NEWS

WIRES

OTHER

A winning season: MIT Sloan Professor Dimitris Bertsimas uses quantitative analytics to predict the Red Sox will win 101 games this season to the Yankees' 93

"I am a big believer that quantitative analytics can have a major impact on businesses, including sports teams," says Bertsimas, who is also co-director of MIT's Operations Research Center. "We calibrated data based on last year's statistics as well as this year's spring training to determine that this year, the Red Sox will win 101 games - which is a rarity in baseball, and makes for a very winning season. In Las Vegas, the odds are for the Sox to win 95 games so we have a pretty good benchmark."

In a new paper* and case study, Bertsimas and Operations Research Center doctoral student, Allison O'Hair, developed a series of three analytic models to determine the outcome of a team's 162 game season. The models take in a variety of statistics from the number of runs a given team scores, to a particular player's on-base and slugging percentage, to defensive figures such as runs allowed.

"A player is a vector of numbers and from that, we can make accurate predictions of how many runs they will score and translate those to overall team statistics," says Bertsimas, who teaches a course at Sloan about how companies from Google to Goldman Sachs to Federal Express use analytics to boost their bottom line. "It's human intuition that tells us which factors to look at to make these predictions, and how much to weigh them, but after that, we let the data speak for itself."

Bertsimas retroactively ran numbers to forecast last year's regular season wins for the Red Sox after spring training in 2010. Bearing in mind that there are some players that rarely play during the games, he only took the top 15 hitters in terms of at-bats and the top 10 pitchers in terms of innings pitched from spring training, which is a reasonable guess as to which players would be used the most during the regular season. Instead of using the statistics from spring training, he used their statistics from the 2009 regular season. According to his models, the Red Sox should have been expected to win 90 games; in fact, the Red Sox won 89 games.

Bertsimas based some elements of his models on the book Moneyball: The Art of Winning an Unfair Game, which is the story of how Billy Beane, general manager of the Oakland A's, used quantitative methods to create a winning team. His models also draw on the work of Bill James, who is credited with popularizing the use of analytics in baseball.

"Now, every major league baseball team has a statistics group and about three-quarters of the teams are believed to incorporate quantitative methods into their decisions," says Bertsimas. "In 2002, John Henry, an extremely successful futures trader who believes in the use of analytics in baseball, bought the Red Sox. It's no accident that after this change in leadership the Red Sox won the World Series in 2004 and 2007, after an 86-year dry spell."

While his models accurately predict which teams will make the playoffs, they fall short of predicting which teams will prevail in the post-season. The problem is sample size. Over the 162 regular season games, luck evens out and skill is more important, but in a series of five or seven games, luck is a much larger factor, according to Bertsimas.

"In a five game series, the worst team in baseball will still beat the best team in baseball 15 per cent of the time," he says. "Analytical principles are very useful for getting a team to the playoffs, but they are much less helpful once the playoffs start because the level of randomness is much higher. Any general manager worth his salt sees his job as getting the team to the playoffs, but once they get that far, luck plays a much larger role."

* The Analytics Edge in Baseball; Sloan School and Operations Research Center, Massachusetts Institute of Technology; Dimitris Bertsimas and Allison O'Hair

MIT Sloan School of Management
Paul Denning, 617-253-0576
Director of Media Relations
denning@mit.edu
or
Patricia Favreau, 617-253-3492
Associate Director of Media Relations
pfavreau@mit.edu