Thursday, February 03, 2011

Why Not Combine WP48 and Usage Rate?

(apologies for startup fans--the following blog post is all about basketball stat-geekery)

One of the major revolutions in basketball fandom over the past decade is the rise of advanced statistical analysis, similar to Sabermetrics for baseball. Thanks to folks like John Hollinger and David Berri, NBA fans now have a plethora of stats. My personal favorite is Berri's WP48 (wins produced per 48 minutes) because it is already position-adjusted (e.g. Centers are measured differently than Point Guards).

However, the arrival of advanced stats hasn't ended the interminable debates. The main criticism of WP48 is that it favors rebounders who don't shoot (much like the book Moneyball implies that Billy Beane's genius lay in signing fat baseball players who walk a lot). We don't like the idea that rebounding specialists like Ben Wallace are superior to balletic guards like John Wall.

A bigger issue is the intuitive argument, "But a team of 5 Ben Wallaces would stick, because none of them could shoot."

Similarly, a lot of stat geeks predicted that the Miami Heat would be unstoppable, thanks to the statistical contributions of LeBron James, Dwyane Wade (damn you, why can't you just spell your name correctly!) and Chris Bosh.

That had the opposite problem from Ben Wallace--all three of those guys want the ball in their hands all the time, which meant that their individual stats weren't additive. We measure this ball-neediness with the statistic of usage rate, which measures what proportion of a team's possessions a player uses while he's playing.

My proposal is simple: Acknowledge the fact that when you construct a team, you have to select a lineup with a cumulative usage rate of 100%.

If you build a team that looks great on paper, but has a usage rate of 200%, recognize that the numbers are going to be off--badly. If you build a Ben Wallace team that has a cumulative usage rate of 50%, recognize that the numbers are also going to be off.

Many stat geeks feel that the best players on the Lakers are Pau Gasol and Lamar Odom. Both generate high WP48 scores without taking a lot of shots. But without Kobe Bryant, both Gasol and Odom would be forced to take more shots than they could effectively use. Just watch the Lakers try to play for long periods without Bryant and you'll see what I mean. (And I still think Kobe shoots too much.)

My principle adds additional complexity. There isn't a single stat for comparing players. You can't simply say, "LeBron James is the best player in the NBA." What you can do is better assemble a lineup of players who cover all five positions on the court, and balance high efficiency with natural usage rate. It's not as easy to explain or argue, but it's probably more effective.

No comments: