Lakers 3pt Strategy

[italia13calcio]

With the Byron Scott saying he wants the Lakers to take between 10-15 threes a game (and with them averaging well below that in the preseason) I was wondering what your thoughts were. The general consensus in the analytics community seems to be that more threes is better, but is there any way you guys see this lack of threes being beneficial? Does it maybe fit their personnel better?

Along these lines, the Grizzlies ranked last in 3pt attempts last year, even with Hollinger leading analytics there and seemingly having a good amount of sway. Is that a case of personnel or is there some new research/studies that maybe shows threes aren't as good as thought?

I am confused by the Lakers decision (but maybe they're just being the Lakers, go Celtics!) but I am even more so by Memphis. Any insights into this would help calm my mind. Thanks

[gzchen]

The Grizzlies were a slow team in which three of their four best players are simply incapable of shooting 3's (and the pace might skew their rankings slightly in terms of 3pa/game). They were able to overcome that because those best players are very good defensive players, and they scored a lot of points in the paint. Unlike 2013-2014 Memphis, the Lakers actually have some capable 3-point shooters but, apparently, will choose not to shoot 3's, while on the other hand, they don't have the defense or ability to create shots at the rim that the Grizzlies do. It's not clear exactly how good the Lakers could be if they actually shoot a more optimal number of 3's (probably still bad), but it's very likely this teamwide strategy to shoot fewer 3's is not the optimal decision for them. This is especially true, since they could probably use the extra volatility as the probable underdog in most games.

[Statman]

With the Byron Scott saying he wants the Lakers to take between 10-15 threes a game (and with them averaging well below that in the preseason) I was wondering what your thoughts were. The general consensus in the analytics community seems to be that more threes is better, but is there any way you guys see this lack of threes being beneficial? Does it maybe fit their personnel better?

Along these lines, the Grizzlies ranked last in 3pt attempts last year, even with Hollinger leading analytics there and seemingly having a good amount of sway. Is that a case of personnel or is there some new research/studies that maybe shows threes aren't as good as thought?

I am confused by the Lakers decision (but maybe they're just being the Lakers, go Celtics!) but I am even more so by Memphis. Any insights into this would help calm my mind. Thanks

If you have some major usage guys who are strong at creating shots and especially strong at drawing fouls - then less three attempts is probably fine.

Most teams aren't in that boat - ball rotation and as many somewhat uncontested threes by good shooters as you can muster is probably is smarter plan. Of course, if you have a couple decent post guys - that'll open up their game when the defense has to keep extending.

If the Lakers shoot less threes - I assume that would require Kobe to create a ton. Drive to the hoop, create shots, draw fouls. I can't imagine his body holding up. Of course, if he turns into an constant 18 foot jump shooter - they are even in bigger trouble.

Lakers are in trouble no matter what I would guess.

[v-zero]

On the contrary I would say the three point shot is still undervalued within the analytics community as a whole. Whilst the fact that a made shot is worth three points rather than two, and that floor stretching is fantastic for helping your efficiency on the inside, especially with a stretch big man, are both well observed and accounted for, the fact that three point shots are more often available for offensive rebounds, and that they may well be easier to rebound for guards (bouncing further from the hoop), is rarely taken into account when the three point movement is spoken about at large.

Let's make a few basic assumptions to take a look at what this might mean. Let's say both 2pt and 3pt shots are offensively rebounded at the same rate, about 25%, and let's imagine two teams, one obsessed with two point shots and one obsessed with three point shots. Each will only take the type of shot which fits their obsession. Now, I'll ignore all sorts of things in this analysis for the purpose of making my point, but I don't think that invalidates my point.

So, what if we say that these two teams always have the same number of expected points per shot. Let's say each average 1pt per shot, for the sake of simplicity. This implies the 2pt guys are hitting shots at a 50% rate, the 3pt guys at 33.3% rate. So, they look equal, until we consider what happens if they miss. If they miss they have a 25% chance of an offensive rebound, so the chance a shot ends in an offensive rebound is 0.5*0.25 = 12.5% for the 2pt guys, and 0.667*0.25 = 0.167 for the 3pt guys. This then is a recursive process in which the possession will only end once the team either makes a shot or fails to make an offensive rebound from a missed shot. As such we can calculate the expected points per possession for each team recursively, and what do we get?

Expected points per 100 pos for the 2pt team: 114.3
Expected points per 100 pos for the 3pt team: 120

Now, that's a really huge difference... We can map the points per shot to points per 100 pos for both, and find that in the points per shot region of 0.7 to 1.3 there is a straight line relationship between points per shot and points per possession which says for the 2pt team they will gain 0.98 pts per pos for every 1pt per shot they gain, whereas the 3pt team will gain 1.08 pts per pos for every 1pt per shot they gain... The difference, in general, is absolutely huge.

Guess which team last year best combined offensive rebounding and quantity of three point shots? Portland, with the number 2 offense in the league.

[Mike G]

Last season's team correlations with 3PAr (3fga/FGA)

Code: Select all

TS%        .649        OpTO%     .101 
eFG%       .644        OpFT/FGA  .084 
ORtg       .484        DRtg      .014
Pace       .372        SOS       .001
PythW      .317        Op.eFG%  -.085
MOV        .315        
SRS        .312        DReb%    -.352
Age        .295        OReb%    -.364
Attendance .286            
FT/FGA     .284            
FTr        .247            
TO%        .240            

Does the 3-point strategy boost team TS%, or do teams with better shooters profit most by it?
It looks as if teams that either (a) have lots of good shooters on their roster, or (b) tend to keep their best shooters on the floor, in general suffer in the rebounding department.

They tend to have average DRtg; they suffer a turnover deficit; they get more FT/FGA. This last one is partly that better FG shooters are also better FT shooters.

Guess which team last year best combined offensive rebounding and quantity of three point shots? Portland, with the number 2 offense in the league.

Average team OReb% last year was 25.5, and avg TO% was 13.6
The 10 most 3-pt-happy teams:

Code: Select all

tm    3PAr   OReb%   TO%
Hou   .330   27.4   14.6
Atl   .316   21.0   14.3
NYK   .302   25.1   12.5
Brk   .301   21.7   14.1   
Phx   .300   26.4   14.0
Mia   .292   20.6   14.6
LAC   .291   25.0   12.7
GSW   .291   25.1   13.8
LAL   .291   20.2   13.7
Por   .290   28.0   12.4 

The Blazers were the 10th most prolific 3-shooting team.
They're one of 3 which were better than avg in turnovers.
The league's 4 worst OReb teams are here.

[Mike G]

Among individual players, correlations with 3PAr are pretty striking. From players w >500 minutes in 2013-14:

Code: Select all

(TO%   -.239)       ORtg    .016        Blk%    -.564
Ast%    .173        Usg%   -.088        DRb%    -.600
Stl%    .106        PER    -.315        TRb%    -.686
eFG%    .083       (DRtg    .367)       ORb%    -.726
TS%     .041        FTr    -.515            

While a given 3-pt specialist may have low FTr, apparently the 'spreading the floor' benefit includes FTA for teammates -- team correlations were positive.
On the flip side, these 3-pt guys tend to have few turnovers; and yet teams that shoot lots of 3's seem to get more TO. This implies perhaps that a bunch of catch-and-shoot players may not get tagged with the TO, and yet their main shot-creators and ball-handlers get stuck with the ball too often.

[Mike G]

The correlations between 3PAr and other ratios among all players of course includes players of all sizes.
In the past 2 seasons, players listed at 6'9" and 6'10" have a combined 160 player-seasons with at least 500 minutes.
Among just these tall guys:

Code: Select all

3fg%   .641        WS     -.075
-TO%   .389        WS/48  -.157
FT%    .346        DWS    -.196
Stl%   .109        PER    -.256
Ast%   .088        DRtg   -.352
ORtg   .007        2fg%   -.366
eFG%   .003       -DReb%  -.437
OWS   -.001        Blk%   -.457
TS%   -.015        FG%    -.558
Usg%  -.035        TReb%  -.621
MP    -.052        OReb%  -.703

It would seem that most big men who play closer to the basket on offense, also play inside more on defense. Or at least they're more effective.

[Statman]

On the contrary I would say the three point shot is still undervalued within the analytics community as a whole. Whilst the fact that a made shot is worth three points rather than two, and that floor stretching is fantastic for helping your efficiency on the inside, especially with a stretch big man, are both well observed and accounted for, the fact that three point shots are more often available for offensive rebounds, and that they may well be easier to rebound for guards (bouncing further from the hoop), is rarely taken into account when the three point movement is spoken about at large.

Let's make a few basic assumptions to take a look at what this might mean. Let's say both 2pt and 3pt shots are offensively rebounded at the same rate, about 25%, and let's imagine two teams, one obsessed with two point shots and one obsessed with three point shots. Each will only take the type of shot which fits their obsession. Now, I'll ignore all sorts of things in this analysis for the purpose of making my point, but I don't think that invalidates my point.

So, what if we say that these two teams always have the same number of expected points per shot. Let's say each average 1pt per shot, for the sake of simplicity. This implies the 2pt guys are hitting shots at a 50% rate, the 3pt guys at 33.3% rate. So, they look equal, until we consider what happens if they miss. If they miss they have a 25% chance of an offensive rebound, so the chance a shot ends in an offensive rebound is 0.5*0.25 = 12.5% for the 2pt guys, and 0.667*0.25 = 0.167 for the 3pt guys. This then is a recursive process in which the possession will only end once the team either makes a shot or fails to make an offensive rebound from a missed shot. As such we can calculate the expected points per possession for each team recursively, and what do we get?

Expected points per 100 pos for the 2pt team: 114.3
Expected points per 100 pos for the 3pt team: 120

Now, that's a really huge difference... We can map the points per shot to points per 100 pos for both, and find that in the points per shot region of 0.7 to 1.3 there is a straight line relationship between points per shot and points per possession which says for the 2pt team they will gain 0.98 pts per pos for every 1pt per shot they gain, whereas the 3pt team will gain 1.08 pts per pos for every 1pt per shot they gain... The difference, in general, is absolutely huge.

Guess which team last year best combined offensive rebounding and quantity of three point shots? Portland, with the number 2 offense in the league.

You don't mention free throw rate at all. Pretty important factor to ignore.

[italia13calcio]

Agreed, you can't ignore free throw rate which is much higher at the basket. I was thinking about this and there's multiple things you need to account for:

Definitely consider: Shooting percentage from each spot, Offensive rebounding percentage for each shot type (higher for shots closer to the rim), whether you are fouled, whether you make the shot while being fouled

Maybe consider: Turnovers (a turnover at the top of the key is much more costly than one under the basket as a team could have a quick fast break) and defense in general

Any other things?

I had found statistics on everything but and-1 opportunities by position. Anyone know where to find that?

[permaximum]

http://www.refanalytics.com/2014/05/14/ ... bad-thing/

Any statistical evidence on free throw rate difference between 3pt and 2pt shots? I couldn't find anything after a rough search... However there are people who work with PBP data here so, some of them may shed light on this issue.

BTW, free throws attempted per offensive play sits around 21%-22%.

[Mike G]

Of players listed as Forward (or F-G; not F-C) at b-r.com, there are 289 player-seasons in the last 2 years with >500 minutes.
Breaking them out by their 3PAr -- 3fga/fga -- and totaling each group, here are their efficiencies and per36 minute rates:

Code: Select all

3fga/fga   ss#   Age   fta/fga   GS%    mpg    2fga   2fg%    3fga   3fg%   eFG%   FTA    FT%
.50+       37    28.9    .15    .39    21.9     4.4    .48    6.2    .39    .54    1.6    .78
.40-.50    32    27.5    .22    .50    25.0     6.4    .49    5.1    .37    .52    2.5    .78
.30-.40    37    25.2    .28    .50    24.6     8.1    .47    4.4    .35    .49    3.5    .78
.20-.30    45    25.7    .30    .63    27.9    10.6    .49    3.3    .34    .50    4.2    .79
.10-.20    19    26.3    .27    .57    24.9    10.9    .51    2.0    .33    .50    3.4    .72
<.10      119    26.3    .32    .48    22.5    11.3    .51     .2    .22    .50    3.6    .69
            
                                    
3fga/fga   ss#    TSA     TS%     Pts    ORb    DRb    TRb    Ast    Stl    Blk     TO     PF
.50+       37    11.3    .561    12.7    1.0    4.1    5.1    2.3    1.1     .6    1.4    2.9
.40-.50    32    12.7    .550    13.9    1.2    4.5    5.7    2.3    1.1     .5    1.6    2.8
.30-.40    37    13.9    .535    14.9    1.3    4.8    6.1    2.7    1.3     .6    1.9    2.8
.20-.30    45    15.7    .545    17.2    1.7    5.3    7.0    2.7    1.3     .7    2.1    2.7
.10-.20    19    14.4    .537    15.5    1.8    5.4    7.2    2.8    1.1     .9    2.0    2.7
<.10      119    13.1    .537    14.1    3.1    6.5    9.6    1.8    1.0    1.2    1.9    3.6

ss# is the number of players in each group
Age is just avg age in the group, not minutes weighted.
GS%: Games Started % = GS/G
TSA : True Shot Attempts = FGA + .44*FTA

[Crow]

Mike G, thanks for the interesting correlations.

Permaximum there ate foul rate estimates in this article http://www.82games.com/locations.htm

v-zero, your hypothetical needs turnover rates, which vary for 2 and 3 pointers, as also indicated in that link (which is mentioned in the newer thread as well).

Ideally in the future the best databases would show ftas and turnovers associated with different floor locations (and types of shoots). And offensive rebounds as Italia suggests.

While Scott's comments have been criticized on their face and appear sincere given the preseason, the real test is the regular season. I don't think he was trying to be deceptive but that might be an interesting strategy or an unintended benefit that the players may eventually taken advantage of (and Scott could come back and try to take credit for).

[permaximum]

Thanks for the link Crow. It's interesting such an old article which I've read before still comes handy at such situations. 82games really started something in this field... at least for amateurs like me.

[v-zero]

v-zero, your hypothetical needs turnover rates, which vary for 2 and 3 pointers, as also indicated in that link (which is mentioned in the newer thread as well).

I was waiting for somebody to say this. It was a basic comparison to point out something often ignored. Adding in TO% will not change the conclusion, but will actually favour the three point shot slightly more, so I see no need to redo anything.

Numerous valid things have been said, however that was why I said this:Now, I'll ignore all sorts of things in this analysis for the purpose of making my point, but I don't think that invalidates my point. I'm not going to exhaustively analyse it as that was not my purpose, my purpose was to point out an oft-ignored but very important difference between two point shots and three point shots.

[Mike G]

3-point specialists are among the weakest offensive rebounders in the league, at any position.
They're almost as bad at defensive rebounding.

3-point leading teams are among the weakest OReb teams, in general. Not surprising, given that their 3-pt shooters are almost all weak at OReb.

3-pt teams are also worse than avg in the TO department.

Given these facts, how is it possible that 3FGx are rebounded by offense as often as 2FGx?