The Battle of the Statisticians

March, 23, 2007
3/23/07
11:30
AM ET

This is one hell of an exciting time to be a basketball statistician. Those precious few experts who understand basketball at a deeper level, guys like Dean Oliver, Dan Rosenbaum, Jeff Sagarin and others, have quickly evolved from, well, nerds into nerds who have some say-so at the NBA teams where they work. Many others, like ESPN's own John Hollinger (Insider), David Berri who co-wrote "The Wages of Wins," and Roland Beech of 82games, are being read far and wide.

This is basketball's Moneyball moment. Our children will marvel that we ever considered good ol' points and rebounds as all that important in assessing players.

One of my favorite things about this whole process is that by and large the stats crowd is impressively cooperative with each other. It's a beautiful thing to see. They share their work, and when they rip each other's theories to shreds they typically do so in civil tones.

There has long been a refreshing we're-all-in-this-together attitude -- which is appropriate, because even though some may work for competing teams, websites, universities etc., they are all in it together. Despite signs like stat-friendly Daryl Morey's taking over of the Houston Rockets, and stat-friendly Mark Cuban's owning the Dallas Mavericks, it remains to be seen how much the old-school veterans who run much of the NBA will truly embrace the findings of these smartypants newcomers. The statisticians might be in the room, but they're still outsiders.

Part of the reason for that, I believe, is that, publicly at least, no one has found the holy grail -- the formula or metric by which you can determine with any degree of certainty that drafting this or that player, or playing this or that lineup, will lead to more wins. The best formulas out there are still up for discussion. (There are still divisions on basic questions, for instance: can you base a good system on the boxscore data the NBA provides, or do you have to chart games a whole different way to get meaningful results? Or is there some third approach that works?) It's an exciting time.

In this environment, however, a couple of unwritten rules seem to have evolved: no one should take too much credit for creating the perfect system, because there isn't one yet. And no one should consider him or herself above joining the discussion to explain their techniques so that the rest of the group can learn from them.

David Berri is a very respected academic who wrote a very interesting book called "The Wages of Wins." He counts Malcolm Gladwell among his fans. (Berri also has a blog he updates frequently.) The book covers a lot of ground, and people who know a lot more about it than I do swear that the science behind the vast majority of the book is extremely solid and helpful.

However, Berri has recently earned the nickname the "the Ann Coulter of basketball" from others in the stat community, apparently because he has broken the two unwritten rules described above. On his blog he often takes credit for his high rate of success in predicting wins using a formula he developed. Simultaneously he has thus far refused to make public the nitty gritty of how he arrived at those results.

Critics say his formula contains a certain fudge factor, an unspecified "team adjustment." A team adjustment, they say, can make almost any halfway decent system look brilliant. Berri has been asked many times to describe that team adjustment, and as of this writing he has refused to do so in a satisfying way.

Yesterday, one of his critics, Dan Rosenbaum (who advises the Cavaliers) essentially did a quickie reverse engineering of Berri's key piece of work, and apparently developed an even more accurate model (based on applying it to several years of data) than Berri's. Several people who know more about this than I do have emailed me about the happenings, which you should read for yourself on the APBRMetrics message board, and declared this more or less the final nail in the coffin of the "Wins Produced" metric at the center of Berri's basketball models.

Will this close this chapter of the discussion? I doubt it. I'm sure there is plenty more to learn from all involved, including Berri. But I have to admit that I do hope this serves as encouragement for everyone involved to keep collaborating as much as possible. It's good for the game.

SPONSORED HEADLINES

Comments

You must be signed in to post a comment

Already have an account?