APBRmetrics

I was perusing 82games and is Durant really only worth +1.1 points to his team over 100 possessions?

On Court / Off Court stats

On Court: +8.4
OFF Court: +7.2
Net: +1.1

http://www.82games.com/1314/13OKC8.HTM

Any feedback would be appreciated. Thank you.

By this version of adjusted +/- (which accounts for the quality of all opponents and teammates on the court with him), second article down here: http://talkingpracticeblog.com/ Durant is estimated to be worth +4.2 points per 100 possessions. By another measure IPV, as explained in first article, he is worth +6.7, the most in the league.

His raw on /off data is affected by

1) playing a lot against opponent starters, usually better than opponent subs on average (maybe by about 3 pts on average for starting lineup vs the lineups with the most subs used),
2) playing a lot with Perkins and Thabo (both modestly negative on RAPM) and in particular in the traditional with Westbrook starting lineup with them which is moderately negative, and
3) the team when he is off the court plays 80% of the time with Collison (very good on RAPM, 2nd best in league) while he plays with Collison only about 25% of the time.

Raw +/- can be subject to uneven quality of on and off environments (i.e. quality of other 9 players) and this is one such case that tends to mask Durant's individual impact.

Reader wrote:I was perusing 82games and is Durant really only worth +1.1 points to his team over 100 possessions?

No.

Crow wrote:By this version of adjusted +/- (which accounts for the quality of all opponents and teammates on the court with him), second article down here: http://talkingpracticeblog.com/ Durant is estimated to be worth +4.2 points per 100 possessions. By another measure IPV, as explained in first article, he is worth +6.7, the most in the league.

His raw on /off data is affected by

1) playing a lot against opponent starters, usually better than opponent subs on average (maybe by about 3 pts on average for starting lineup vs the lineups with the most subs used),
2) playing a lot with Perkins and Thabo (both modestly negative on RAPM) and in particular in the traditional with Westbrook starting lineup with them which is moderately negative, and
3) the team when he is off the court plays 80% of the time with Collison (very good on RAPM, 2nd best in league) while he plays with Collison only about 25% of the time.

Raw +/- can be subject to uneven quality of on and off environments (i.e. quality of other 9 players) and this is one such case that tends to mask Durant's individual impact.

This is a great explanation. Thank you Crow.

After reading your post I have been looking for RAPM #'s but I guess they are hard to come by in-season before the All-Star Game. Is there anywhere I can find #'s that approximate a player's value to his team? If you have any ideas, please let me know and thanks again.

The only in-season public RAPM I know about is at the link I posted above http://talkingpracticeblog.com/.

My data is a couple games old and estimates possessions, but with my new "indirect plus-minus" (an attempt to made an adjusted plus-minus that is more understandable), I have Durant's +/- as 8.2 points per 100 higher than we would expect given his teammates' statistical +/-.

ON per 100: +7.2
SOS modifier (based on team SOS and % minutes played): +0.2
Teammate estimated SPR (my statistical +/-): +1.83
Team Adjustment: +2.68 (entire team gets this)

ON of 7.2, plus 0.2 for SOS, and 2.68 for team adjustment, minus teammate SPR (i.e. how much his +/- should be if his value were 0 per 100) = 8.2.

This makes him the 2nd-highest on the team, after Nick Collison. The randomness that still persists could largely be chalked up to LOTS of minutes with Kendrick Perkins, etc...also playing the vast majority of minutes means his +/- will be very very close to his team's overall +/-.

Is Teammate estimated SPR, the estimated minutes-weighted average for his actual mix of teammates on the court with him or just the straight average teammate value for all their minutes played? And that is the sum for the other 4 positions? If so, are you saying then that the average SPM of the rest of the team is barely positive? What is the derivation and interpretation of Team Adjustment? (is the coach any part of that?) I am concerned that there may be a considerable amount or overlap or double-counting in this formula, though I understand you are going for a rough estimate by simpler means. Anything more you care to say or point at to explain further would be welcome.

Did your offensive SPM method change significantly from this much earlier post?
http://www.apbr.org/metrics/viewtopic.p ... 212#p11212
What did you use for defensive side?