How the NBA’s love for advanced stats has slowed our understanding of the game.
Photo by Jonathan Daniel/Getty Images
This weekend, many of the NBA’s sharpest minds will gather at the Sloan Sports Analytics Conference, an increasingly splashy affair held each year by MIT, and now sponsored by ESPN. They’ve come a long way. Just over a decade ago, several of the most prominent among them first gathered in a much humbler spot: a Yahoo Groups message board called “APBR Analysis,” for the Association of Professional Basketball Researchers. In the very first post, on Feb. 10, 2001 at 10:32 p.m., a former Cal Tech hoopster turned statistics Ph.D. named Dean Oliver laid out an ambitious agenda of 12 issues. “To start off the group, I think that it is most appropriate to identify some of the outstanding questions in basketball,” he wrote. Some questions were practical. “Does Hack-a-Shaq work?” “Why has Charlotte had such a good record without Derrick Coleman in the lineup and a mediocre one with him in?” Others were more theoretical. “What additional statistics could be taken to improve individual defensive evaluation?”
Oliver and his cohorts on the message board wrestled with these questions and countless others, logging on at all hours to debate the relative merits of Allen Iverson or how best to calculate a new metric called usage rate. Long before Moneyball author Michael Lewis wrote a New York Times Magazine cover story on the topic, the board wondered why Shane Battier had such a positive impact on his teams despite not appearing to be all that good at basketball. The message board was a veritable think tank. “You could tell that this was a place where there was going to be a serious level of discussion about NBA statistics,” says Kevin Pelton, who would become one of the original writers for Basketball Prospectus. “It was literally the only place in the world it was happening.”
The NBA establishment quickly took notice. Oliver, who published the seminal Basketball on Paper in 2003, seven months after Moneyball hit stores, was hired full time by the Seattle Supersonics in 2004. Another frequenter of the board, John Hollinger, was hired the following year by ESPN—and recently became a vice president of basketball operations for the Memphis Grizzlies. Hollinger’s ESPN gig was filled by Pelton, who, after making his name at Basketball Prospectus did a consulting stint with the Indiana Pacers’ front office. Roland Beech, who created the popular website 82games, was hired by the Dallas Mavericks in 2009 as director of basketball analytics. (His boss, Mark Cuban, is regularly one of the biggest names at the Sloan conference.)
As soon as each statistician joined an NBA squad, sharing in public became off-limits—and so, gradually, the think tank closed shop. What were the teams paying for, after all, if their new stat gurus were just posting their ideas online for the other 29 franchises to read? This has had a paradoxical result: Because NBA teams embraced advanced stats so quickly, progress on basketball analytics has actually slowed down. The top minds are now all working in silos, not only unable to collaborate but actually competing against each other.
Major League Baseball teams were hidebound enough to ignore Bill James and sabermetrics for a full quarter century—as a result, he and others hashed out ideas out in open, public forums. By the time MLB executives finally embraced advanced baseball statistics, the movement was fully formed. But advanced basketball stats were just getting started when NBA teams tuned in. And though many of those teams are now collecting the kind of data that outsiders can only dream of, they lack the manpower to fully harness it. Certainly there have been advances: Teams’ internal stats generally blow away what’s available publicly. But they haven’t come as fast as they otherwise might have. And we, as fans, don’t understand the game as well as we could.
Dean Oliver estimates that between 22 and 24 NBA teams currently employ some form of analytics, with about one-half that number seriously incorporating their findings into the team’s approach to the game. Most analytics departments are small, which makes it hard to tell when your research is headed down the wrong path, says Aaron Barzilai, a former MIT player who started the site BasketballValue before joining the Philadelphia 76ers in November as their Director of Basketball Analytics. “You often just don’t have a ton of feedback on how you’re doing, especially if you’re one person on a team by yourself,” he says. And asking for help isn’t an option. Desperate for any competitive advantage, NBA teams guard their data—and whatever conclusions they draw from it—with about the same paranoia as a government official sitting on bomb codes. When asked how many analysts he employs, Houston Rockets General Manager Daryl Morey, the first stats acolyte to be hired to run a franchise, replies, “It’s not something we talk about.”
Fans have lost out in the bargain, too, with the newest ideas mostly staying locked up inside team offices. Just glancing at the homepages of leading advanced stat sites makes it clear they’re not getting enough love—BasketballValue and 82games both look like they were designed by a 14-year-old sometime in 1998. Barzilai hasn’t updated the numbers on his site since the Sixers hired him, and while the stats 82games remain current, Beech stopped posting articles there when he joined the Mavericks. Hollinger’s analysis, too, has now disappeared behind the league’s veil. NBA.com just launched a much prettier stats portal, complete with advanced metrics—but what’s on the site pales in comparison to what’s available behind closed doors.