improved players from 2023 to 2024

[Crow]

OK, thanks. I forgot where GmBPM was.

[Mike G]

https://www.basketball-reference.com/le ... otals.html
It's a rout!

Code: Select all

eWin+   per36      tm   Eff%   Sco   Reb   Ast   e484       Eff%   Sco   Reb   Ast   e484
2.57   Shai G-A   OKC  .634   34.9   5.7   5.2   3.01      .610   30.4   4.8   4.5   2.28
2.05   A Şengün   Hou  .582   23.7  10.1   4.5   2.04      .588   16.5  11.7   3.8   1.43
1.94   S Barnes   Tor  .563   18.5   8.4   4.3   1.51      .518   13.9   7.3   4.2    .98
1.60   T Maxey    Phl  .576   24.0   3.4   4.9   1.69      .597   21.4   3.4   3.2   1.25
1.57  V Williams  Mem  .602   11.8   7.9   2.4    .80      .366    6.7   4.8   1.1    .02

1.54   C Sexton   Uta  .599   23.7   3.9   4.8   1.50      .604   19.8   3.3   3.4    .92
1.50 Karl-A Towns Min  .630   27.2  10.1   2.7   2.02      .607   21.8   9.0   4.2   1.59
1.47  P Banchero  Orl  .529   21.7   8.0   4.1   1.55      .517   18.4   7.8   3.2   1.16
1.45 Jal Williams OKC  .615   21.4   4.3   3.9   1.43      .594   15.6   5.2   3.2    .97
1.40   A Edwards  Min  .575   27.4   5.7   4.4   1.83      .555   22.3   5.9   3.6   1.44

eWin+   per36      tm   Eff%   Sco   Reb   Ast   e484       Eff%   Sco   Reb   Ast   e484
1.37  Coby White  Chi  .580   18.1   4.8   4.0   1.03      .568   13.8   4.4   3.4    .69
1.34  DiVincenzo  NYK  .608   18.8   5.1   2.8   1.27  GSW .589   11.8   6.1   3.8    .75
1.25   F Wagner   Orl  .556   20.8   6.9   3.5   1.49      .581   19.1   4.8   3.1   1.07
1.23   J Brunson  NYK  .588   27.6   4.2   5.2   1.93      .587   23.9   3.8   5.4   1.61
1.19  D Robinson  Mia  .607   16.5   3.3   3.0    .74      .530   12.1   3.8   2.0    .31

1.11   A Caruso   Chi  .625   13.6   4.9   2.8    .87      .580    8.1   4.6   3.5    .42
1.11   J Walker   Por  .560   11.7  10.8   1.1    .90      .490    9.5   7.6   1.6    .37
1.06  D Schröder  Tor  .557   14.3   3.1   5.3    .80  LAL .535   13.2   2.9   4.3    .48
1.02 T Haliburton Ind  .619   23.3   4.7   9.4   2.09      .616   20.6   3.9   8.9   1.71
.99   Jab. Smith  Hou  .576   14.2   9.7   1.3   1.02      .508   11.9   8.7   1.2    .70

What disqualifies Shai from taking this award?

[Crow]

Voter awareness, tradition.


V Williams got large opportunity in yr 2 he didn't in yr 1.

[Mike G]

Season at 64%
Players whose productions are down but have gotten minutes. The eWin deficit may be doubled to estimate how many more wins their team might have now, given the same minutes at last season's rates.

Code: Select all

eWin+   per36      tm   Eff%   Sco   Reb   Ast   e484       Eff%   Sco   Reb   Ast   e484
-1.99  D Lillard  Mil  .586   24.3   4.3   5.2   1.56  Por .630   30.4   4.9   6.0   2.17
-1.82  J Holiday  Bos  .580   14.7   6.2   4.1    .89  Mil .580   20.6   5.4   6.9   1.48
-1.69   C Wood    LAL  .560   12.3  10.2   1.4    .76  Dal .612   22.1  10.5   2.2   1.77
-1.62   J Poole   Was  .509   15.2   2.6   3.0    .48  GSW .564   22.0   3.3   4.3   1.05
-1.58   B Lopez   Mil  .599   13.8   6.3   1.4    .85      .623   19.0   7.6   1.3   1.37

-1.32   D Ayton   Por  .568   14.3  12.4   1.4   1.19  Phx .609   21.2  12.1   1.7   1.80
-1.30  J Jackson  Mem  .550   23.3   6.3   1.9   1.44      .602   23.1   8.4   1.0   1.89
-1.26   A Gordon  Den  .582   15.3   7.7   2.6   1.01      .604   19.0   8.3   2.9   1.46
-1.25  A Wiggins  GSW  .513   13.6   5.9   1.6    .57      .560   17.1   5.6   2.1   1.10
-1.21  K Thompson GSW  .550   17.5   4.1   2.0    .75      .572   21.6   4.5   2.1   1.20

eWin+    per36     tm   Eff%   Sco   Reb   Ast   e484       Eff%   Sco   Reb   Ast   e484
-1.16   G Trent   Tor  .543   13.3   2.9   1.4    .46      .553   17.7   3.1   1.5    .94
-1.13   J Butler  Mia  .614   23.9   6.0   4.0   1.77      .629   25.3   6.9   5.0   2.24
-1.05   K Dunn    Uta  .548    9.3   6.0   6.2    .74      .598   16.8   6.2   6.2   1.41
-1.01  O Okongwu  Atl  .665   13.3   9.3   1.3    .94      .665   14.9  11.1   1.2   1.34
-.98   St Curry   GSW  .630   31.0   4.9   3.9   1.90      .647   30.4   6.4   5.2   2.24

-.97   A Davis    LAL  .604   24.5  12.3   2.9   2.26      .613   26.6  13.1   2.2   2.56
-.97   B Hield    Ind  .584   14.4   4.6   2.6    .71      .610   17.8   5.8   2.6   1.09
-.93   D DeRozan  Chi  .559   20.6   4.1   3.9   1.34      .580   23.6   4.7   4.1   1.61
-.88   B Portis   Mil  .568   16.2  10.2   1.1   1.24      .570   18.3  12.7   1.7   1.61
-.86   N Powell   LAC  .628   18.9   3.7   1.1    .85      .600   22.4   4.1   2.1   1.21
-.86   J Harden   LAC  .625   19.3   5.5   6.8   1.42  Phl .595   20.3   6.5   9.0   1.72
-.86   Dinwiddie  Brk  .522   13.0   3.9   5.4    .74      .567   16.8   3.7   5.8   1.05

Centers Lopez and JJackson were already at the bottom end of rebounding, for players with their size and shotblocking. They've managed to do even less of that job this season.

Shai's all-over improvement has gotten his team about 5 wins more than expected.

Code: Select all

eW+     per36       tm   Eff%   Sco   Reb   Ast   e484       Eff%   Sco   Reb   Ast   e484
2.56   Shai G-A    OKC  .635   35.2   5.7   5.3   2.95      .610   30.4   4.8   4.5   2.28
2.03   A Şengün    Hou  .581   24.0  10.4   4.4   1.99      .588   16.5  11.7   3.8   1.43
1.93   S Barnes    Tor  .564   18.9   8.3   4.4   1.45      .518   13.9   7.3   4.2    .98
1.69   F Wagner    Orl  .568   22.6   6.8   3.5   1.57      .581   19.1   4.8   3.1   1.07
1.63   DiVincenzo  NYK  .611   20.2   5.0   2.8   1.32  GSW .589   11.8   6.1   3.8    .75

1.55   A Edwards   Min  .576   28.4   5.7   4.4   1.84      .555   22.3   5.9   3.6   1.44
1.50   C Sexton    Uta  .597   24.4   3.9   4.9   1.45      .604   19.8   3.3   3.4    .92
1.50   C White     Chi  .580   18.9   4.8   4.0   1.03      .568   13.8   4.4   3.4    .69
1.50   P Banchero  Orl  .530   22.0   7.9   4.2   1.52      .517   18.4   7.8   3.2   1.16
1.47   K-A Towns   Min  .626   27.9   9.8   2.6   1.98      .607   21.8   9.0   4.2   1.59

eW+     per36       tm   Eff%   Sco   Reb   Ast   e484       Eff%   Sco   Reb   Ast   e484
1.43  Jal Williams OKC  .613   21.9   4.4   3.9   1.40      .594   15.6   5.2   3.2    .97
1.36   T Maxey     Phl  .567   23.8   3.5   4.8   1.59      .597   21.4   3.4   3.2   1.25
1.29   J Brunson   NYK  .588   28.4   4.2   5.2   1.92      .587   23.9   3.8   5.4   1.61
1.26   D Robinson  Mia  .605   17.2   3.2   3.0    .75      .530   12.1   3.8   2.0    .31
1.18   J Kuminga   GSW  .593   20.9   6.7   1.9   1.23      .588   15.6   5.9   2.5    .81

1.13   J Collins   Uta  .597   17.0  11.0   1.0   1.26  Atl .585   14.2   7.9   1.1    .89
1.11   J Walker    Por  .549   11.6  10.7   1.0    .84      .490    9.5   7.6   1.6    .37
1.09   A Caruso    Chi  .619   13.3   5.2   2.9    .81      .580    8.1   4.6   3.5    .42
.99    J Allen     Cle  .660   21.1  12.8   2.5   1.89      .658   17.3  11.7   1.6   1.58
.99   Hartenstein  NYK  .636   10.7  12.9   2.2   1.18      .560    8.2  11.9   1.8    .84

[Mike G]

Season at 76%

Code: Select all

eWin+   per36       tm   Eff%   Sco   Reb   Ast   e484       Eff%   Sco   Reb   Ast  e484
2.54   Shai G-A    OKC  .633   33.2   5.6   5.4   2.84      .610   30.4   4.8   4.5  2.28
2.17   A Sengun    Hou  .577   22.5  10.5   4.5   1.94      .588   16.5  11.7   3.8  1.43
1.98   A Edwards   Min  .575   27.8   5.7   4.5   1.87      .555   22.3   5.9   3.6  1.44
1.88  Jal Williams OKC  .618   21.8   4.4   4.1   1.45      .594   15.6   5.2   3.2   .97
1.85   DiVincenzo  NYK  .594   19.3   5.1   2.9   1.28  GSW .589   11.8   6.1   3.8   .75

1.83   J Collins   Uta  .599   16.8  11.6   1.2   1.40  Atl .585   14.2   7.9   1.1   .89
1.77   S Barnes    Tor  .558   17.8   8.5   4.7   1.37      .518   13.9   7.3   4.2   .98
1.75   D Robinson  Mia  .618   17.1   3.2   3.2    .81      .530   12.1   3.8   2.0   .31
1.64   C Sexton    Uta  .603   23.3   3.8   5.1   1.40      .604   19.8   3.3   3.4   .92
1.62   F Wagner    Orl  .569   21.3   6.9   3.6   1.47      .581   19.1   4.8   3.1  1.07

eWin+   per36       tm   Eff%   Sco   Reb   Ast   e484       Eff%   Sco   Reb   Ast  e484
1.61   J Kuminga   GSW  .587   20.1   6.9   2.3   1.28      .588   15.6   5.9   2.5   .81
1.60   Ja Walker   Por  .542   11.6  11.1   1.1    .96      .490    9.5   7.6   1.6   .37
1.54   C White     Chi  .579   18.2   4.8   4.2    .99      .568   13.8   4.4   3.4   .69
1.54   K-A Towns   Min  .618   26.3   9.7   2.8   1.95      .607   21.8   9.0   4.2  1.59
1.52   P Banchero  Orl  .539   21.8   8.0   4.5   1.48      .517   18.4   7.8   3.2  1.16

1.41   Ja Smith    Hou  .556   13.9   9.7   1.3   1.05      .508   11.9   8.7   1.2   .70
1.25   J Duren     Det  .646   15.7  14.7   2.3   1.53      .642   12.5  12.8   1.3  1.13
1.22   T Maxey     Phl  .567   22.9   3.6   4.9   1.51      .597   21.4   3.4   3.2  1.25
1.21   A Caruso    Chi  .597   12.1   5.1   3.0    .79      .580    8.1   4.6   3.5   .42
1.20  Hartenstein  NYK  .627   10.1  12.9   2.6   1.23      .560    8.2  11.9   1.8   .84

Jalen Brunson was #13 in the previous installment, and he's way gone since. How his last 6 games compare to his previous 50:

Code: Select all

G    mpg   TS%   ORb   DRb   TRb   Ast%   Stl%   TO%   Usg%   ORt  DRt
50  35.8  .597   1.8  10.2   6.0   30.2   1.3    9.1   30.3   125  117
6   37.9  .555   1.4   6.3   3.8   42.3   1.6   14.5   35.9   115  121

Carrying the Knicks, as almost their only true starter, hasn't done him very well.

Early leader Scottie Barnes has had distinct halves of the season. Showing also last year:

Code: Select all

Gms     ts%   ORb   DRb   TRb   Ast%   Stl%  Blk%   TO%   ORt  DRt   BPM
2023   .524   7.1  15.2  10.8   20.2   1.5   2.2   12.0   113  115   0.4
2024
1-30   .587   8.3  21.6  14.9   26.2   2.1   3.5   12.7   118  112   6.9
31-60  .544   6.1  16.3  11.1   26.2   1.4   3.8   14.4   110  120   0.4

https://www.basketball-reference.com/pl ... nced/2024/
As a rookie, he was mostly PF, last year mostly SF, this year equally SG/SF/PF w a few minutes at C.

[Mike G]

Corroboration from other summary stats. Last 4 columns are (mostly) improvements from last season.
Same 20, same order as in previous post:

Code: Select all

. player      tm    min   e484   PER   WS/48   BPM     e484   PER   WS/48  BPM
Shai G-A     OKC   2138   2.84   30.7   .296  10.0      .56   3.5   .070   2.7
Alp Sengun   Hou   2016   1.94   22.1   .165   4.8      .50   2.4   .050   3.4
Ant Edwards  Min   2138   1.87   19.6   .132   3.5      .43   2.2   .068   2.5
Jal Williams OKC   1820   1.45   18.7   .155   1.9      .48   3.1   .036   1.6
D DiVincenzo NYK   1638   1.28   16.2   .133   3.4  GSW .53   3.4   .027   2.7

John Collins Uta   1663   1.40   16.5   .103  -1.5  Atl .51   2.9   .008  -0.2
Sco Barnes   Tor   2094   1.37   19.4   .106   3.7      .38   3.9   .016   3.3
Dun Robinson Mia   1639    .81   12.8   .097  -0.5      .50   5.0   .062   4.5
Col Sexton   Uta   1590   1.40   20.4   .148   2.4      .47   4.5   .043   3.7
Franz Wagner Orl   1855   1.47   17.4   .121   1.2      .40   1.5   .022   1.3

. improved    tm    min   e484   PER   WS/48   BPM     e484   PER   WS/48  BPM
Jon Kuminga  GSW   1571   1.28   17.0   .113  -0.3      .47   3.7   .029   1.4
Jab Walker   Por   1260    .96   13.3   .084  -3.4      .59   3.7   .069   2.7
Coby White   Chi   2320    .99   15.3   .095   0.2      .29   2.8   .008   1.0
Karl-A Towns Min   1969   1.95   19.1   .158   2.8      .35   0.3   .025  -0.2
Pao Banchero Orl   2161   1.48   17.7   .091   1.5      .32   2.8   .044   3.0

Jabari Smith Hou   1861   1.05   13.4   .108  -0.6      .35   2.2   .071   3.1
Jalen Duren  Det   1415   1.53   18.6   .130   0.1      .40   1.3   .001   0.9
Tyrese Maxey Phl   2056   1.51   19.9   .145   3.0      .26   2.9   .017   2.4
Alex Caruso  Chi   1476    .79   14.0   .111   2.2      .37   2.5   .002   1.1
Hartenstein  NYK   1425   1.23   17.3   .183   2.2      .39   3.1   .047   2.5

BPM doesn't think John Collins and KAT are improved over last year.

[Crow]

Low assists are holding the improved overall Jabari Walker to replacement level. He is low in group on other metrics too but are his ratings replacement level on them? Not sure that can be said fully. Probably not.

Always felt BPM over penalized low assists for players who don't have it as a significant role. Is that criteria rewarded / penalized by position? Similar to with defensive rebounds or less so? With the way assists are handed out, often for nothing special passes, often for really big credit, somebody is going to get credit about 70% of time for often nothing special pass that somebody is probably going to do. And all the possible assists, routine or strong, get zero, even in a background way. Because THAT potential assist pass to a guy who actually got the shot and made it becomes special in the eyes of the metric as prescribed. Hypothetical ly, assists as a ratio of potential assists or total passes of a player or the lineup or team could be considered is crediting (and debiting).

Is by position crediting going to be more or less pronounced in new BPM? Is new BPM coming for next season? Is it agreed that it is coming to BRef, replacing previous version?


Banchero improves from previously average. Because of a few things but probably heavily for big move up on assist-making. 18th biggest improvement here on e484, 5th biggest on BPM. About 12th on PER. 9th on WS48.


A Edwards at +3.5. Nice but not stratospheric. SGA's is stratospheric based mainly on prior advancement.

Big move for D Robinson, but not quite up to league average. Heat hold on to him but try to take advantage of improvement for trade? It could go either way. It may or may not matter much.

Kuminga, modest improvement mainly based on usage increase. Slight decline in modest assist rate probably holding his increase down. 9th biggest increase on e484, just 14th on BPM. 11th biggest on WS48. PER loves and overrated usage, absolutely and compared to other metrics, so tied for 3rd biggest increase there.


With player tracking, contested defensive rebounds could get more credit than for uncontested rebounds. I don't know of a metric doing that, but they could. All are probably still diminishing credit for all defensive rebounds based on most but not all being uncontested something that somebody is going to do. Or at least that is the logic sold and accepted
conventionally for this one and only specific stat.


Variation in sensitivities to stat change among metrics is a good argument for metric comparison, as presented here, or metric blending, as sometimes done here and elsewhere, or metric review and / or replacement.

[Mike G]

... contested defensive rebounds could get more credit...

If a player or players are properly blocking out the opponent, a rebound may be viewed as "uncontested". However, I don't see why failure to do part 1 of the job should make part 2 seem more valuable.
The logic of claiming something like "Yes you got 10 rebounds, but half of those could have gone to someone else on our team" is in the same league as: "I got 10 rebounds, but I could have gotten 15 if I wanted". So 10 = 10 in every real sense.

Similar pseudo-arguments are made that "someone else" could have made the assist or the shot; and yet some are better than others, and coaches and (most) teammates will prefer the more competent player does each job as opportunity allows.

PER doesn't so much over-value "usage" as it entirely disregards defense. A huge GameScore in a 150-140 game -- or high PER for a team that allows 125 ppg -- doesn't have the real impact as the same number in a more defensive game or season.
I may give PER equal weight, though, when it comes to awards and popular perception, and to keep it simple.

When you make good passes and maintain position for rebounding, you tend to get the numbers you deserve.
If you're not sensible with shot selection and passing and playing D, you probably don't get the minutes a more diligent player gets with similar 'skills'.

Check these 4 stats against player min/G. Just players who last year got 500+ minutes, dividing teams into equal thirds based on their wins last year. Here are the correlations:

Code: Select all

wins '23   e484  PER  WS/48  BPM
45 - 58    .56   .53   .40   .63
40 - 44    .55   .55   .31   .56
17 - 38    .55   .51   .27   .54

[Crow]

Those correlations of metrics to minutes are only moderate and only more than mildly better for good teams on half the metrics. Would want to compare to the conceptually comprehensive RAPM. Even then, I think it is quite possible to question / critque the degree of optimization Coaches are achieving with their minute (and lineup) choices. All metric scores should be adjusted based on quality of teammates on court and opponents but only RAPM does that comprehensively. Ewins sonewhat does a game level version of half that but not the full opponent quality question for season or by who is on court play by play for each person and lineup in and against.


PER over-values "usage" and disregards shot defense including help defense. And under-values actual rebounding and ignores box out data. (Tracking data for that and other things is available for any metric builder that wants to do the work to include / not ignore it.)

[Mike G]

Correlations with minutes from last year's 1st round of playoffs:

Code: Select all

avg    tm     W   L    PER   WS/48   BPM   e484
.55    Phl    4   0    .42    .39    .74    .64
.41    Brk    0   4    .39    .29    .39    .56
.36    NYK    4   1    .40    .19    .25    .59
.63    Cle    1   4    .76    .36    .61    .80
.21    Mia    4   1    .27    .07    .18    .34
.43    Mil    1   4    .55    .38    .26    .54
.06    Bos    4   1    .24   -.50   -.22    .73
.23    Atl    1   4    .29    .11    .07    .43

avg    tm     W   L    PER   WS/48   BPM   e484
.63    Phx    4   1    .68    .54    .64    .69
.30    LAC    1   4    .34    .11    .40    .34
.62    Den    4   1    .64    .54    .56    .75
.69    Min    1   4    .73    .64    .58    .81
.78    LAL    4   2    .82    .69    .74    .88
.33    Mem    2   4    .45    .08    .12    .69
.29    GSW    4   3    .29    .10    .19    .57
.43    Sac    3   4    .57   -.07    .58    .63

Includes those who got at least 5 min per team game. Exception: Kawhi (LAC), best player but 9th in minutes due to early injury; excluded from sample.

And later rounds;

Code: Select all

avg     tm    W   L    PER   WS/48   BPM   e484
.57    Mia    4   2    .60    .42    .57    .71
.64    NYK    2   4    .63    .60    .60    .73
.60    Bos    4   3    .73    .35    .52    .80
.62    Phl    3   4    .60    .52    .72    .63
.66    Den    4   2    .68    .62    .67    .68
.81    Phx    2   4    .84    .71    .80    .89
.67    LAL    4   2    .69    .59    .58    .80
.10    GSW    2   4    .03   -.15    .17    .35

avg     tm    W   L    PER   WS/48   BPM   e484
.33    Mia    4   3    .37    .17    .25    .54
.26    Bos    3   4    .23   -.04    .22    .63
.89    Den    4   0    .92    .88    .92    .84
.81    LAL    0   4    .85    .77    .79    .85

.27    Den    4   1    .33    .02    .15    .59
.23    Mia    1   4    .22   -.14    .21    .64

[Crow]

So Celtics correlation of minutes to performance in 1st round was the absolute lowest by BPM but 5th highest by e484?

Not sure what to make of that, other than maybe don't go by one metric.

Metrics are however fairly in agreement than Heat correlation of minutes to performance was quite poor / limited and that 4 teams in the west were pretty highly correlated relatively speaking.

In later rounds, there probably was higher correlation and the Nuggets were very highly optimized in 3rd round. Warriors were ridiculously low in second round. In the finals, the correlations were generally pretty bad for both.

[DSMok1]

Correlations between BPM and playing time are mostly irrelevant at the coaching level. Fit matters far more than BPM. BPM is generic across all contexts the player finds themselves in. If your five highest BPM players are all big, your best lineup is not to play all bigs. Even if the five highest BPM's come across five different positions, your best lineup is not necessarily to play those five together. Fit and skill set optimization matters more.

Note that those things have to be assessed by the coach and by viewing how the lineup works together; seeing things like the quality of shots that are obtained.

[Mike G]

Correlations between BPM and playing time are mostly irrelevant at the coaching level. Fit matters far more than BPM. ...If your five highest BPM players are all big, your best lineup is not to play all bigs. ...

These points are likely true across all these summary stats. Hence, a team may have high or low correlations across all (or most) of these 4 measures.

As I am presently refining these playoff correlations, it's apparent I should be using player min/G rather than total minutes. Several key players missed some games -- Kawhi, Giannis, Paul ... -- and thus are out of order in the minutes allocation.
This postseason I will try to keep mpg in my summaries.

It's both vital and kind of arbitrary whether to count the top 7 or top 11 players in a correlation study. So I am now generating lines for top-5 thru top-11; deleting those with fewer than 5 mpg; and averaging the lines.
In this way, the bottom line -- generally a player going 5-10 mpg -- gets counted one time, the next one up 2 times, etc. Starting 5's get counted the most.

I was going to just edit my previous post, but I will re-submit here.
Round 1 of 2023 playoffs:

Code: Select all

avg     tm    W   L    PER   WS/48   BPM   e484
.46    Phl    4   0    .41    .36    .57    .52
.35    Brk    0   4    .42    .09    .29    .60
.33    NYK    4   1    .32    .21    .33    .46
.49    Cle    1   4    .67    .12    .51    .68
.40    Mia    4   1    .49    .27    .35    .50
.25    Mil    1   4    .37    .01    .19    .46
.16    Bos    4   1    .35   -.37   -.02    .68
.12    Atl    1   4    .22   -.21    .05    .43
                            
avg     tm    W   L    PER   WS/48   BPM   e484
.49    Phx    4   1    .55    .32    .45    .64
.25    LAC    1   4    .38   -.12    .30    .44
.69    Den    4   1    .70    .63    .67    .78
.62    Min    1   4    .67    .58    .48    .75
.56    LAL    4   2    .63    .32    .48    .80
.27    Mem    2   4    .34    .09    .10    .52
.17    GSW    4   3    .27   -.10    .05    .46
.50    Sac    3   4    .67   -.01    .59    .74

These exclude Giannis, Kawhi, and Herro.
Rounds 2, 3, 4, minus Chris Paul:

Code: Select all

avg     tm    W   L    PER   WS/48   BPM   e484
.49    Mia    4   2    .59    .43    .33    .62
.70    NYK    2   4    .72    .59    .70    .79
.58    Bos    4   3    .67    .38    .50    .76
.53    Phl    3   4    .58    .42    .52    .60
.51    Den    4   2    .54    .41    .50    .57
.82    Phx    2   4    .88    .58    .87    .94
.59    LAL    4   2    .63    .46    .52    .77
.24    GSW    2   4    .20    .04    .30    .41
                            
avg     tm    W   L    PER   WS/48   BPM   e484
.46    Mia    4   3    .51    .32    .40    .62
.09    Bos    3   4    .18   -.34   -.11    .65
.79    Den    4   0    .81    .70    .83    .84
.87    LAL    0   4    .89    .85    .87    .87
                            
.53    Den    4   1    .55    .44    .48    .66
.38    Mia    1   4    .48    .09    .24    .71

Total/avg by round:

Code: Select all

rd   PER   WS/48   BPM   e484
1    .47    .14    .34    .59
2    .60    .42    .53    .68
3+4  .57    .35    .45    .72

[Mike G]

In each of their 3 playoff rounds, the Celtics had the widest range of correlations in that round; which is to say, the 4 summary stats don't agree so much.
This may be very hard to interpret, but I'm copying my worksheet with all players and correlations available. Toward the bottom (fewer minutes) they get increasingly less viable.
Totals for entire postseason:

Code: Select all

. Celtics   G  Min   PER  WS/48  BPM  e484    PER  WS48   BPM  e484   PER   WS  BPM  e484
Tatum      20  799  22.6  .168   5.3  2.17    .57  -.08   .27  .80          
Brown      20  751  15.5  .051  -0.9  1.21                    
Smart      20  679  15.4  .112   0.8  1.01  Correlations from Tatum    Average of correl.
Horford    20  617  12.8  .118   2.8   .78   to player in each line    down to each line
White      20  593  16.2  .151   3.0   .83    .72  -.11   .05  .88    .72  -.11  .05  .88
Brogdon    19  474  13.4  .076  -1.6   .76    .72   .24   .47  .80    .72   .06  .26  .84
R Williams 20  417  19.6  .208   3.4  1.04    .25  -.29   .17  .66    .56  -.06  .23  .78
G Williams 15  265  12.0  .137   1.1   .20    .47  -.26   .20  .79    .54  -.11  .22  .78
Hauser     15  104   9.3  .079  -2.2   .02    .66   .04   .48  .85    .57  -.08  .27  .80

Pritchard  10   57  18.4  .140   5.0  1.20    .39  -.03   .11  .60    .54  -.07  .25  .76
Kornet      8   32  28.5  .337   8.2  1.18   -.11  -.41  -.22  .46    .44  -.12  .18  .72
Muscala     6   21  17.9  .210   4.9   .52   -.13  -.46  -.29  .50    .37  -.16  .12  .69
.   total     4809  16.4  .125   1.9  1.10

Celts went mostly 8 deep, with Grant W slipping out of the rotation for a spell.
The bottom 3 are included for illustration sake -- apparently they were awesome in mop-up duties -- and Hauser is on the cusp of relevancy. Below 5 mpg the correlations plummet.

Team totals are pretty consistent across the stats. But was Brown the #2 Celt? or the very worst?

[Crow]

If "Fit matters far more than BPM..." then Coaching lineup decisions should be really driven by lineup data and the desire to play the best fit lineups should be very high. But they aren't. Coaches throw out hundreds of lineups based considerably on in the moment guesses and not on apparent best fit to performance data and trying to optimize it.


Projecting a team's future performance on sum of BPM for 5 biggest minute players will also be a very crude estimate, dependent on repeating past average lineup pattern.


If minutes and BPM and / or other metrics are not relatively highly correlated, a team's roster construction balance is suspect and trades for better fit players to form high use / higher performance lineups should be pursued.