I suggest this because under the old, two term system, a worse offense than defense gets you called defensive biased which casually can suggest some measure of defensive quality even though they may be below average on defense compared to the league. (Using ratings per 100 possessions for applying the terms removes the pace coloring effect. ) I would think the winning percentage of "defensively biased" teams defined as just those who are above the league in defense and with a league comparison stronger on defense than offense would be far stronger than the results shown when the term "defensively biased" included teams that are bad because they are below average on both but slightly less bad on defense. I guess the offensive biased numbers are affected by this too and wonder how the results would look for offense and defense under this revised definition. Better for both certainly by removing the bad both ways teams, but what about the comparison between them, closer or farther apart? It would seem to be worthwhile if the debate is really about winning expectations (making the playoffs or championship) of good offenses vs. good defenses. Or if you use offense, rebounding, and defense, you could be biased in 3 directions (if you were above average in any one) or below average in all. Separating into 4 groups might be helpful for explaining why some teams that dont look playoff material on offense and defense ratings, still pull it off (if rebounding is relatively higher in importance). The membership in the now more restricted definition offense and defense biased circles would be more selective and affect any statements made about their importance compared to each other and rebounding arguments could be made and compared against its rivals.
Back to top

Johnny Slick



Joined: 15 Oct 2005
Posts: 45
Location: Seattle
Posted: Tue Oct 18, 2005 1:58 pm Post subject:

As a follow-up to what jambalaya's talking about... since 1988 and not counting the strike-shortened season of 1999, there are 114 teams that had a defensive efficiency of 0.5 SDs above the mean and were better on defense than on offense (again measured in terms of SDs). The average regular-season record of those teams was 50.4-31.6. Conversely, offensive-minded teams who were decent at scoring were 50.5-31.5. Teams 0.5 SDs or more above the mean at defensive rebounding were 47.4-34.6, regardless of whether they were offense or defensive minded.
Back to top

jambalaya



Joined: 30 Jan 2005
Posts: 282
Posted: Tue Oct 18, 2005 2:25 pm Post subject: Reply

That data goes a way toward making an apparent case that the three categories are of nearly equal impact on winning. Or at least that winners have the specific category strengths about as often. Rebounding is slightly lower but still shows good strength. It would be interesting to see within the main categories how the winning percentage of a team meeting the .5 D definition of strong varies based on their score on the other two variables. I assume that teams that are pretty good on those other two categories will be much stronger than those that are relatively weak on them. And that would add the balance is important argument back, alongside with the case that the specific category is important. The results of a team with strength in a category is due to that strength and the relative strength in the other categories compared to teams not strong in that category and perhaps weaker on others too compared to the "strong" team. In Kevin's article, he provides the results of a regression: "Improving an offense by one standard deviation improves the team’s winning percentage by 0.654 standard deviations. By comparison, one standard deviation of defensive improvement only improves winning percentage by 0.460 standard deviations. (For the sake of posterity, the same improvement in offensive rebounding percentage produces a 0.204 deviation improvement, in defensive rebounding percentage a 0.158 deviation improvement.) This method shows offense to be about 40% more important than defense." This is a starting point not an ending one. A new regression study removing below average teams would seem to be called for to help with the analysis of relative importance of this 4 categories for winning teams. And using playoff data rather than regular season would seem appropriate as well since the game changes. It is worth noting that rebounding returns to a level of importance pretty close to its box score weight of importance. It would be interesting to know if strength of offense over defense in the league as a whole largely carries forward or weakens progressively step by step in the analysis of just playoff teams or teams that make the final 4 or win the championsip to square it with the strong ultimate performance of defensive biased teams in the championship winner /loser rank order study that started the thread. It could well be that offense is relatively more important for winning regular season games against the entire league while defense (and perhaps rebounding too I would be interested in checking) matter more in the playoffs. That seems in line with alot of convetional wisdom.Last edited by jambalaya on Tue Oct 18, 2005 2:48 pm; edited 1 time in total


Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 2 posts ] 

All times are UTC


Who is online

Users browsing this forum: Google [Bot] and 5 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum

Search for:
Jump to:  
Powered by phpBB® Forum Software © phpBB Group