Sports

Stat prof Wang shares unique perspective on managers

In print | March 27, 2008

As the Major League Baseball season prepares to swing into action, one Swarthmore professor has certainly been getting in the spirit.

Last month, Assistant Professor of Statistics Steve C. Wang presented at an American Association for the Advancement of Science conference about statistical work he had been doing to evaluate Major League managers.

It’s an interesting challenge. Managers are a very visible, oft discussed part of every baseball game, yet even the most avid baseball fan has no grasp of a manager’s numbers. “People don’t keep stats [about managers] very often. Without the statistics, you really don’t know,” Wang said.

For example, what was the highest number of pitchouts called for by a manager in 2007? An observant watcher might be able to give a ballpark figure but not much more than that (it was 56, by the way).

By contrast, players’ statistics are so ubiquitous that they alone can largely identify a player. As Wang put it, “Player statistics have language.”

For instance, figure out which stat line corresponds to Red Sox slugger David Ortiz and which belongs to Red Sox outfielder Coco Crisp: 35 home runs/3 stolen bases, 6 HR/28 SB. Any casual fan of the sport will know that the first set corresponds to Ortiz, the latter to Crisp.

Wang seeks to bring some of that same power of numbers to managers.

His study, which he presented to the AAAS, sought to find differences and similarities among managers and see if such tendencies were consistent over time.

The study compared the data of every Major League manager across a wide spectrum of categories, from number of lineups used to number of runners sent in motion on a pitch to number of pinch hitters used.

But there’s a significant challenge to be overcome in such a study. Comparison of managers based on one stat can be visualized on a number line; such a comparison for two stats can be represented on a Cartesian plane. But what about six or seven dimensions, or more?

“It’s an interesting, challenging problem to make understandable pictures,” Wang said. His solution was very interesting as well, and appealing to the eye.

First, he used Chernoff faces. The idea behind this technique is that humans are very adept at recognizing subtle differences in human faces.

So, a statistician might present multi-dimensional data as faces, where each category is represented by a facial feature. For example, big ears might mean the manager has a propensity to pull pitchers early in a game, while small ears mean he lets them go for a while. In this way, Wang constructed a Chernoff face for each manager.

Aside from being somewhat comical, the faces did allow for patterns to be seen—large similarities or stark differences were immediately evident. However, these cartoon faces provided only so much insight.

Wang found a more effective way to compare these managers over multiple dimensions. He created a table (a “heat graph”) in which each cell was specific to a certain manager and a certain trait. For instance, the first column was Jerry Narron, manager of the Reds, while the first row was “long starts,” the proclivity of a manager to leave in his starting pitcher. In that way, each manager had a cell for each category.

However, those cells did not contain numbers; they were colored in. So if Jerry Narron tends to leave starters in a long time (which he does), cell 1A is a stark red. If a manager has a low rating in a given category, that cell is colored in blue. In this way, the various dimensions of these managers leap out of the table and are very visually apparent. This allows them to be grouped and analyzed in much more precise and directed ways.

Though Wang’s study only scratches the surface of available research on managers, he believes he has found some interesting results.

While managers certainly adapt to their players, he thinks there are definitely strong tendencies that a manager will show throughout his career, even with different sets of players. For instance, Tony LaRussa of the St. Louis Cardinals has consistently used among the most different lineups over the course of a season, while Phillies’ manager Charlie Manuel uses very few.

Wang hopes to continue his research by using his visualization methods to compare managers who have worked together. For instance, while working on his current study, he noticed that Willie Randolph of the New York Mets and Joe Torre of the New York Yankees (until this year) have very similar profiles. Since Randolph worked as a coach for Torre with the Yankees for years, Wang wonders if managerial styles are passed down. He will try to examine more teacher-student pairings in the future.

Like all baseball fans, Wang is eagerly anticipating the 2008 season. He grew up in New York state, but does not think that influenced his rooting interest.

“I’m a Yankees fan because that’s the first team I ever heard of,” he said. His favorite player was Reggie Jackson for the same reason. The 1970s, the time of his childhood, was a good time to be a fan of Jackson and the Yankees, as they won three pennants and two World Series from 1976-78.

Like the rest of the baseball world, he is eager to see how the Yankees’ trio of young pitchers — Phil Hughes, Joba Chamberlain and Ian Kennedy — will perform on the big stage over the course of a full season.

Wang’s passion for baseball and admirable statistical work are well valued here at Swarthmore. Andrew Quinton ’08, an avid baseball fan who has worked with Wang as a student, grader and clinician, appreciates the opportunity to speak with the professor about their mutual passion (even though Quinton roots for the Red Sox).

He has found Wang to be very knowledgeable and accessible. “It’s just been cool to know that someone I know personally is also reading Web sites like The Hardball Times and Baseball Think Factory, and to have another stathead-type fan to talk with from time to time,” Quinton said.

In addition to being interesting to baseball fans, the quality of Wang’s statistical work is impressive. His skill with data visualization is unique.

“Steve is constantly asking how best to display a set of data, and from these displays he gains insight into what differences or similarities might be interesting to examine in a more formal way,” Phil Everson, Associate Professor of Statistics, said. “Data visualization is something all researchers should do more often, and we are fortunate at Swarthmore to have an expert among us.”

So while you enjoy the start of the 2008 season, try to think critically about your favorite team’s manager. Does he seem to go out to the mound to spell his pitcher earlier than his counterparts? Is he willing to use a bench option in a tight moment? Do players seem to be moving around the lineup a lot?

Managers have tendencies, and though we rarely notice them, they affect games and seasons in very important ways. After all, their decisions, big and small, shape the context in which the players play.

The more you recognize these men’s tendencies, the deeper your understading of the your team, and the sport in general, will be.