Wednesday, November 24, 2010

Offensive and Defensive Rebounds

Some statisticians treat offensive and defensive rebounds as equally valuable when evaluating NBA players. I don't see the logic. If I'm a player and I fail to get a defensive rebound, odds are (something like 70% chance) that someone else on my team will rebound the ball so the value of the rebound was 0.3 possessions. If I get an offensive rebound, then I created a possession that would have only been created about 30% of the time, so it's worth about 0.7 possessions. Theory says offensive rebounds are a lot more important. (One caveat: the kind of player who is available to get a lot of offensive rebounds may set few picks, create few of his own shots, and take few shots and thus draw fewer defenders. So offensive rebounds might be correlated with other bad, but unmeasurable attributes.)

Defense of the equivalence base it on regressions. The theory is, I guess, that the regressions are showing an implicit correlation between offensive rebounding and negative attributes (discussed above). Another theory, which is my guess (without looking closely at the data) is that defensive rebounds are overvalued because they are correlated with good defense (low opponent FG%), which is also unmeasured. If that is the case defensive rebounds are overvalued and that's why the equivalence shows up in the regressions.

No comments:

Post a Comment