May 24, 2006

Basketball statistics and Malcolm Gladwell

In response to Gladwell's book review of The Wages of Wins in the New Yorker credulously accepting three economists' formula for ranking basketball players, an algorithm under which Allen Iverson came out 91st best in his MVP season and Dennis Rodman much better than Michael Jordan in their last season together, readers write:


If you want to understand basketball and stats, you're much better off reading the likes of John Hollinger (Pro Baskeball Forecast) and Dean Oliver (Basketball on Paper) and 82games.com. If you haven't already, that is.


In Hollinger's rating system, Iverson was the 13th best player per minute in 2001 (his MVP season), and he always plays a huge number of minutes per game (so a per minute ranking is a little unfair to him and to other little guys -- big guys often have to rest more, in part because of strain on the knees). And Jordan was #4 in 1998 while Rodman didn't make Hollinger's top 50 (which seems to underestimate Rodman, who led the league in rebounding).

My impression that Iverson, who is 6'0" and 165 pounds is clearly one of the world's greatest athletes, but it's a little less clear that he is one of the world's greatest basketball players, simply because he is of average size. Still, he is an awfully good basketball player, much better than Gladwell's gurus' stats claim.


For example, Hollinger points out that Iverson's true value is that he can create a shot at will. He doesn't depend on getting set up by others, which is highly useful for when the shot clock is running down. The downside is that he's a low-efficiency player [career .421 field goal shooting percentage]; Jordan was great because he was high-efficiency as well [career .497 shooting percentage], with a low turnover rate [career 2.6 per game vs. 3.7 for Iverson].


One question would be whether Iverson cooperates with his coaches. If a coach told him he wanted him to forego his six hardest shots in each game so that his final shooting line would end up as 8 made out of 18 shots instead of 10 out of 24, would he do it?

Something that is often forgotten about Wilt Chamberlain is that he'd do whatever his coach asked him (except show up for early morning practices!) One year, for example, he became the only center ever to lead the league in assists, because that's what the coach wanted. Chamberlain might have been a little too cooperative for his own good. In contrast, early in his Laker career, Magic Johnson organized a coup that got his coach fired (who'd already won a championship) and replaced with Pat Riley, who changed the team strategy to better suit Magic. They went on to win four championships.


Another reader writes:


I was shocked by Gladwell's acceptance of a flawed idea by mid-article. He seems, at first, to grasp perfectly well the complexity of rendering judgments about which it is uncertain which data to choose or how to weigh it. He notes well that the Rookie of the Year award is poorly judged. Since it's useless to anyone but the winner, and seems to depend on how bad your team needed a scorer, it's low hanging fruit. Fine.

However, he doesn't stop here. Gladwell is so impressed by the mental machinations of these economists that he's willing to suspend his own common sense and judgment, dismiss a career (Iverson's) that has many of the hallmarks of a great one, and abandon his prior skepticism about the limits of experimental design. He does this without presenting much evidence that the model has a shred of predictive value, or can adapt to changing attitudes....

Instead of accepting the limitations of current hoop accounting techniques, the idea we are left with is that we should trust the admittedly flawed data, and, as long as the economists do something really neat with it, not our own eyes. Sounds like a hot trend. Also, dismissing a man's life work in toto often cries out for one to make a rational case of one's own.

However, it would be very unsexy for Malcolm Gladwell to say "I can see the use of this model to improve the statistical measure of the following five elements of pro basketball." or "Here are five reasons why Allen Iverson's statistics are not particularly useful." Hell, he might have gotten a book with these guys out of the first. Rather, he seems to buy the whole thing, and uses bad examples to make an argument that shows other theory's weaknesses replacing it whole hog with a theory that at least he has not proven to be stronger in many areas.


Another reader comments:


I've had some success looking at team performance using two key metrics, scoring capacity and scoring efficiency. Scoring capacity is the point total a team would score if they hit all their shots. Scoring efficiency is a team's actual point total divided by their scoring capacity.

The University of Kentucky is a great test environment for this model, since they have detailed stats back to the 1940s and examples of great (and poor) performances over the years.

When Rick Pitino was coach at UK in the 1990s, his obvious strategy was to maximize his scoring capacity margin over his opponents. There are only three drivers for this: offensive rebound margin, turnover margin and 3PT attempt margin. Pitino coached a 14-14 team in 1989-90 to a scoring capacity margin of 24 points per game (unbelievable!) by shooting the 3PT, defending the 3PT, and full-court pressing to force turnovers. But that team had a 6 percentage point DEFICIT in scoring efficiency against the opposition. However, by 1992 Pitino had recruited better talent, so his 1991-92 team had a 21 point scoring capacity advantage over the opposition and was dead even in scoring efficiency.

If you look at that great Kentucky-Duke game in the 1992 East Regional Finals [Duke won 104-103 as they went on to win the national championship for the second straight year], it's the clash of two different basketball strategies. Kentucky maximized scoring capacity margin and Duke maximized scoring efficiency. That's what made it such a great game, Kentucky pressed, forced Duke into a boatload of turnovers, and had a much higher scoring capacity. Duke was extremely efficient, with Christian Laettner going 10-10 from the field and hitting the famous game winner. Duke was a big favorite but was lucky to win. Kentucky showed that scoring capacity matters and most analytical systems focus only on efficiency.


My published articles are archived at iSteve.com -- Steve Sailer

No comments: