I am attending InterACT San Francisco 2007 this week and blogging live (or nearly live) from the sessions. This session is Michael Lewis, author of Liar's Poker: Rising Through the Wreckage on Wall Street,
The Blind Side: Evolution of a Game, and Moneyball: The Art of Winning an Unfair Game
. Michael said that writing about sports is a way to tackle broader issues. He particularly sees a value revolution in sports. The big problem is that athletes are paid huge sums yet the mechanism for assessing their value has not kept pace. Michael started with Baseball because as salary levels rose, the discrepancy between teams became significant. Everyone worried that the game would become an economic issue not an athletic one. The commission set up to look into this in the late 90s had Paul Volker (an economist) who asked how come a team like Oakland with a low payroll was winning? Turned out that the problem was not that some teams had lots more money but that the game was "inefficient" because the data was not well understood and so lots of the money was spent inefficiently.
The trick was not how the Oakland As played, though they had a different strategy, but how they used data off the field. Oakland looked at existing data differently, by considering on base percentage not hitting percentage for example, because the on base percentage was a better predictor of runs scored. They also realized that they needed more data. For instance, with the traditional statistics, poor fielding could result in a batter and pitcher having stats that reflected that fielding not their actual ability. Oakland had been capturing more data about hits e.g. angle off the bat and speed of the bat. They used this data to correct for "luck" in a game by considering how hits like this do overall. Found, for instance, that pitchers had lucky and unlucky years and used this to swap those that looked better than they were (thanks to a lucky season) for those who looked worse than they really were. Also started to take a very analytical approach to picking players - replacing traditional judgment with math-based analysis. Focusing on very specific statistics for players to find their value.
There is a culture clash as scouts feel they are being replaced by people with computers. This is interesting because the over/under valued asset is a person not a stock for instance. The As were looking for players who had a glaring defect that meant they were not valued correctly. For instance a relief pitcher who had two club feet and so could not move fast. But, in fact, his feet made his pitches unique.
What if these people were not athletes but regular employees? Despite all the data about the performance of these athletes they were still mis-valued. But who then could be mis-valued - answer, anyone. People relying on subjective judgment:
- Don't differentiate between luck and skill
- Are often deceived by appearances
- Have "fame bias" - 80% of MLB players could be replaced by a minor league player without a negative impact!
- Have a number bias - as soon as you count something it becomes a fetish. For instance, creating a "save" as a statistic led to it become a fetish and As would exploit this to drive up the value of a player
- Hard to commit to value you cannot see
Use of data can remove human feelings but can also reveal someone more deeply.
Anyway, Michael was a very interesting speaker. I would recommend two books I have reviewed for anyone interested in Michael's - The Halo Effect and Blink - as both discuss the challenges of judgment over analysis. You can find the full set of posts from InterACT in this category.
Technorati Tags: analytics, InterACT, Michael Lewis
