Sports analytics: a real game-changer

As 2,700 people settled into their seats on Friday morning at the MIT Sloan Sports Analytics Conference, author Michael Lewis surveyed the scene from the dais and reminded everyone of how this massive annual event got started.

“This is Daryl’s class,” Lewis said.

Daryl, in this case, is Daryl Morey MBA ’00, general manager of the NBA’s Houston Rockets and the co-founder of the conference. In 2006, when he got the Rockets job, Morey was an executive with the Boston Celtics and a lecturer at the MIT Sloan School of Management — and arranged to create the conference, first held in 2007, as a way of compensating for the fact that he could no longer teach at MIT Sloan.

The result is a sprawling conference on a sprawling subject: Sports analytics, which once simply meant reformulating baseball statistics, now draw upon a flood of information, like optical-tracking systems, and are used in all major sports. This year’s conference had 30 panels, a research-paper competition, and representatives of more than 90 professional teams in attendance, and revolved around the theme of “big data.”

Analytics are visibly reshaping how sports are played. Consider Morey’s own team: This past offseason, the Rockets signed three players — James Harden, Jeremy Lin and Omer Asik — who were especially appreciated by stats-geeks. The team subsequently shifted to a seemingly freewheeling offensive style that is actually an efficiency expert’s dream: The Rockets are now among the NBA leaders in attempting shots from both the three-point range and very close to the hoop — both logical places to shoot from, according to the numbers. And they are taking fewer long two-point shots (which numbers-crunchers hate) per game than any team in NBA history.

“I think what teams are going to be running in 10 years will be totally different,” Morey said on Friday, talking about the way offensive styles in basketball are evolving based on statistical feedback.

The event’s opening panel on Friday, which included Lewis, Morey and political analyst and author Nate Silver, served as a reminder that the lessons of sports analytics can flow into other realms of life.

“Sports is a very good, pure laboratory” for working on analytics problems and techniques, noted Silver, who worked as a baseball analyst before shifting to political forecasting.

Morey: ‘Black-box models are very dangerous in sports’

The popularity of sports analytics owes a lot to Lewis’ 2003 book “Moneyball,” which illuminated how the Oakland Athletics used the 1980s-era insights of pioneering baseball analyst Bill James to compete with wealthier teams. However, one theme emerging during this year’s panel discussions was the fleeting nature of such advantages — which spurs a need for new metrics.

At the time of “Moneyball,” few baseball teams still fully appreciated that on-base percentage (which takes into account how often a player draws walks) was a more fundamental component of scoring runs than batting average, so the A’s could make high-value acquisitions on that basis.

Today, however, “Things have evolved and we have to be more sophisticated about how we get an edge,” Farhan Zaidi, Oakland’s director of baseball operations, acknowledged during a panel on baseball analytics. 

It also means that whereas sports analytics once had an open-source ethos — James used to publish his models in the back of his books — more information today is proprietary. But that trend, Morey said at a Friday panel called “True Performance and the Science of Randomness,” is double-edged: People running teams need to understand the assumptions behind the metrics their analysts produce.

“Black-box models are very dangerous in sports,” Morey said, adding that as a general manager, “You shouldn’t accept them from your staff.”

To be good at analytics, said sports consultant Jeff Ma, a former member of the MIT blackjack team that became the subject of the book “Bringing Down the House,” you need a cool detachment about your own work. “Every model you make,” Ma said, you have to keep asking yourself, “Does it still apply?”

And the trend toward proprietary sports statistics may not help the analytics movement in general.

“The implications of that are not all positive,” said Voros McCracken, a baseball analyst who worked for the Boston Red Sox when they won the World Series in 2004. “I think the openness that Bill James [had] led to growth. The more people who look at something, the more likely it is that somebody’s going to find something.” Proprietary research, by contrast, means “the growth of knowledge will be slightly slower.”

Showing their Spurs

In some sports, advances in knowledge may come from recognizing the limits of analytics. “There is nothing like on-base percentage in soccer,” said Chris Anderson, a professor of government at Cornell University and co-author of a forthcoming book on soccer analytics, “The Numbers Game.” On the other hand, he suggested, research shows that the price of fielding a bad player on a soccer team may be greater than the value of having a superstar in the lineup.

The conference’s research-paper competition highlighted some other specific advances. The winning entry, by Timothy Chan and Douglas Fearing, adapted a manufacturing model developed by MIT professor Stephen Graves to quantify the potentially large extent to which defensively versatile players can help a team’s performance.

The MIT Sloan event also yields insights about the behind-the-scenes work of some prominent teams. The NBA’s highly successful San Antonio Spurs have long been thought to be practitioners of analytics, while being far more low-key about it than the Oakland Athletics. The Spurs finally placed their general manager, R.C. Buford, on this year’s basketball-analytics panel, the first time a team official has spoken at the event.

Buford suggested that San Antonio’s investment in analytics fell somewhere short of having a platoon of mathematics PhDs locked in a room doing research, but involved study of whether the numbers backed the intuitions of head coach Gregg Popovich, such as the idea that three-point shots taken from the corner of the court are desirable. (Indeed, statistics show they are the highest-efficiency shot in basketball.)

“[Popovich] got interested when the data supported him,” said Buford, who also made clear that the Spurs have used analytics to try to upgrade their defense this season.

Indeed, much of the interest — and charm — of the conference comes in those moments when teams or former players and coaches disclose their openness to new forms of knowledge. Twenty years ago, Jack Del Rio was a veteran NFL linebacker mostly interested in leveling running backs. On Saturday, Del Rio appeared on a football panel next to Brian Burke, a former Navy fighter pilot who runs the web site Advanced NFL Stats; the two eagerly talked about the use of metrics such as win probability — the odds of a team winning at any moment in a game, depending on the circumstances — as a guide to making key coaching decisions.

“It’s just the direction things are evolving in this sport,” Del Rio said.