Explaining minutes & money (and values)

Home for all your discussion of basketball statistical analysis.
Post Reply
Crow
Posts: 5491
Joined: Thu Apr 14, 2011 11:10 pm

Explaining minutes & money (and values)

Post by Crow » Thu May 24, 2018 1:20 pm

There are various forms of boxscore metrics, adjusted (and unadjusted) plus minus, player tracking and blends. All seek to solve, all seek reasoned facts to help solve.

But there is also coaching allocation of minutes and team allocation of money. These may or may not line up well with the first group. Help me remember, is there an academic study of what the weights are for discrete stats are in coaches allocation of minutes on average? I can't recall if that has been done. It should be done and common knowledge. It could be done for each and every coach and every season and maybe every month and game. But, unlike the first question, I KNOW it hasn't been done, at least in public. Anybody up to it?

Same with explaining the allocation of money. We know points get paid more but what are the accurate weights on average (today & trend) and for specific teams, seasons, etc.

Which coaches & team are closest to league averages for allocation of minutes and money? Farthest from? Most in / out of agreement between coaching and spending?

It would also be possible to compare these weightings to weightings in various performance metrics and to RPM (and O & D splits). We don't know what coaches and teams use and what balance (and they may not either consciously) but we can see how it compares, appears. Who is most aligned with PIPM, BPM, WS/48, Wins Produced, PER or yay points? Offense or total impact?

Coaches and GMs may have different ideal weights than the apparent effective weights and they might consider more tha one season and could anticipate future trends. GM spending decisions are made at specific times that could become dated as performance levels change and individual players & decisions impact one another including the choice of coach and play style.

Mike G
Posts: 4140
Joined: Fri Apr 15, 2011 12:02 am
Location: Asheville, NC

Re: Explaining minutes and money

Post by Mike G » Thu May 24, 2018 2:25 pm

With few exceptions I find WS/48 correlates worst with minutes. eWins is most often the best correlation.
Partly this is due to starters getting more credit than bench guys, for their production. BPM also rewards minutes and comes close to eWins. PER is generally better than WS.

It's important to have some sort of cutoff -- at least 10 mpg, or only the top 9 players, for example. One extreme outlier with few minutes can greatly skew the correlations.

A good blend of metrics would probably try to achieve higher correlation with MPG. And weigh toward Win% of the teams.
The novice may think his metric is better than NBA coaching, but in general that is wrong.

Crow
Posts: 5491
Joined: Thu Apr 14, 2011 11:10 pm

Re: Explaining minutes and money

Post by Crow » Thu May 24, 2018 3:07 pm

Let's see how they compare in far more detail and in different ways before we decide either way or anything.

How do the titlist minutes & money align with the performance metrics? More tightly? More uniquely?

Crow
Posts: 5491
Joined: Thu Apr 14, 2011 11:10 pm

Re: Explaining minutes & money (and values)

Post by Crow » Fri May 25, 2018 7:48 pm

Leaguewide the correlation of minutes to RPM (for player over 500 minutes) this regular season is .462. Single year PIPM . 467 BPM is slightly higher at .480, for the win here. PER is actually lower at .433. WS/48 and DRE are significantly lower. None of these values are particularly impressive. Multi-year PIPM didn't improve the correlation over single year and was at.459. The defensive split correlations where available were really bad, near zero or even below. If both these metrics and coaches know defense, the coaches weren't making decisions based on these defensive metrics and probably not defense in any other way either.

At team level, the correlation of minutes to RPM ranged from near .800 (BKN, BOS, MIL, OKC, SAS with BKN highest at .795) to three negative (DAL, PHX and SAC with PHX at a tankeriffic and / or out of touch -.654. Dallas was also absurdly negative at -.515). On single year PIPM, the most consistent with minute allocation were BOS CHA, MIL, OKC, PHI and SAS (all over.750). The lowest were ATL, DAL, NYK, PHX and SAC with Dal and PHX again leading the way in negative correlation between minutes and performance. Milwaukee was the only team close to .800 correlation between minutes and performance on BPM. DAL and PHX again with biggest negatives but PHX reduced theirs and DAL leads here. OKC, with their points oriented big 3, had the by far highest correlation with PER and minutes at .870. CHA was the only other much over .700. PER is not a driver. The lowest were DAL and NY. The highest with single year O-PIPM was OKC at a very high .950. There were still a few negatives but they were small. With single year D-PIPM the leader was MEM at .898. Yep, they really believe in defense. DAL was at a ridiculous -.696.

With respect to salary, I'd have to add some salary data to make more comparisons with other metrics. (Maybe later.) But using PIPM Value, the overall correlation of money and minutes with that metric is . 643. Coaches / teams are playing the guys who got paid more often than those who are performing well. No real surprise there. The tightest alignment of money and minutes (all over .800) are BOS, CHA, LAC, MIL, OKC (at .910), PHI, TOR and UTA. DAL and PHX again have the biggest negative correlations with PHX at -.405.

It is easy to spot the tanking teams. They probably can't be this whack otherwise. Of the four conference finalists, only BOS is near the top in performance value per dollar. Cavs are in bottom 10. Rockets modestly above average. Warriors barely top 10.

Nathan
Posts: 135
Joined: Sat Jun 22, 2013 4:30 pm

Re: Explaining minutes & money (and values)

Post by Nathan » Fri May 25, 2018 8:26 pm

How do the Cavs look? Lots of shock/dismay among fans at decisions to play Thompson (and Green) over Nance, Smith over Korver, Clarkson over...well...anybody.

EDIT: Just saw that you mentioned they're bottom-10 at the end. I imagine they're even worse in the playoffs.

Crow
Posts: 5491
Joined: Thu Apr 14, 2011 11:10 pm

Re: Explaining minutes & money (and values)

Post by Crow » Fri May 25, 2018 9:17 pm

Cavs (former employers of APM, RAPM producers) were 8th lowest on correlation of minutes and RPM. Way lower on single year PIPM. Pretty high correlation with BPM. Very bad correlations on most defensive metrics. Meh on DBPM but DBPM is meh or worse. Looks like a case where the boxscore is the driver but that isn't enough and not even that consistent a guide.

Mike G
Posts: 4140
Joined: Fri Apr 15, 2011 12:02 am
Location: Asheville, NC

Re: Explaining minutes & money (and values)

Post by Mike G » Sat May 26, 2018 12:11 am

Here's how the Cavs look in these ECF thru 5 games, in eWins/480, PER, WS/48, and BPM
Below that list (inverted order) are correlations with minutes for their top-5, top-6, thru top-12 in mpg.

Code: Select all

Cavs     e480   PER   WS/48   BPM   mpg
Calderon  .06    0.1  -.043  -11.1    3
Osman     .03    8.9   .022   -2.2    4
Hood      .79    6.7   .013   -9.4    7
Clarkson  .38    7.0  -.035   -8.2   11
Nance    1.51   23.6   .289    9.8   13
Korver   1.06   18.2   .213    3.4   22
Green    -.22    4.9  -.002   -3.9   22
Thompson  .95   18.8   .185     .3   27
Smith    -.45    1.7  -.038   -5.9   28
Love     1.54   13.0   .045   -2.4   31
Hill      .16   10.3   .102     .4   32
James    2.70   29.8   .230   12.5   39

top  to  e480   PER  WS/48  BPM   avg
5    TT   .76   .72   .56   .89   .73
6    JG   .79   .76   .66   .83   .76
7    KK   .62   .52   .27   .59   .50
8    LN   .25   .07  -.24  -.01   .01
9    JC   .30   .22   .08   .27   .22
10   RH   .26   .32   .21   .45   .31
11   CE   .35   .36   .28   .42   .35
12   JC   .41   .47   .38   .53   .45
all avg   .47   .43   .27   .50   .42
Correlations are not bad thru 7 players; but #8 is Nance who is their #1 in WS/48, strong #2 in PER and BPM.
The bottom row is just the average of the 5-12 lines; Calderon is counted just once, Osman twice, etc.

Nathan
Posts: 135
Joined: Sat Jun 22, 2013 4:30 pm

Re: Explaining minutes & money (and values)

Post by Nathan » Sat May 26, 2018 12:23 am

Mike G wrote:Here's how the Cavs look in these ECF thru 5 games, in eWins/480, PER, WS/48, and BPM
Below that list (inverted order) are correlations with minutes for their top-5, top-6, thru top-12 in mpg.

Code: Select all

Cavs     e480   PER   WS/48   BPM   mpg
Calderon  .06    0.1  -.043  -11.1    3
Osman     .03    8.9   .022   -2.2    4
Hood      .79    6.7   .013   -9.4    7
Clarkson  .38    7.0  -.035   -8.2   11
Nance    1.51   23.6   .289    9.8   13
Korver   1.06   18.2   .213    3.4   22
Green    -.22    4.9  -.002   -3.9   22
Thompson  .95   18.8   .185     .3   27
Smith    -.45    1.7  -.038   -5.9   28
Love     1.54   13.0   .045   -2.4   31
Hill      .16   10.3   .102     .4   32
James    2.70   29.8   .230   12.5   39

top  to  e480   PER  WS/48  BPM   avg
5    TT   .76   .72   .56   .89   .73
6    JG   .79   .76   .66   .83   .76
7    KK   .62   .52   .27   .59   .50
8    LN   .25   .07  -.24  -.01   .01
9    JC   .30   .22   .08   .27   .22
10   RH   .26   .32   .21   .45   .31
11   CE   .35   .36   .28   .42   .35
12   JC   .41   .47   .38   .53   .45
all avg   .47   .43   .27   .50   .42
Correlations are not bad thru 7 players; but #8 is Nance who is their #1 in WS/48, strong #2 in PER and BPM.
The bottom row is just the average of the 5-12 lines; Calderon is counted just once, Osman twice, etc.
Seems like the correlations are propped up by LeBron's datapoint? I imagine 2-8 would be seriously negative, and then probably back to neutral or so once you include the deep bench.

Clarkson's not playing as many minutes as it feels like he's playing, I guess. Though a basketball-reference search reveals he has the worst BPM of any player in playoff history with at least 200 minutes. Possibly unprecedented to have such a bad player getting non-garbage minutes this deep into the playoffs.

Thanks for taking the time to put this together!

Mike G
Posts: 4140
Joined: Fri Apr 15, 2011 12:02 am
Location: Asheville, NC

Re: Explaining minutes & money (and values)

Post by Mike G » Sat May 26, 2018 1:22 pm

Clarkson is also backing up perhaps the weakest player (statistically) ever to start and play so many minutes.
Since 1984, most minutes in the Conf. Finals by players with fewer than 40 pts, 20 reb, and 10 ast:

Code: Select all

<40-20-10  yr   Min  Pts  Reb  As  St  Bk
JR Smith   18   179   31   15   6   1   1
B Bowen    08   172   36    5   5   6   1
B Bowen    05   171   35   16   8   1   1
M Pietrus  12   159   21   19   0   5   2
D Fisher   07   152   31    5   7   7   0

D Majerle  97   150   26   16   7   8   0
D McKey    98   150   38   14   7   4   0
D Scott    96   141   29   10   6   5   0
J Anthony  11   139    8   19   4   5  15
B Bowen    07   139   30   15   3   5   0
http://bkref.com/tiny/MWYU4
Other than Bowen (one time), they all got some steals or blocks, at least. JR seems to do nothing, and he has the worst shooting% of all these.
Clarkson has more points in less than half the minutes, in the ECF.
Here are some stats for both in the ECF and the whole postseason:

Code: Select all

ECF   TS%  OR%   DR%  As%  St%  Bk%  TO%  ORt  DRt
JR   .331   .0   9.6  4.8   .3   .6  7.9   69  113
JC   .398  8.6  11.3  0.0   .8  1.4  2.4   85  109

all   TS%  OR%   DR%  As%  St%  Bk%  TO%  ORt  DRt
JR   .492  1.4   8.4  5.5  1.5   .5  9.5   98  112
JC   .380  3.8   9.6  7.6  1.2   .7  9.0   80  112
Clarkson has picked up a wee bit? Smith had a good series vs Toronto.

Updated summary stats:

Code: Select all

ECF  mpg  PER  WS/48   BPM  e480
JR   30  -0.9  -.066  -7.3  -.45
JC   11   7.8  -.040  -9.9   .43
These are the 2 worst WS rate and BPM on the team. Jeff Green also has worse PER and eW than Clarkson.
Negative PER over 179 minutes might also be an alltime worst of some sort.

Nathan
Posts: 135
Joined: Sat Jun 22, 2013 4:30 pm

Re: Explaining minutes & money (and values)

Post by Nathan » Sat May 26, 2018 2:38 pm

Mike G wrote:Clarkson is also backing up perhaps the weakest player (statistically) ever to start and play so many minutes.
Since 1984, most minutes in the Conf. Finals by players with fewer than 40 pts, 20 reb, and 10 ast:

Code: Select all

<40-20-10  yr   Min  Pts  Reb  As  St  Bk
JR Smith   18   179   31   15   6   1   1
B Bowen    08   172   36    5   5   6   1
B Bowen    05   171   35   16   8   1   1
M Pietrus  12   159   21   19   0   5   2
D Fisher   07   152   31    5   7   7   0

D Majerle  97   150   26   16   7   8   0
D McKey    98   150   38   14   7   4   0
D Scott    96   141   29   10   6   5   0
J Anthony  11   139    8   19   4   5  15
B Bowen    07   139   30   15   3   5   0
http://bkref.com/tiny/MWYU4
Other than Bowen (one time), they all got some steals or blocks, at least. JR seems to do nothing, and he has the worst shooting% of all these.
Clarkson has more points in less than half the minutes, in the ECF.
Here are some stats for both in the ECF and the whole postseason:

Code: Select all

ECF   TS%  OR%   DR%  As%  St%  Bk%  TO%  ORt  DRt
JR   .331   .0   9.6  4.8   .3   .6  7.9   69  113
JC   .398  8.6  11.3  0.0   .8  1.4  2.4   85  109

all   TS%  OR%   DR%  As%  St%  Bk%  TO%  ORt  DRt
JR   .492  1.4   8.4  5.5  1.5   .5  9.5   98  112
JC   .380  3.8   9.6  7.6  1.2   .7  9.0   80  112
Clarkson has picked up a wee bit? Smith had a good series vs Toronto.

Updated summary stats:

Code: Select all

ECF  mpg  PER  WS/48   BPM  e480
JR   30  -0.9  -.066  -7.3  -.45
JC   11   7.8  -.040  -9.9   .43
These are the 2 worst WS rate and BPM on the team. Jeff Green also has worse PER and eW than Clarkson.
Negative PER over 179 minutes might also be an alltime worst of some sort.
Hmmm...hard to compare between high-volume and low-volume terrible players. Clarkson really stands out because he's 2nd on the team in USG (behind LeBron of course) but with sub-40% true shooting and 0 assists in the conference finals. JR has been even worse shooting the ball, but it's less conspicuous because he takes fewer shots and actually passes the ball every once in a while. Fun video to illustrate:

https://twitter.com/CaineLovesCali/stat ... 7799826432

Mike G
Posts: 4140
Joined: Fri Apr 15, 2011 12:02 am
Location: Asheville, NC

Re: Explaining minutes & money (and values)

Post by Mike G » Sat May 26, 2018 10:42 pm

Crow wrote:Leaguewide the correlation of minutes to RPM (for player over 500 minutes) this regular season is .462. Single year PIPM . 467 BPM is slightly higher at .480, for the win here. PER is actually lower at .433. WS/48 and DRE are significantly lower. None of these values are particularly impressive...
Yah, eWins rate correlates at .466 with min>500
But with min/G it's .541

Playoff teams are .491 with minutes and .569 with min/G
Non-playoff teams were .422 and .529

Players traded mid-season were counted with their final team, but with all their minutes, and min/G for the whole year.

Mike G
Posts: 4140
Joined: Fri Apr 15, 2011 12:02 am
Location: Asheville, NC

Re: Explaining minutes & money (and values)

Post by Mike G » Sat May 26, 2018 11:03 pm

For last season, I have RPM along with the others. For players with 200+ minutes, I get these correlations with total minutes and with Min/G

Code: Select all

corr  e484  PER  WS/48  BPM   RPM
min. .538  .504  .417  .578  .562
mpg  .606  .555  .420  .615  .583
Well everyone does better with MPG than with straight minutes, some more than others.

Separating players into their teams, looking at correlations of top 5 in mpg, top 6, thru top 11, we get these average correlations across 30 teams:

Code: Select all

#   e484  PER  WS/48  BPM   RPM    avg
5   .57   .57   .35   .43   .40   .462
6   .60   .60   .46   .53   .52   .540
7   .56   .55   .43   .52   .52   .515
8   .58   .56   .45   .55   .55   .539
9   .61   .58   .43   .57   .56   .550
10  .60   .56   .40   .56   .55   .535
11  .58   .52   .38   .58   .56   .524
It seems best correlations are with 1 thru 9. Player #9 is on avg an 18 mpg guy.
With shorter rotations, such as playoffs, 7 or 8 might be stronger.

eminence
Posts: 85
Joined: Sun Sep 10, 2017 8:20 pm

Re: Explaining minutes & money (and values)

Post by eminence » Sun May 27, 2018 1:58 pm

Did a super basic analysis on salary vs win%/Orating/Drating once that's kind of close to what you're talking about. Can't find it right now, but in short teams don't pay successfully for defense at all (~0 correlation between the two), while they're decent at buying offense.

Crow
Posts: 5491
Joined: Thu Apr 14, 2011 11:10 pm

Re: Explaining minutes & money (and values)

Post by Crow » Sun May 27, 2018 7:30 pm

The weakness of RPM and BPM with the top 5 is concerning, worth looking at further.

Also low on RPM / minutes correlation (outside the tankers or just awful), Utah. Very low correlation with PIPM too and even lower with BPM. Does Snyder freelance too much?

With respect to the slight BPM advantage over some other metrics, are coaches valuing "versatility" and like BPM are they doing it appropriately or too much? May be affected more by randomness than reason.

Of the 4 finalists, Celtics are highly aligned between minutes and RPM, the rest below average. Celtics minutes highly aligned with PIPM too, GSW & HOU modest alignment, Cavs very low. Cavs highest alignment with BPM but not especially high; GSW lowest. Overall GSW with the least alignment between minutes and these metrics; Boston highest.

Post Reply