APBRmetrics

http://stats-for-the-nba.appspot.com/rebounding_study
New ranking is here
http://stats-for-the-nba.appspot.com/rebounding (same URL as the old one, but a different ranking method)

Things to take away: Adjusted rebounding is better at forecasting than using individual rebounding numbers. A mix of both methods does best overall. Most overrated and underrated players are listed in the writeup

Appreciate comments and suggestions.

bump

So a meta-metric of individual rebounds with team rebounds distributed among the players and Adjusted rebounding numbers did best on the Briers test.

I like seeing both the test of the blend and the strong performance. Evan has recently been blending EZPM and RAPM as well.

In both cases, there is the opportunity to explore beyond equal / even blends of the two to see if a different ratio does even better.


On how to allocate team rebounds, might distributing them at exactly in line with the actual rebounding %s players get on their team be a good option? That would introduce no change in rank or scale of player performance on teams compared to where they are with respect to actual rebounding %s.

Using average player performance is another option. It could shift results some among players within a team. But it could be argued that it might be better on average than distributing them at exactly in line with the actual rebounding %s players get on their teams if one was trying to capture perhaps "truer" 'player rebounding impact in a position', not dependent on a specific team context.

The equal distribution of team rebounds among all positions avoids heightening the position impact of rebounding. There may be arguments in favor of that and against given what actual rebounding %s show by position.

I see your comments on the diminishing returns issue. I hope others who have researched that issue will read your comments and respond in general and on that.

Thanks for the data and the extended, thorough explanation.

I am just beginning to look at the player level data. if I have other observations from that, I'll post them later.

bump

bump

bump