Free-Floating Hostility

Monday, April 11, 2005


Moneyball on the Hardwood

I am watching the Oakland A's play the Toronto Blue Jays on television right now. Both teams are run by Bill James disciples, which means that most of the decisions made on the field have some statistical foundation and that there's lots of emphasis on the most predictable and measureable part of the game, hitting. The teams, therefore, are taking lots of pitches, conserving outs, risking little on the basepaths, and threatening to turn every defensive play into a Monty Python sketch. Seriously, earlier in the game Oakland outfielder Eric Byrnes was nearly knocked over by a bloop single that fell eight feet in front of him. It was ugly.

Sophisticated statisical analysis of baseball has existed for years. The gold standard is baseballprospectus.com, which was great until it became a pay site. Now, fans of the other sports are trying to get into the act. There's a site called footballoutsiders.com, which attempts to derive statistical meaning from the NFL, something that's already been discussed on Jeff's old blog, to which I cannot link for technical reasons.

A few months ago, Form sent me a link from Mark Cuban's blog where he details some of the stats the Dallas Mavericks coaching staff use when developing game plans. I can't make head or tail of it. Recently, some guy has started 82games.com, which is meant to be a companion to NBA.

There are some interesting insights, such as analysis of NBA shot clock usage that shows how many more points teams can expect to score early in the shot clock. This suggests that the most important defensive concept is transition defense. The Detroit Pistons starting five is so crucial to the team's success that I'll spend the bulk of my time between now and the playoffs hoping none of them get hurt. But it also suggests that someone named Brian Cook is more helpful to the Lakers' fortunes than Kobe Bryant. And that Yao Ming actually leaves the Rockets at a deficit. That's the fun parts of SABRmetrics, making brazen (and likely wacky) statements and then backing them with statisical evidence.

The question I have is whether a team in a league as dominated by its stars as the NBA, could ever operate like the Oakland A's. My guess is no. That's in part because of the salary structure, which forces teams to make basketball decisions for reasons that have very little to do with basketball. More fundamentally, player statistics are impossible to isolate in basketball. Take L.A.'s Lamar Odom, whose scoring is down 2 points from last year. His shots attempted are also down two and he's taking one fewer foul shot per game. His shooting percentage, however, is up. All of those changes have something to do with Kobe Bryant's dominance of the ball and volume shooting. But how much? It's also impossible to truly separate phases of the basketball game because defense and offense are not discrete as they are in baseball. It's interesting stuff, but, in terms of the NBA, the new frontier of knowledge may end up teaching us very little.

2 Comment(s):

  •   Posted by Blogger Form at April 12, 2005 5:04 AM | Permanent Link to this Comment
  • The NBA game does not work well for this type of analysis. I went to 82games and found out that Amare Stoudamire and Shawn Marion score a lot off of passes by Steve Nash. That is a statically insignificant fact bacause you need zero information from statistics to know it.

    What I would really like to see is a behavioral economics analysis of NBA coaching, showing how the endowment effect and control heuristics influence coaching. I'd be willing to bet they have a lot to do with why teams only try to score late in the shot clock.

  •   Posted by Blogger Mike at April 14, 2005 9:13 PM | Permanent Link to this Comment
  • At the heart of the whole enterprise is the goal of assigning value to each play. That Shawn Marion and Amare Stoudamire have accepted lots of passes from Steve Nash and deposited them in the basket is a counting stat and has no meaning on its own. But, under the assumption that a player is worth only what he does statistically, then one has to figure out how much Stoudamire's game is actually predicated on dimes from Steve Nash.

    The points per 100 possessions conversion is something lots of coaches look at. When talking about a basketball team's efficiency, it's important to know that in the NBA points are bought in bulk, which is different than say, runs. I don't have the average margin of victory in the NBA in front of me, but some gambling site says that in 2001-02 more than half the games were decided by between 2-9 points.

    In the Shaq/Kobe argument, many people (people who are not bright) thought Kobe was the better because he hit the important shots late in games. But what they forget is that to get to those late game situations, you need bulk scoring early. It creates a situation where no possession is especially valuable, but every possession is valuable because it comes down to possibly less than three possessions per game.

Post a Comment