There is a (relatively) large upheaval in sports right now. The situation is similar to Wall Street a decade or more ago, when traders bought and sold stocks based on gut feeling and inside knowledge about a particular industry. Then came the quants — PhDs in quantitative disciplines such as Physics and Mathematics from the ivory tower — and their mathematical models that told people when to buy and when to sell. Wall Street has never been the same since…

The quants are now making a headway in sports. The success story has been baseball, where quantitative analysis has even made its way into the mainstream media. The relatively-avid baseball fan knows a bit about Billy Beane, Bill James, Baseball Prospectus, and SABR. It’s not uncommon for baseball to hire people with outstanding quantitative skills and limited traditional baseball experience (and this). It’s safe to say that quants have very much influenced baseball — people are now paying attention to the various statistics and measures that they’ve come up with (value over replacement, etc.).

Football seems to be the next sport to topple. One could actually argue that it’s already been toppled, what with the BCS already being decided by a bunch of freaking computers. Coaches are already paying attention to quants; Belichick might not be running up the score this season when he goes for it on fourth down in the fourth quarter — he might just be following the optimal dynamic programming schedule. It’s known that he’s read it.

And of course, basketball is seeing the eggheads. There’s Daryl Morey, who taught a course at MIT titled “Analytical Sports Management”, now as GM of the Rockets. And just this season, Ken Pomeroy has started a website Basketball Prospectus (named after the wildly successful Baseball Prospectus site) that takes an analytical approach to college basketball.

Now, don’t get me wrong. I am a man of science, and I truly believe in the power of reason over gut feelings. But I have strong reservations about such quantitative approaches to basketball. Baseball is a discrete game very amenable to analysis — each at-bat is reasonably well-approximated by just two players interacting (the pitcher and the hitter), each at-bat can reasonably be taken to be independent of previous at-bats, etc. It’s these simplifying assumptions that allow observations such as on-base percentage is what matters, not your hitting percentage. It allows people to reasonably simulate a baseball game and determine what order of hitters is optimal. My belief is that it’s these reasonable assumptions that have allowed quants to have such a large effect on baseball. I’m less of a believer that quantitative analysis can model football well, but I believe two things make football amenable to quantitative analysis — the first is that it’s discrete, so that events can be separated from one another with ease. The second is that most everything is planned — and it’s the plays that largely dictate play, rather than individual improvisation.

Basketball, on the other hand, has nothing going for it — something that happens on one side of the court directly affects the play on the other side, so the separability from play to play doesn’t exist. Individual improvisation is everything; sure, college run offenses and defenses, but more often than not those are general methodologies of play, rather than specific instructions as in football or baseball. Statistics that quants have been pushing for — pace-independent measures of ball security (turnovers divided by posessions), effective field goal percentage (weight three point makes more than two point makes) — make sense, but I don’t believe they capture basketball in the same way the measures do in baseball and football. Of course, I would love to be proven wrong.

Categories: Commentary, donald