What are the best metrics for evaluating defense?
What are the best metrics for evaluating defense?
Currently when I want to evaluate a players' defensive ability, I look at synergy, counterpart data (from either 82games or Evan's site) and defensive RAPM. Are there any metrics I'm missing that give a better idea of defensive ability? Also, how reliable are these 3 different metrics? I've found that they do conflict quite a bit (which makes sense to a degree). In particular, I find the counterpart data seems to differ from RAPM and synergy (for example, Battier is the 2nd best defender via opponent PER but has very average synergy/DRAPM numbers). Has anyone done any studies on the correlation between these 3 defensive metrics? And as mentioned earlier, am I missing any other defensive metrics (besides stl%, blk%, etc.)?
Re: What are the best metrics for evaluating defense?
Those are among the best ones.
I definitely would only use "defensive rating lightly", if at all. An outdated, deeply flawed metric.
Rewards gamblers. Doesn't recognize individual impact on shot defense much (uses average team shot defense including for time when the player isn't even on the court).
Look for some of Jon Nichols work. I think he or someone else reported correlations here or elsewhere (at http://sportskeptic.wordpress.com/, evan's site, the wins produced site or back to the kettle calling the pot black paper).
He also constructed a simple blend. http://basketball-statistics.com/aboutcs.html That probably helped him get a NBA job. The blend can be done differently and probably better.
you big around the archives here. also look at this maybe http://basketball-statistics.com/sereda ... rttwo.html
there is more to do on defense. but something more than average and it can get noticed. scott got an nba job too.
I definitely would only use "defensive rating lightly", if at all. An outdated, deeply flawed metric.
Rewards gamblers. Doesn't recognize individual impact on shot defense much (uses average team shot defense including for time when the player isn't even on the court).
Look for some of Jon Nichols work. I think he or someone else reported correlations here or elsewhere (at http://sportskeptic.wordpress.com/, evan's site, the wins produced site or back to the kettle calling the pot black paper).
He also constructed a simple blend. http://basketball-statistics.com/aboutcs.html That probably helped him get a NBA job. The blend can be done differently and probably better.
you big around the archives here. also look at this maybe http://basketball-statistics.com/sereda ... rttwo.html
there is more to do on defense. but something more than average and it can get noticed. scott got an nba job too.