Features

Greg -- if you really want to get the scoop on data mining in baseball, forget about the NYTimes article, which is frankly old news.

The real analytics leader in the baseball industry is the Oakland A's, as chronicled in the GREAT book "Moneyball" by Michael Lewis. (Moneyball: The Art of Winning an Unfair Game, Michael Lewis, W.W. Norton & Company; 2003, Editor)

They are years ahead of any other major league franchise, with the exception of the Red Sox (and then only because the new GM of the BoSox was the assistant GM for the A's before he got hired to go to Boston).

The A's have also (like a lot of the more successful corporate entities who've embraced data mining, such as the credit cards and some of the banks) converted their organizational culture to run their business based on the data and the analytics, instead of simply running the analyses in the back office but going about their decision-making via business as usual. This means, in baseball terms, running their recruiting organization to find the right players with the right profile that ends up winning games. Most baseball organizations have not made that transformation, and as a result, most of the analytics they run go nowhere in terms of really changing the way they play the game.

The result: the A's, on the second smallest franchise budget in the majors, consistently make it into the playoffs (and note they lead the AL West this season right now too), while their peers based on salary ranges, never consistently perform at that level--for instance, the Expos, the Orioles, the Pirates. Even some teams that have extremely high salaries, like the Astros, can't turn that investment into a consistent high performing business.

Success at analyzing statistics in baseball, as Billy Beane--the GM of Oakland--has found, isn't just about applying new types of analytics to the same old data, it's really about getting the right data (which is often not the kind of data that is traditionally collected) and applying some of the more traditional analytic approaches to it--and then following through on the results and testing/measuring around the ideas that get generated.