Usage and efficiency stats

[Mike G]

From this page -- http://www.basketball-reference.com/lea ... anced.html
-- and pruning it down to players with at least 200 minutes this year, the mid-season leaders in Usage% and Offensive Rating are:

Code: Select all

ORtg   O Rating         Tm   Usg%      Usg%    Usage          Tm   ORtg
133   Brandan Wright   DAL   18.0      32.9  DeMarcus Cousins SAC   107
127   Chris Andersen   MIA   14.2      32.9 Russell Westbrook OKC   103
127   Darius Miller    NOP    9.6      31.9   Kevin Durant    OKC   125
126   Josh Harrellson  DET   12.6      31.9   Carmelo Anthony NYK   112
125   Kevin Durant     OKC   31.9      31.2   Derrick Rose    CHI    88
125   Robin Lopez      POR   14.0      29.9   LeBron James    MIA   121
124   Chris Paul       LAC   25.0      29.5 LaMarcus Aldridge POR   111
124   Wesley Matthews  POR   18.8      29.3   Kyrie Irving    CLE   108
124 Miroslav Raduljica MIL   17.3      28.6   Stephen Curry   GSW   113
122   Jose Calderon    DAL   15.7      28.4   Paul George     IND   112

On the left, top O-Rtg, are just 2 above-average 'users', and some who seldom 'use', in spite of their efficiency.
On the right, offensive stars. It makes sense that top players would take the most shots, in general.
Is it practical to fuse these 2 stats? Like Usage * Efficiency?
Ranked by Usg% * ORtg/105.8 ( = league avg ORtg) --

Code: Select all

U*OR     Offense       Tm   Usg%  ORtg     U*OR     Offense     Tm   Usg% ORtg
37.7   Kevin Durant    OKC   32   125      30.1   Paul George   IND   28   112
34.2   LeBron James    MIA   30   121      29.9   Kyrie Irving  CLE   29   108
33.8   Carmelo Anthony NYK   32   112      29.4   James Harden  HOU   27   114
33.3  DeMarcus Cousins SAC   33   107      29.3   Chris Paul    LAC   25   124
32.0 Russell Westbrook OKC   33   103      29.3   Blake Griffin LAC   27   113
31.6   Kevin Love      MIN   28   121      28.6   Isaiah Thomas SAC   26   115
30.9 LaMarcus Aldridge POR   30   111      28.3   Tony Parker   SAS   27   113
30.9   Brook Lopez     BRK   27   121      28.2   DeMar DeRozan TOR   28   107
30.5   Stephen Curry   GSW   29   113      27.7  Jamal Crawford LAC   27   107
30.2   Dirk Nowitzki   DAL   27   117      27.7   Dwyane Wade   MIA   27   108

Does this basically represent the most proficient (productive and efficient) offensive players in the league? I don't see any slackers, interlopers, or quirks.

We do see Westbrook just above Love. A high-usage guy over a high-efficiency guy.
What if we give efficiency a boost, relative to Usage? Change the formula to Usg% * (ORtg/105.8)^2

Code: Select all

k2       Player        Tm   Usg%  ORtg      k2      Player        Tm   Usg% ORtg
44.5   Kevin Durant    OKC   32   125      31.8   Paul George     IND   28   112
39.1   LeBron James    MIA   30   121      31.7   James Harden    HOU   27   114
36.1   Kevin Love      MIN   28   121      31.3   Blake Griffin   LAC   27   113
35.7   Carmelo Anthony NYK   32   112      31.2 Russell Westbrook OKC   33   103
35.3   Brook Lopez     BRK   27   121      31.1   Isaiah Thomas   SAC   26   115
34.3   Chris Paul      LAC   25   124      30.5   Kyrie Irving    CLE   29   108
33.7  DeMarcus Cousins SAC   33   107      30.5   Damian Lillard  POR   25   118
33.4   Dirk Nowitzki   DAL   27   117      30.2   Tony Parker     SAS   27   113
32.6   Stephen Curry   GSW   29   113      30.0   Goran Dragic    PHO   24   118
32.5 LaMarcus Aldridge POR   30   111      29.8   Anthony Davis   NOP   24   117

Love has surged past Westbrook. Maybe an exponent between 1 and 2 is ideal.

A more complete rating of course involves defense. There's more guesswork, but player Defensive Rating is surely better than nothing. We can substitute DRtg for the league avg ORtg in the denominator.
uod = Usg% * ORtg/DRtg , and uod2 = Usg% * (ORtg/DRtg)^2

Code: Select all

uod      Player        Tm   Usg%   O/D      uod2     Player        Tm   Usg%  O/D2
39.9   Kevin Durant    OKC   32   1.25      49.8   Kevin Durant    OKC   32   1.56
34.5  DeMarcus Cousins SAC   33   1.05      39.7   LeBron James    MIA   30   1.33
34.5   LeBron James    MIA   30   1.15      39.5   Paul George     IND   28   1.39
33.6 Russell Westbrook OKC   33   1.02      38.8   Kevin Love      MIN   28   1.41
33.5   Paul George     IND   28   1.18      36.9   Chris Paul      LAC   25   1.48
33.1   Carmelo Anthony NYK   32   1.04      36.2  DeMarcus Cousins SAC   33   1.10
32.7   Kevin Love      MIN   28   1.19      35.1   Stephen Curry   GSW   29   1.23
31.7   Stephen Curry   GSW   29   1.11      34.3   Carmelo Anthony NYK   32   1.08
31.5 LaMarcus Aldridge POR   30   1.07      34.2 Russell Westbrook OKC   33   1.04
30.4   Chris Paul      LAC   25   1.22      33.9   Brook Lopez     BRK   27   1.26
30.4   Blake Griffin   LAC   27   1.11      33.6   Blake Griffin   LAC   27   1.23
30.3   Brook Lopez     BRK   27   1.12      33.6 LaMarcus Aldridge POR   30   1.14
29.9   Dirk Nowitzki   DAL   27   1.09      32.6   Dirk Nowitzki   DAL   27   1.20
29.1   James Harden    HOU   27   1.07      31.5   Anthony Davis   NOP   24   1.29
29.0   Kyrie Irving    CLE   29    .99      31.0   James Harden    HOU   27   1.14
28.4   DeMar DeRozan   TOR   28   1.02      30.0   Tim Duncan      SAS   25   1.19
28.1   Dwyane Wade     MIA   27   1.04      29.6   Tony Parker     SAS   27   1.12
28.1   John Wall       WAS   28   1.02      29.2   David Lee       GSW   24   1.20
28.0   Al Jefferson    CHA   27   1.03      29.2   Dwyane Wade     MIA   27   1.08
28.0   Tony Parker     SAS   27   1.06      29.0   DeMar DeRozan   TOR   28   1.04

Does one list 'look better' than the other? Squaring the O/D ratio, we lose Irving, Wall, and Jefferson; replaced by Duncan, Davis, and Lee.

[v-zero]

Obviously Usage*Efficiency is basically just points, and we know that they alone are not a good indicator.

[Mike G]

Hey, that's pretty good. Certainly, at a team level, that's all there is to it.
Player ORtg includes turnovers and offensive rebounds, and their resultant (estimated) points or loss of points.
Efficiency is generally considered to be a ratio, and here I've offered various possibilities: eventually, it's points as a fraction of opponent points, again including those estimated via O-Reb and TO.

Durant scores 31.3 points in 38.1 minutes. That's a rate of 39.6 per 48.2 minutes (an avg OKC game).
OKC Opponents avg 98.0
Durant therefore averages 40.4% of what opponents score while he's on the court. That's very close to his 39.9 in the last table above (Usg*ORtg/DRtg)

DMC gets 22.6 Pts in 32 Min. In a Sacto game (48.45 Min) of time, he scores 34.2
In that same span of time, opponents score 104.8 -- Cousins scores 32.6% as much.

Paul George plays for a team that allows just 90.2 ppg. His 23.5 in 36.6 Min is = 31.2 in 48.2 Min
He scores 34.6% of what Pacers surrender.

Is George some 3 Pts/48 less of a scorer than Cousins? Or is he 2 Pts better?
That's a 5 point swing between 'just points' and 'points relative to the opponent'.
Include OReb and TO for additional fine tuning.

[v-zero]

..."basically just points"...

In the end a high volume of poor shooting is rewarded, but it's not quite just points, I'll give you that one.

Anyway, I'm not about to rehash old arguments, so I'll get back under my bridge.

[Mike G]

It just seems that Usg% and ORtg have languished without much universal utility. And yet in combinations, they can reconstruct some quality that in the OP I called offensive proficiency -- proactive efficiency. In contrast to the 'high efficiency, low usage' player who seems to only be on the court and do nothing the vast majority of the time, the medium-high player in both areas is likely to get more minutes and actually boost the team offensive efficiency more.

Throwing the player DRtg in there is also interesting. Has anyone made use of the ratio ORtg/DRtg?
Highest and lowest this year (>200 min) --

Code: Select all

O/D      Player      Tm   ORtg   DRtg     O/D     Player       Tm  ORtg  DRtg
1.25   Kevin Durant  OKC   125   100      .62  Anthony Bennett CLE   66   106
1.24  Brandan Wright DAL   133   107      .64   Marquis Teague CHI   68   106
1.23  Chris Andersen MIA   127   103      .72   Otto Porter    WAS   77   107
1.22   Chris Paul    LAC   124   102      .73   Tyshawn Taylor BRK   80   110
1.20  DeAndre Jordan LAC   118    98      .73   Gal Mekel      DAL   83   113
1.19   Kevin Love    MIN   121   102                  
1.18   Paul George   IND   112    95      O/D    >1000 min.    Tm  ORtg  DRtg
1.18   Andrew Bogut  GSW   113    96      .85   O.J. Mayo      MIL   95   112
1.17   Jeff Adrien   CHA   117   100      .86   Dion Waiters   CLE   94   109
1.16   Kyle Lowry    TOR   121   104      .87   Ben McLemore   SAC   97   112
1.16   George Hill   IND   115    99      .87  Tayshaun Prince MEM   95   109
1.16   Robin Lopez   POR   125   108      .89   J.R. Smith     NYK   98   110
1.16 Josh Harrellson DET   126   109      .90   Victor Oladipo ORL   94   105
1.15   Nick Collison OKC   120   104      .90   Trey Burke     UTA  102   113
1.15   LeBron James  MIA   121   105      .90   Gerald Wallace BOS   94   104
1.15 Bismack Biyombo CHA   115   100      .90   Josh Smith     DET   95   105
1.15  Andre Drummond DET   117   102      .91   Evan Turner    PHI  100   110
1.14  Tiago Splitter SAS   116   102      .91   Brandon Knight MIL  100   110
1.14  Kris Humphries BOS   116   102      .92   Glen Davis     ORL   97   106
1.14   Anthony Davis NOP   117   103      .92   Avery Bradley  BOS   98   107
1.14   Joakim Noah   CHI   109    96      .92   Enes Kanter    UTA  101   110
                 

All the very lowest O/R players are also low-minutes guys. So I jumped to those over 1000 min (there are 151, 5 per team)

[Mike G]

Ranked by the formula Usg%*(ORtg/DRtg)^2 , the top 30 player-seasons since 1978

Code: Select all

uod2   Player  Season  Usg%  ORtg  DRtg      uod2   Player  Season  Usg%  ORtg  DRtg
51.3   LeBron   2009   33.8   122    99      45.4   Jordan   1990   33.7   123   106
51.2   Jordan   1996   33.3   124   100      45.0   Robinson 1996   28.8   120    96
50.6   Jordan   1988   34.1   123   101      44.6   Malone   1997   32.7   118   101
49.4   Jordan   1991   32.9   125   102      44.6   Jordan   1992   31.7   121   102
49.0  *Durant   2014   31.9   124   100      44.4   Durant   2013   29.8   122   100
48.5   Jordan   1987   38.3   117   104      43.9   Robinson 1995   29.9   120    99
47.4   LeBron   2012   32.0   118    97      43.9   Garnett  2004   29.6   112    92
47.2   Jordan   1993   34.7   119   102      43.8   Jordan   1998   33.7   114   100
47.2   Robinson 1994   32.0   119    98      43.8   McGrady  2003   35.2   116   104
47.1   LeBron   2010   33.5   121   102      43.7   Robinson 1998   29.7   114    94
46.7   Jordan   1997   33.2   121   102      43.7   Shaq     2002   31.8   116    99
46.3   LeBron   2013   30.2   125   101      43.6   Malone   1990   32.6   118   102
45.8   Jordan   1989   32.1   123   103      43.4   Wade     2009   36.2   115   105
45.7   Shaq     2000   31.2   115    95      43.0   Yao      2007   33.5   111    98
45.6   Kobe     2006   38.7   114   105      42.9   Duncan   2004   29.7   107    89

League ORtg/DRtg = 1, so isn't a factor.
Only player-seasons of >1000 minutes and Usg%>25 are considered.

And top values by players in a season:

Code: Select all

uod2   Player  Season  Usg%  ORtg  DRtg      uod2   Player  Season  Usg%  ORtg  DRtg
51.3   LeBron   2009   33.8   122    99      38.5   Erving   1981   28.4   113    97
51.2   Jordan   1996   33.3   124   100      38.3   Carmelo  2013   35.6   112   108
49.0  *Durant   2014   31.9   124   100      38.2   Mourning 2000   28.1   112    96
47.2   Robinson 1994   32.0   119    98      38.1  Westbrook 2013   32.8   111   103
45.7   Shaq     2000   31.2   115    95      37.6   Gervin   1982   35.0   114   110
45.6   Kobe     2006   38.7   114   105      37.5   Parish   1981   27.1   113    96
44.6   Malone   1997   32.7   118   101      37.3   Wilkins  1987   32.3   115   107
43.9   Garnett  2004   29.6   112    92      37.3   Brand    2006   27.7   116   100
43.8   McGrady  2003   35.2   116   104      37.2   Moses    1983   26.0   116    97
43.4   Wade     2009   36.2   115   105      37.0  *Cousins  2014   33.0   107   101
43.0   Yao      2007   33.5   111    98      37.0   Drexler  1992   28.7   117   103
42.9   Duncan   2004   29.7   107    89      37.0   Webber   2001   31.6   105    97
42.8   Nowitzki 2006   30.0   123   103      37.0   Dantley  1986   30.0   121   109
41.2   Iverson  2001   35.9   106    99      36.9   J O'Neal 2005   36.2   102   101
40.6   Olajuwon 1993   28.8   114    96      36.9   Carter   2001   30.7   114   104
40.2   Ewing    1994   29.8   108    93      36.9   King     1984   28.7   119   105
40.1   Amar'e   2008   28.2   124   104      36.8   Hill     1997   28.3   114   100
39.9   Paul     2009   27.5   124   103      36.5   Sabonis  1996   25.5   116    97
39.5  *George   2014   28.4   112    95      36.2   J Drew   1978   31.0   107    99
39.4   Bird     1988   30.2   121   106      36.0   Magic    1987   26.3   124   106
39.3   Dwight   2011   27.2   113    94      36.0   Brandon  1996   27.6   121   106
39.2   Barkley  1991   29.1   123   106      35.9   P Pierce 2003   33.2   104   100
39.0   Kareem   1978   27.0   113    94      35.7   Kemp     1999   28.9   110    99
38.8   Rose     2011   32.2   113   103      35.6   R Pierce 1994   30.8   114   106
38.6   Ginobili 2007   27.1   117    98      35.4   Cummings 1985   28.8   112   101
38.6  *Love     2014   27.4   121   102      35.4   Westphal 1978   30.9   107   100

[Mike G]

A similar list of best playoffs, according to Usg*(ORtg/DRtg)^2
Minimum 500 postseason minutes.

Code: Select all

uod2    Player  yr    Usg%  ORtg   Rtg      uod2    Player   yr    Usg%  ORtg  DRtg
59.6   LeBron  2009   36.4   128   100      40.9   Malone   1996   34.9   105    97
51.7   Jordan  1991   32.7   127   101      40.8   Malone   1998   34.1   105    96
50.0   Jordan  1990   36.1   120   102      40.5   LeBron   2008   34.7   107    99
48.8   Jordan  1993   38.0   119   105      40.5   Kareem   1980   28.5   118    99
48.8   Jordan  1996   32.9   123   101      40.2   Shaq     2002   31.4   112    99
47.2   Jordan  1998   36.6   117   103      39.9   LeBron   2013   29.2   118   101
47.1   Jordan  1989   35.4   120   104      39.6   Olajuwon 1994   31.4   109    97
45.9   LeBron  2012   33.4   116    99      39.3   Moses    1983   26.0   118    96
45.4   Jordan  1992   37.1   115   104      39.0   Nowitzki 2006   26.9   124   103
45.2   Jordan  1997   35.5   114   101      38.7   Wade     2005   33.9   109   102
44.3   Shaq    2001   32.0   113    96      38.4   Nowitzki 2011   32.0   115   105
42.0   Duncan  2003   26.4   116    92      38.3   Duncan   1999   26.8   110    92
41.9   Wade    2006   31.7   115   100      38.2  GWilliams 1979   31.1   112   101
41.6   Kobe    2009   32.9   117   104      38.0   Rose     2011   35.2   107   103
41.6   Kobe    2001   30.3   116    99      37.9   Olajuwon 1986   27.3   119   101

And top postseasons by players:

Code: Select all

uod2    Player   yr    Usg%  ORtg  DRtg      uod2    Player   yr    Usg%  ORtg  DRtg
59.6   LeBron   2009   36.4   128   100      34.4   Barkley  1993   27.3   119   106
51.7   Jordan   1991   32.7   127   101      34.3   Kemp     1996   27.3   111    99
44.3   Shaq     2001   32.0   113    96      34.0   Garnett  2008   26.6   112    99
42.0   Duncan   2003   26.4   116    92      33.9   Pierce   2002   30.1   103    97
41.9   Wade     2006   31.7   115   100      33.8   Drexler  1992   28.4   120   110
41.6   Kobe     2009   32.9   117   104      33.8   Isiah    1987   29.2   114   106
40.5   Kareem   1980   28.5   118    99      33.4   English  1985   26.8   125   112
39.6   Olajuwon 1994   31.4   109    97      33.2   Brand    2006   25.1   122   106
39.3   Moses    1983   26.0   118    96      33.0   Carter   2001   29.6   112   106
39.0   Nowitzki 2006   26.9   124   103      33.0   Ewing    1993   29.5   109   103
38.2  GWilliams 1979   31.1   112   101      33.0   Hamilton 2004   27.6   105    96
38.0   Rose     2011   35.2   107   103      32.8   Erving   1980   27.6   108    99
37.4   Ginobili 2005   26.3   124   104      32.7   Pippen   1991   25.2   114   100
37.4   Reggie   1994   28.0   119   103      32.7   Smits    1995   26.6   122   110
36.9   Malone   1994   29.5   113   101      31.7   Boozer   2007   27.3   112   104
36.7   Iverson  2003   36.7   105   105      31.7   Jermaine 2004   27.9    98    92
36.3   Carmelo  2009   31.5   116   108      31.0   F Brown  1978   28.8   109   105
36.2   Robinson 1995   29.8   108    98      30.8   W Davis  1984   27.3   119   112
36.1   Amar'e   2005   30.4   122   112      30.7   Russell  2011   34.4   103   109
35.9   Durant   2011   29.0   119   107      30.6   Parker   2013   28.9   108   105
35.8   Gervin   1979   28.8   116   104      30.0   Zach     2011   26.4   112   105
35.2   Bird     1984   25.9   120   103                     

The qualifier Usg% > 25 may be a bug here. Magic doesn't appear. Pippen barely got in; Moses and Bird by not much.
http://bkref.com/tiny/1EPZH

[Metsox]

Thought this might be the right place to post this...

http://regressing.deadspin.com/the-dura ... 1515940922

[Mike G]

That's interesting, thanks. It's not surprising that others have noticed what Durant and LeBron have done in the last 1.5 years.

However, the article frequently refers to TS% as 'efficiency'. But that's just shooting. There's also a player's turnovers to consider.
On one of the charts at that link, Artis Gilmore is shown above the pack of data points, with his splendid TS% near .700. But the A-Train also had a TO% in the 17-19 range. He fumbled away more than 1/6 of his used possessions.

That's why I used ORtg, instead of TS%: It incorporates TO.
When we balance ORtg with DRtg -- and DRtg incorporates Steals -- it's especially meaningful to include TO among 'offensive contributions', along with shooting efficiency.

A player can just increase his turnovers, and thus boost his Usg%, without damaging his TS%. This could push him toward or beyond the Zamir Horizon, even though his offense is suffering, by any definition.

Here's short list of 12 players -- Barkley, Bird, Dantley, Durant, Harden, LeBron, Jordan, Malone, Martin, Yao, Shaq, and Robinson -- who have managed to have a season with Usg% > 29 and TS% > .600
http://bkref.com/tiny/1VT8N
They total 21 player-seasons. Jordan, LeBron, and Durant are in there 4 times, and everyone else just once.

If you sort by TO%, note that Jordan takes the top 3 spots. At the bottom (worst) are Harden '13 and LeBron '14.
Durant '14 is a bit worse than the median in this elite group. In fact, the Big 2 dominate the bottom half of the table.

This keeps their ORtg at less than superhuman level. They aren't extreme outliers after all. Just somewhere out there with the other alltime greats.
They're young. Give 'em time.

[schtevie]

If you slightly lower the USG% to 27.8 however, you get an elite list of 34 player seasons, fifteen different players I think, and good old Adrian Dantley leads the list with 5 appearances. (Couldn't. Resist.)

http://www.basketball-reference.com/pla ... _by=season

[Mike G]

That's right; AD may be quite overrated by TS%.
By Usg%*(ORtg/DRtg)^2, Dantley's highest mark falls between those of Webber and Jermaine; behind about 38 others.
See 5 posts up.

[schtevie]

From all I have looked at, box score stats and on/off data (in terms of games played), it is very difficult to tell a coherent and persuasive story that Adrian Dantley was not a great offensive player. And TS% is as good of a single marker as any of his excellence.

Considering anecdotal and the same on/off data, it is, at the same time, very difficult to tell a coherent and persuasive story that Adrian Dantley was not (in his Utah prime anyway) a really bad defensive player. DRtg is not a good marker in the instance, and more generally isn't a good metric for this type of exercise.

I am not sure what Usg%*(ORtg/DRtg)^2 is intended to rank, exactly, but if it is greatest, individual seasons of play, I am quite sure that AD at 38 is very, very incorrect.

[Mike G]

(ORtg/DRtg)^2*Usg% would seem to be a very good marker for this sort of exercise. Much better than any single component or sub-component.

Again referencing the table above:
- Shaq's 2000 season is shown as his best. In ORtg and Usg% it's just 7th; in DRtg easily #1; TS% was 13th best of his career!

- For KG, 2004 shows up as his best. It was only his 5th best ORtg; highest Usg% and best DRtg; 9th best TS%.

- 1994 David Robinson: His 4th best ORtg, highest Usg%, 11th best DRtg, 7th best TS%

- 2006 Kobe: #4 ORtg, #1 Usg%, #10 DRtg, #6 TS%

- 1997 Malone: #4 ORtg, #1 Usg%, #11 DRtg, #3 TS%

I believe these are all MVP seasons

[schtevie]

It is an enduring mystery why ORtg and DRtg are still considered to be equally good approximations of what they purport to measure. They aren't. Full stop. ORtg is pretty good, whereas DRtg is really quite terrible when put to present purprose (Dean - if this is his nomenclature, I cannot recall - really ought to have labeled it DRtg*.)

And it is really not hard at all to persuade yourself of the terribleness of DRtg as a "precise" measure of individual defensive contribution. Go to http://stats-for-the-nba.appspot.com/ and see how DRtgs correlate with these data, which actually measure defense. And in particular, make note of the positional "bias" in DRAPM, the most striking feature of which is that true defensive greatness only potentially obtains to those with the privilege of protecting the paint. This is a structural feature of basketball that is obscured in the formulation of DRtg.

[Crow]

Yes Defensive Rating is very crude and imprecise.

I wonder if Dean Oliver will develop better ratings using the vast resources at ESPN to at least take the ratings down to play by play precision and maybe incorporate the new data from synergy and sportsVU. John Hollinger was never willing to advance PER, preferring to accept its limitations and age. I know ESPN has added some advanced analytics and intends to go slowly because only a small part of their audience really wants much advanced analytics but there is much they could do if they want to lead.