Two years ago, at the fifth annual MIT Sloan Sports Analytics Conference, the backdrop in the main panel room featured a single image, repeated like wallpaper: Kobe Bryant putting up a fadeaway shot as the Houston Rockets' Shane Battier conspicuously sticks his hand in his face. To the roughly 1500 attendees, it was a loaded image—a picture backed by a thousand words. That simple move forms the centerpiece of Michael Lewis's 2009 article casting Battier as "The No-Stats All-Star," a player who exploited analytics favored by Rockets general manager Daryl Morey. Those analytics reveal, among many other things, that Bryant's shooting percentage goes down when you put a hand in his face. The image's message was simple, progressive, ameliorative: Battier played the tendencies, and this gave him an edge.
At this year's conference, held last weekend in front of nearly twice as many attendees, the backdrop to the first panel (moderated by Lewis) presented a much more jumbled impression. There were photos of a number of sports, only some of which featured famous athletes (a shot of LeBron James driving, for instance), and lists of what appeared to be the gambling odds of NFL match-ups. But this backdrop, too, reflected its conference's overriding theme: the manic angst of having more information, which makes finding the optimal ways to use it—and, more to the point, to use it usefully—more elusive than ever.
Each year, as ESPN's Kevin Arnovitz has noted, more and more people are coming to Sloan. That's literally true, but could be said figuratively of sports analytics in general. The days are long gone when seemingly unremarkable players could be signed on the cheap by the few teams smart enough to understand the value of, say, a high on-base percentage. "You used to know how other teams operated," complained stats-friendly Dallas Mavericks owner Mark Cuban at the opening panel. "Now you have to reverse-engineer what they did to see how they do it."1
Or, as Nate Silver put it on the same panel, "There's not the low-hanging fruit anymore of having some teams that are totally stupid."
Over the course of the two-day conference, I asked many people which sport—out of the four major U.S. team sports (baseball, basketball, football, hockey) and soccer—is least amenable to an advanced analytical interpretation, where little is to be gained by looking at the game from a new, maverick angle. In short: Which sport can't be Moneyball-ed?
Nearly every sport was mentioned, and all for different reasons.2
"Let me rank 'em," Morey said. "Baseball, basketball. Now there's [sports] where there's the least data, and that's football—fewest number of games. There's soccer and hockey, where it's so hard to do immediate outcomes." He concluded, "Soccer's hardest, hockey four, football three. They're all so hard, though."
The only sport nobody mentioned was basketball, to which statheads are flocking just as they did to baseball a decade ago. Analytically speaking, basketball is in a sweet spot. There is a great deal of data, and some smart people have put a great deal of thought into how to use that data; but not all the teams use the data, so there is still a significant edge to be had. Twenty-nine of the 30 National Basketball Association teams sent representatives to the conference. (The lone holdout? The Lakers.) But only about half of NBA arenas, for example, have installed SportVU technology—a series of cameras that captures the precise locations of every player on the court.3 In five or ten years, when every arena inevitably has them and every team is using the information they collect, the available advantage will shrink.
Meanwhile, hockey, soccer, and football remain analytical backwaters. In the case of football, this is not for lack of trying—for reasons I'll explain shortly—whereas hockey and soccer seem to shrug at this stuff. A "Soccer Analytics" panel featured four experts essentially discussing how little had been done (an author did present one interesting study of English Premier League players that showed a correlation between "visual exploratory behaviors"—surveying the field, essentially—and performance). And although there were panels on the analytics of ticketing, gambling, injuries, fans, and mixed martial arts, there was not one on hockey.
Hockey and soccer pose unique challenges to statheads. The greater number of soccer players on the pitch and the larger field could make taking measurements easier than hockey's six-on-six in a small rink. But the frequent substitutions in hockey, versus the three substitutions permitted per game in soccer, make it easier to develop something like basketball's five-man-unit stat, which measures every five-player lineup by how many points they score minus how many they allow. For that reason, and because there are more scores in hockey, I'm inclined to give it the analytical edge. (But there's hardly universal agreement here. When I asked my original question of former Indianapolis Colts general manager Bill Polian—e.g., the guy who drafted Peyton Manning—he replied, "Hockey. Put that down as a guess.")
Much work has been done in football, both at outside sites like Football Outsiders and, much more quietly, within teams. But some of football's top analysts said that the incredible intricacy of football—the different types of scoring, the uncountable potential game situations, the byzantine mess of eleven men working in tandem, intangibly but undeniably dependent upon each other—means that only so much can ultimately be accomplished. "The sheer complexity of football makes it to my mind the hardest," said Brian Burke of Advanced NFL Stats.
Two people I spoke to, one from football and one from baseball, argued that football teams might be particularly stubborn about adopting analytics because of the number of people that must sign off on virtually anything. "There's so many people involved in decision-making," said Grantland football writer Bill Barnwell, who got his start at Football Outsiders. "You can get one person in the front office to buy in, one coach to buy in. But it's hard to get seven coaches, and the G.M., and the owner." Farhan Zaidi, Oakland Athletics director of baseball operations, said, "In soccer and football, the on-field manager has more influence"—and, therefore, stats have less room to thrive. The exception, he noted, is when the on-field coach and front-office manager are the same person, the most prominent example of which is the New England Patriots' Bill Belichick. Not coincidentally, the Pats would make anyone's list of the top five teams most interested in analytics.
That leaves baseball, with its 30 teams, 162 games per season per team, more than 30 at-bats per game, relatively few potential game situations, scoring in increments of one—in sum, a very long series of discreet events that mostly involve a grand total of two players (pitcher and hitter; arguably, also the catcher) and can't possibly involve more than eight or nine. It is something between a stathead's dream and a parody of a stathead's dream. There's a reason Bill James did not write about football, and Billy Beane did not general-manage a soccer team (although now, as it happens, he would probably like to).
But Albert Larcada, an ESPN analytics specialist, contended that baseball was in fact the least amenable to analytics—as of now, that is. "In soccer, no one's doing it," he told me following the "Soccer Analytics" panel (on which he appeared), "so if someone wanted to spend $100,000 on a guy, they'd make a dent. In baseball, all 30 teams are doing it—there's no dent." It was a refreshing reminder of the entire point of the Sloan conference, and of sports analytics in general: "It's about winning games," as former NBA coach Stan Van Gundy put it at another panel. "The goal is to win."
Amid the tsunami of cool numbers and counterintuitive insights, that fact is frequently overlooked. Keeping victory in mind is crucial to using analytics effectively in sports, especially as we remember that there has only been one championship team that could fairly be called anachronistically stat-obsessed: the 2004 Boston Red Sox, which also happened to have the league's second-highest payroll. The team that practically introduced analytics to Major League Baseball, the Oakland A's, consistently has one of the lowest payrolls—and hasn't won a World Series since the '80s. The Houston Rockets, one of the NBA's early adopters of analytics, haven't made the playoffs in three seasons.
Moreover, seeing how analytics perform in a field as black and white as sports, where "the goal is to win," is useful to understanding the relevance of analytics elsewhere. Viewing things through a prism that is relentlessly quantitative and single-mindedly obsessed with outcomes is probably not the best way to operate in fields where the goal is not, or not only, to win. Which happens to cover just about every field that isn't sports.
The introduction of younger owners who made their money in fields like finance or tech—like Cuban, Stuart Sternberg of the Tampa Bay Rays, and Joe Lacob of the Golden State Warriors—has been crucial to analytics' increasing acceptance.
Cuban's answer: "I have no idea. I just pay attention to basketball and nothing else."
The technology is derived from an Israeli missile-defense firm, and was also used to create those infamous CNN holograms.