Last Wednesday night, in a cozy room inside Harvard’s Winthrop House, a group of 20 students gathered around a table to talk about sports. Wearing sweatshirts and a few days of stubble, they scarfed down Subway sandwiches and foil-wrapped burritos, though with soft green walls and a framed portrait above the fireplace, the space looked more appropriate for afternoon tea. This was the weekly meeting of the Harvard Sports Analysis Collective, a student organization that describes itself as “dedicated to the quantitative analysis of sports strategy and management,” which is a modest way of saying that the people in this room aspire to one day preside over sports franchises.
They were sliding around a box of miniatured glazed donuts—the club spends almost all of its university funding on snacks—when David Roher, one of the group’s two presidents, called the meeting to order. He asked everyone to introduce themselves and, as an icebreaker, name their favorite sports books. He also added an unusual caveat.
“For the sake of variety,” Roher said, “it can’t be Moneyball.”
“How about the Moneyball screenplay?” whispered someone else.
Roher was referring to Michael Lewis’ 2003 book about the Oakland Athletics, an organization that used sabermetrics to become one of baseball’s best teams with one of the sport’s smallest payrolls. “It’s probably the most important book I’ve ever read,” he admitted afterward.
The book spawned a revolution of its own, lionizing a community of eggheads and enraging the scouts who believed firmly they had found baseball wisdom in the bottom of a spittoon. Other clubs adopted Oakland’s system of analytics, snapping up Ivy League graduates like Paul DePodesta, the brains of Oakland’s front office, a Harvard economics major. The Red Sox and their 30-year-old general manager Theo Epstein, a Yale man, won the World Series in 2004 just months after Larry Summers, then the president of Harvard, cited Moneyball in a speech about heath policy. “What’s true of baseball is actually true of a much wider range of human activity than has been the case before,” Summers said.
It then took a few years for Harvard to be littered with The Moneyball Generation: nerds whose thinking wasn’t just influenced but wholly shaped by the book. These students are the first who were young enough to encounter Moneyball as impressionable teenagers, meaning they’ve hardly experienced sports—all sports, and not just baseball—as anything other than a grand experiment in quantitative analysis.
Maybe they’re too late to infiltrate baseball’s front offices, considering the boom of sports-minded quants and fall of certain icons in the time they took to graduate high school. DePodesta lost his job as the Los Angeles Dodgers’ general manager, Billy Beane’s Athletics have made the playoffs only once since 2004, and even the Red Sox, for all their sabermetric might, wouldn’t have broken their 86-year championship drought without John Henry’s deep pockets. It’s even easier to pronounce Moneyball dead than it was to dismiss it in the first place.
Yet here, finally, are students riding a fresh wave of Moneyball. Some of them, between classes about robot systems and games of beer pong, have already been hired to crunch numbers for professional teams. Of all the long-lingering effects of Moneyball, intentional and otherwise, this might be the most profound. “In the abstract, sure, I thought there’d be lots of job openings and a change in sports,” Lewis said recently, unaware of the group at hand. “But in the concrete, I never really thought that there would be clubs at Harvard training people to work for the Oakland A’s.”
When he was a sophomore at Harvard, like most sophomores at Harvard, Rohit Acharya had no idea what to do with his major. It was 2006, and his concentration was applied mathematics, a cross between the theoretical and practical. He came back from Christmas vacation and his roommate handed over a copy of Moneyball that his mother had given him. It wasn’t long before Acharya was breezing his way through Oakland’s calculations. “It wasn’t that complicated,” he says.
This didn’t surprise him as much as the two passing references to someone named Carl Morris. Acharya had never heard of Morris, so he wouldn’t have known that this Harvard statistics professor had long doubled as an amateur baseball analyst. In a tucked-away corner of his office, hidden by a row of filing cabinets, Morris stores a small collection of baseball almanacs. Somewhere around here is Bill James’ first Baseball Abstract, purchased for $2 after Morris spotted a three-line advertisement in the back of a Sporting News issue. The forefather of the field even included a handwritten note with the book.
James was hired as a Red Sox consultant in 2003, around the time that Harvard Magazine profiled Morris, who had devised a model that showed the expected number of runs a baseball team would score in every game situation. The alumni magazine’s story about Markovian analysis somehow made it to the desk of Billy Beane, only a high-school graduate. He dismissed it with the all the ceremony he reserved for a used wad of tobacco.
“We knew this three years ago,” Beane says in the book, “and Harvard thinks it’s original.” (“Can I get my side of it?” Morris said recently. “It said in the paper that we were citing work that had been done 40 years ago!”)
Acharya saw this exchange in Moneyball and immediately emailed the statistics professor, hardly expecting a response. He wanted to discuss the possibility of starting a sports analytics club at Harvard. Actually, Morris replied, this was a wonderful idea. Acharya soon was a salesman pitching the club to anyone who would listen. There was one last obstacle: a name. One night, in the dining hall, Acharya and his buddies were kicking around a few possibilities and agreed to start with Harvard Sports Analysis. Someone suggested tacking on Collective. It stuck, for no other reason than it sounded good. Anyway, the first word was probably more important than the last.
About 10 students, tops, showed up to the early meetings, as HSAC focused on producing research papers. But these studies required time and manpower that exhausted the club’s resources, and the pace of academic publishing, which remains as glacial as a July baseball game, didn’t suit college students.
As the group evolved from a circle of wonks to a generalist’s club, mixing pre-professionals with those who just wanted to chat about sports, HSAC’s focus started to shift. It launched a blog, encouraging members to pursue quirky sports problems without spending an entire semester poring through data. This is where the club made a name for itself, with posts like, “Strikeouts and the Anna Karenina Principle, or: Why K’s Don’t Hurt MLB Batters” and “Momentum in College Basketball: Do Late Rallies Carry Over to Overtime?” (Another favorite: “No Less Worth Despite Their Girth,” a study of overweight college-football coaches.) The blog took off. These were provocative questions with engaging answers, and it didn’t matter that they came from undergraduates. These ideas—well, most of them—could help teams win more games and make more money.
In August, the first comment on a post about college basketball came from someone curious about a particular strategy. The commenter was Mark Cuban, the owner of the Dallas Mavericks. “We’re probably more well-known within the sports analytics community than we are within the Harvard community,” Roher says. He meant it as a deprecating remark before he reconsidered. “Actually, I hope we are.”
The club still convenes every Wednesday night in the same room in Winthrop House, where Acharya lived. At one recent meeting, an economics major invited criticism of her thesis about revenue sharing in Major League Baseball before the conversation moved to college basketball (simulating a mid-major through a power-conference schedule) and then professional football (adding an extra week in the NFL to increase revenue, decrease injuries and—voila!— stave off a lockout). It seemed there were simultaneous conversations, always, until an idea was so tantalizing that everyone else stopped to pay attention. The week before, the topic of this robust discussion was the basketball value of Jimmy Chitwood, the star of the movie “Hoosiers.”
HSAC now boasts about 30 regular members, including six undergraduate officials, a president emeritus and two graduate advisers, former presidents enrolled at Harvard Law School. Then there’s Morris, the interim chairman of the university’s statistics department, who attends as many meetings as he can. He’s the club’s longest-serving member.
One day last week, dressed in a tweed coat, Morris reclined in a desk chair in his sunny office. A pair of turtleshell glasses rested low on his nose, accenting his tufts of white hair. He was talking about how the ambitions of many HSAC members are largely the same, despite the inevitability of long hours, low pay and even less initial recognition.
“Some might be writers, some might be front office, some might be owners, some might be general managers,” Morris said. “Some of them will catch on, and my guess is, they’ll be a force.”
Last month, HSAC elected elected two presidents who will serve until January. They are John Ezekowitz and David Roher, two sports geeks who live in Cabot House, a 10-minute walk from Harvard Yard. For all their similarities—they are the leaders of a group obsessed with quantitative analysis in sports—they couldn’t be much more different.
Even as a sophomore, Ezekowitz comes off as professorial. Last week, he put on a blazer, a striped oxford shirt and khaki pants to meet his summer boss, formerly the assistant secretary of the Treasury, for dinner at Harvard Square’s swankiest restaurant. As a teenager, he was a champion Scrabble player and voraciously consumed sports blogs, another medium that came of age with him. He walked-on to Harvard’s varsity golf team for a year, and sometimes, on Saturdays, he wakes up at 7 a.m. to watch English soccer matches. Also, he doesn’t drink coffee.
Ezekowitz knows college basketball probably better than anyone in the club knows any other sport. He’s a total junkie. When his younger brother visited him this month, they caught four hoops games in one weekend, and his laptop was lagging recently because of a spreadsheet with the 35 relevant statistics for every college team since 2004.
This summer, Ezekowitz logged 120 hours manually building an extensive data set that he used in four blog posts that changed his life. This college-basketball study received attention far and wide, and not long afterward, the Phoenix Suns contacted Daniel Adler, a former HSAC president, looking for someone who could help with basketball analytics. Instead of crunching numbers for Harvard’s basketball team on a volunteer basis, Ezekowitz was soon hired as an intern, working out of a remote office 2,000 miles from Arizona. It’s his dorm room.
This is all unexpected, considering Ezekowitz wasn’t much interested in math as a high schooler. He only learned of HSAC when a girl on his hall told him about it a month into his freshman year. He was enjoying his introductory statistics course, and she was headed to a meeting that night, and—
“Wait! You didn’t like math?” interjected Roher.
Roher, a junior, is a baseball geek who might have even founded a club like this at another college. Unlike Billy Beane—or even DePodesta, a baseball and football player at Harvard—he never so much as suited up for Little League. He was born with Charcot-Marie-Tooth disease, a muscular disorder that leaves him with limited feeling and partial numbness in his extremities and prevented him from playing organized sports as a child. “All the energy that would’ve gone into being a mediocre high-school baseball player instead went to all this stuff,” he says. “There are a lot fewer people in sports analytics than there are trying to be athletes. Basically, for me, sports are the mathematical abstraction. If I can’t think about something in terms of baseball, I probably won’t be able to solve it.”
In the eighth grade, right after he made perfect sense out of Moneyball, he needed extra credit in a math class and approached his teacher with an idea: What if he established a system of baseball power rankings? The project enthused him, and Roher continued to dabble in sports as he took to computer science in high school. As a senior, he drove to Boston ostensibly to tour Harvard, but also to attend a conference called the New England Symposium on Statistics in Sports, where he spotted a sign for HSAC. It was the first he had heard of such a club. When he arrived on campus, in addition to showing up at HSAC meetings, Roher joined the lightweight crew team as a coxswain, making him one of the club’s only varsity athletes. He jokes that he signed up for the irony of it—“I’m a varsity athlete in the loosest sense of the term,” he says—even if he hasn’t found a way to quantify rowing.
“That would be a nice story, right?” he said recently, walking to an HSAC meeting in a zipped-up hooded sweatshirt. “It’s kind of like the robot that learned to love.”
He had just mentioned at dinner with the HSAC board that he wanted to take a very humble stab at reverse-engineering one component of Watson, the IBM machine competing on “Jeopardy!” The concept was certainly interesting, but no one else at the table was listening. While pitching the idea, Roher had let slip a secret. Like most people at the table, he didn’t exactly love mathematics. The club’s other officials basically choked on their spaghetti when they heard this.
“You’re going home tonight to reengineer Watson,” said Jake Fisher, HSAC’s president emeritus, suddenly leaning on the table. “For fun!”
It was true that Roher wasn’t placed in his high school’s advanced math track, and that he prefers the logical nature of numbers rather than the computations, but he backed down with a conciliatory shrug. That night, he stayed up until about 3 a.m., powered by a large, late cup of tea, and he published his initial thoughts on HSAC’s blog in the morning. “After hashing out the theory, my next step will be to code a Jeopardy simulator that can solve for the ideal buzz threshold in all situations, and perhaps solve for sports-related thresholds as well,” he wrote. “Provided I can get it to work, stay tuned.”
If you scour HSAC’s archives, you’ll find a clever piece by Roher about the effect of old age on college-football coaches. Using a statistic he created, he found that gimmicky coaches typically fizzle out within a decade. “But if there’s a coach who’s really good because he’s constantly innovating and coming up with good ideas,” Roher says, “he’ll get better as he gets older.”
In some ways, while these kids aren’t geriatric or qualified to coach college football, they’re staring down the same problem. They are the first students who have known their entire adult lives that they could work in a baseball team’s front office without having played professional baseball, so long as they possessed the same résumé as, say, an investment banker.
But the benefit of youth was also a curse. While they were consuming Moneyball as Gospel, the market became saturated with wannabe Paul DePodestas. The baseball blogger David Pinto, a Harvard graduate and HSAC visitor, put it another way. “The problem,” he says now, “is that it’s harder to get a GM job than it is to get into Harvard.”
This conundrum will be discussed in detail next weekend, for the fifth straight year, at the Sloan Sports Analytics Conference, hosted by the Massachusetts Institute of Technology. Most HSAC members will ride over to the Boston Convention Center on a university subsidy to watch the brightest minds in the business analyze metrics in every major sport. The ideas of Moneyball, in short, are no longer limited to baseball by any means.
In 2007, at the first conference, basketball was the only other sport represented, and it’s the game in which analytics have advanced most since then. Already the most progressive analysis originates from NBA front offices, still on the lookout for talent. Last week, the Houston Rockets advertised an opening for an intern with three qualifications: basic knowledge of basketball, experience with statistics and familiarity with programming. Jason Rosenfeld, a Harvard junior and former HSAC president, worked for the Shanghai Sharks on a gap year in China, and there’s Ezekowitz, who won’t discuss the particulars of his job with the Suns. This is partly because the community is competitive about finding and maximizing efficiencies, even if the business side of sports can be more collaborative.
“Three years ago, there were three teams in the NBA doing basketball analytics,” Ezekowitz says. “Now, there are 10 to 12 teams that have full-time stat guys and probably 18 to 20 that have an employee of some kind. It’s hard to imagine over the next three years that it won’t be 95 percent of the league. So let’s put it this way: As compared to baseball, basketball has only scratched the surface of what analytics can do, and the comparative advantage of finding something in basketball, finding something really good, is enormous.”
He went on to give an example.
“The box-score stats that capture defense are woefully insufficient: steals, blocks, turnovers forced, defensive rebounds. Even to say something like defensive plus-minus, which is pretty much as far as the field has gone—it’s nowhere near as good as offense, and nowhere near as good enough. If you were to come up with something that quantified defense better, even if it’s just 50 percent better, that can win an NBA team”—he paused to think about his phrasing—”well, enough games to make that team a lot of money.”
The basketball frontier, in other words, was broached before someone like Ezekowitz even graduated from high school. In four years, or as long as it takes someone to make it through Harvard, it will be swarming with the mathematically inclined. So what’s next?
“Soccer,” Ezekowitz said like someone who’s thought about it. “And after that is golf.” The challenge in soccer, like football and basketball, is the fluid nature of player interactions. Baseball, by contrast, is more mechanical; there are many scenarios, but only so many. (“Cricket is discrete!” Roher added.) Soccer is the beautiful game because it’s almost impossible to model the way a midfielder works his magic. How could such a sport be ripe for quantitative analysis? HSAC recently formed a partnership with an English company that coded hundreds of hours of college soccer into a massive database and turned it over to these Harvard kids, responsible for coding the regressions that will determine a player’s true value. These quantitative statistics in soccer are limited. It will be up to these students, in part, to invent them.
Last week, around 10 p.m. in a quiet hotel lobby, Ezekowitz and Roher couldn’t help but speculate about what the data would reveal. Neither student had parsed through the numbers—this was all very new, and they hadn’t planned to talk about it—but suddenly, without so much as a calculator, they were trading ideas back and forth. How many times, for example, did a midfielder send a cross into the box? That’s something a soccer manager might want to know. But then, how often did that midfielder have the opportunity to hit that cross? This statistic was more specific and probably more valuable. Well, did it matter if the midfielder received a crisp pass himself? The question was complicated. Perhaps it didn’t have an answer. “There are all sorts of factors that have to be quantified,” Ezekowitz said, and Roher nodded. They sounded like they could have kept talking long into the night.