It's an NBA stats throwdown, as David Berri critiques John Hollinger's PER formula, backed up by Malcolm Gladwell. Hollinger fires back saying Berri's misunderstood how PER is calculated. On the issue at hand -- the "break even point" for shooting efficiency -- the two formulas are actually very close, Hollinger is only slightly more forgiving of missed shots. This long post by Dan Rosembaum on Wins Produced gets into some interesting business; Rosenbaum's point seven, dealing with defense, seems especially persuasive.
Obviously, as problematic as trying to quantify individual contributions to team offense based on box score numbers may be, quantifying individual contributions to defense is way more problematic. One point Berri makes time and again on his blog is that payroll size and wins correlate only very weakly in the NBA even though individual players' statistical production is fairly consistent from year to year. That, he points out, is good evidence that conventional thinking about player evaluation is often mistaken. So far as it goes, that seems correct, but my guess what be that an awful lot of the mis-evaluating just has to do with the extreme difficulty of assessing people's contributions to team defense. Beyond defensive rebounding, the numbers available strike me as next-to-worthless; giving people credit for steals, for example, doesn't make much sense unless we understand something about steal rates and the costs of failed steal attempts.
| 
Heads in the Sand
One point Berri makes time and again on his blog is that payroll size and wins correlate only very weakly in the NBA even though individual players' statistical production is fairly consistent from year to year.
Consistent player production might mean that there are standard models of both basketball and player at work. When you see a totally different understanding of basketball deployed--like that of the Suns--you see player production numbers go up. It's not clear that the important thing to decide isn't playing style.
Posted by somecallmetim | November 23, 2006 11:41 AM