NY Times: N.B.A.’s Haves and Have-Nots
Posted by Justin Kubatko on January 14, 2011
This week in the NYT, I take a look at the disparity between the best and worst teams:
Keeping Score: N.B.A.’s Haves and Have-Nots
The column appears in this morning's print edition as well.
January 14th, 2011 at 10:20 pm
Justin: If you're going to compare the current SD of winning percentage to that of past years, you need to take account of the fact you have a smaller sample now (38 games) which will add some variance. An observed SD of .157 in the past implies a true talent SD of .147 (with an 82 game season), while the current SD of .179 implies a true talent SD of .159 this season (adjusted for smaller 38 game sample). So the observed difference of .022 is roughly cut in half to .012 once you take the partial season into account. That may still be real, but clearly is a smaller difference.
January 15th, 2011 at 7:29 am
I was going to make the same point about the sample size. There's also the possibility that there's some vague regression to the mean at some point in the last few games of the season as teams locked into their seeds begin benching their stars for the playoffs. I can think of a number of occasions this seems to have happened with 70+% teams. I'd guess that the more spread out teams are the more likely they'd be locked into seeds in the first place.
It would be interesting to see these historical comparisons based on a similar number of games played. Which I think would be a really awesome bball_ref.com content addition. The play index you guys have designed is phenomenal, but I often find myself wondering about this partial-season kind of thing. What if there was a feature that enabled us to compare teams and players after N games? This would be interesting in terms of player evaluations (eg, how does LBJ's 7.2 Winshares after 41 games in '10-11 compare with the first 41 games of his last five years?), or for tracking teams (eg, how does the Cetics 30-9 record compare with 39 games in seasons' past?). Such a feature would also be useful, I think, for the more statistically-oriented fan who's looking for intra-year player/team trends.
Granted, I have no idea how much work it would take the site to incorporate such a feature, and maybe there's not enough interest to justify it. But, I think it's one of the very few interesting statistical apps. that bball-ref hasn't yet added to its extraordinary, encyclopedic content base.