For Top Players, Mets Count on Him

Mets statistician and Hunter College instructor Benjamin Baumer is trying to improve basic forumlas used to evaluate players' fielding skills.
Growing up in Massachusetts, Benjamin Baumer, a boy with "a pretty strong feel for quantitative stuff," collected baseball cards. He still watches ball players by the numbers.

As baseball season swings into summer, the doctoral student and Hunter College statistics instructor spends much of his time conferring with New York Mets General Manager Omar Minaya and his management team, analyzing and evaluating players.

"I try to verify independently what people are seeing -- what appears to be true," says Baumer, the team's full-time statistician. Baumer -- a teacher of elementary statistics and probability whose interests include graph theory, combinatorics (the study of counting), discrete geometry and graph algorithms -- is a player in an emerging field that uses statistical methods to give a fuller picture of ball players' skills and performance. "It's actually two jobs," he says of his Mets position. First, he serves as an analyst and adviser on player evaluation. His second job involves building a website as well as a database of player statistics from feeds he gets from major league baseball sources and proprietary database providers.

Baumer, who grew up in Northampton, Mass. -- amid Red Sox Nation -- played ball in high school, as an infielder and pitcher. He majored in economics at Wesleyan, then earned a master's degree in applied math from the University of California, San Diego, in 2003. That summer, with the release of Moneyball, the best-selling book by Michael Lewis, the baseball establishment was turned on its head by statistics.

The book chronicled the unconventional approach of Oakland A's general manager Billy Beane, who touted advanced statistical analyses of player performances over the collective (but often subjective) wisdom of managers, coaches and scouts.

Beane's strategy was to identify and scoop up undervalued players in order to compete with big-market clubs with more financial resources. Such tactics came to exemplify a burgeoning movement in baseball, known as the "sabermetrics," which attempted to create more objective measures of a player's ability to help his team win, giving rise to a new generation of statistical acronyms like WHIP (walks plus hits per inning pitched).

When Baumer returned home to Massachusetts, a sports economist introduced him to the Mets general manager and in 2004 he was hired as the team's statistician. A year later, he entered the Ph.D. program in mathematics at the CUNY Graduate Center.

At last year's CUNY Statistics Seminar, Baumer presented a research paper that analyzed several sophisticated methods being proposed by researchers to evaluate players' sometimes-elusive defensive skills.

This graphic represents one type of mathematical model being used to assess the skills of baseball players. The bell-shaped image shows the probability of an average center-fielder catching a fly ball hit to various locations on the field. CUNY researchers Benjamin Baumer and Dana Draghicescu are working on methods to improve such models.
Traditionally, a player's defensive ability has been calculated by fielding percentage -- the number of assists and putouts a player records, divided by the total chances a player had to make these plays. But this only tells you how players handled the balls they were able to put a glove on; it gives little credit to a fielder who can run down a ball that is out of range for most players. Baumer examined two basic mathematical models -- "discrete" and "continuous" -- being developed by researchers to better assess fielding ability. The discrete model divides the field into zones and calculates the probability that an average player will catch balls hit in a particular zone, compared to how a player actually played. The continuous model treats the field as a continuous playing surface. Using existing data that describe all the balls a player has fielded, the model then calculates the probability that a player would catch other balls across a continuous plane. While all these methods work to varying degrees, the metrics still credit players in different ways, so statisticians have not agreed on any "gold standard" yet.

Baumer has now teamed up with another CUNY researcher, Dana Draghicescu of Hunter College's Department of Mathematics and Statistics, to develop better statistical models to assess batters' skills. Draghicescu, who grew up in Romania, earned a Ph.D. at the Federal Institute of Technology in Switzerland and then did a postdoctoral stint at the University of Chicago. Although she was trained as a theoretical statistician, she was interested in adapting her methodology to real-world applications, such as environmental and medical research. When she met Baumer, she says she knew little about baseball, but she did know one thing: Her expertise in analyzing and predicting environmental data, like precipitation patterns, through space, could also be useful in mapping "hot zones" -- the areas in or around the strike zone in which baseball batters are likely to hit the ball.

Draghicescu and Baumer recently discussed some of their hot-zone mapping techniques at the CUNY Statistics Seminar and in August will present a joint paper on their research at a major conference in Vancouver. While predicting outcomes in sporting events is always tricky, Baumer points out that baseball is "a very good game to do this kind of analysis." Not only are there hundreds of statistics, but given the rules and structure of the game, "data sets" are easier to assign to specific players than in some other sports. Nevertheless, "plenty of teams are drifting in that direction [of statistical analysis]," Baumer says. "You're seeing it a lot more in basketball, football and hockey."

If there's one thing that's certain, says Draghicescu, it's that Baumer will leave his mark as a statistician on America's pastime. "I predict that he will become an expert in this area," she says.