“The world of A.D. 2014 will have few routine jobs that cannot be done better by some machine than by any human being. Mankind will therefore have become largely a race of machine tenders."—Isaac Asimov, 1964
There’s a staggering amount of information an NBA coach would have to hold in his head to do his job optimally, a volume that—given the rapidly increasing sophistication of the tools that produce it—likely grew in the time it took you to read this sentence.
Consider this: In 2014, 25 times a second, in every NBA arena, the location of the 10 players on the floor, as well as the ball and the officials, are captured by a network of SportVU motion-tracking cameras, then translated into facts about the game. Additional technologies are poised to add layers of detail to this already rich picture.
The sheer number of concerns, calculations and considerations—in other words, opportunities for a competitive edge—presented by this data is overwhelming. The average person can hold seven discrete bits of information in his or her head at a time; someone with a genius-level working memory and ample training can manage up to 80. The iPhone in your pocket right now can hold trillions.
And so as the game becomes more data-centric, a strange possibility is creeping into view: Like mortgage appraising, high-frequency trading and assessing credit risks, in-game management of a professional basketball team is a job that may no longer be fit for a human.
Which suggests the question: Who, or what, is it fit for?
Rise of the Machines
On the surface, the brave new world of coaching will probably look a lot like the present. Picture this: a coach, say Erik Spoelstra, standing on the sidelines, dictating direction to his players, making substitutions and barking orders, flop sweat accumulating on his forehead—to the naked eye, coaching as its always been practiced—but lean in a bit.
Periodically, maybe constantly, Spoelstra glances down at his wrist to consult an electronic device. The tablet is running an app that analyzes, in real time, the reams of data the game is producing. The program, written by a coterie of medical experts, analytics heads and software engineers the Heat employ, gives the coach direction on—maybe tells him outright—what calls to make.
Player X is tired, running at a speed 17 percent slower than usual and should be removed; he’s at an elevated risk of injury. Player Y is not being aggressive enough on help defense and needs to be chastised. The team’s shot selection has been dismal, costing it eight expected points through three quarters; call a timeout to address this.
A group of assistants monitors the same information and keeps an eye peeled for alerts that might slip the notice of the coach. While he still holds veto power—the decision is, ultimately, his to make—Spoelstra has become in many ways marginalized; his work outsourced to some algorithm.
This strange future may not be too far off. Tracking cameras are mounted in all 29 arenas and feed teams data during games, while firms like Catapult Sports—which uses GPS sensors to track player movement with granular precision—promise to complement SportVU with tools for monitoring players’ biometric information, such as heart rate. (At least five teams wear Catapult’s OptimEye sensors during practices, but the NBA does not yet allow its use during games.)
The final leap that’s required is the creation of systems for processing this information in real time and turning it into immediate strategy. While teams are incredibly tight-lipped about how they’re using SportVU and similar data sources (in the age of Wikleaks, NBA intel might actually be protected more closely than state secrets), the league’s more forward-thinking front offices are currently locked in a quiet arms race to do just that.
“[SportVU] is going to have a big impact, and the scary thing is, we don’t know how. It’s too early. I just hope I figure it out before everybody else does,” Memphis executive John Hollinger told The Washington Post in the preseason, capturing the urgency of the chase.
If this sounds implausible, that a piece of software would or could be trusted to make the tactical decisions now left to the Gregg Popovich’s of the world—and improve upon them—consider what happens when human chess masters square off against super computers. As Garry Kasparov can tell you, the humans, brilliant they may be, don’t come out of it well.
In many ways, the NBA’s Kasparovs have already lost the match. In numbers-savvy organizations, schematic decisions are now determined more by mathsketball than the expert intuition of head coaches. Take the three-pointer: The shot is en vogue (teams are shooting more of them than at any point in league history) not because coaches suddenly warmed to it, but because statisticians did.
Up in Toronto, as Zach Lowe laid out expertly last March, the Raptors have written a program that calculates, then translates to video, how the defense should be positioned relative to the offense at all times given the statistical tendencies of the players involved. Meanwhile, in Portland, the Blazers curl up with iPads during stints on the bench to watch video of that night’s game and make corrections. The trend is clear: Look ma, no coach.
Portland head coach Terry Stotts, while a vocal supporter of his team’s creative use of video, understandably views the encroachment of analytics—“which used to be called statistics,” he deadpanned—with a wary eye.
“I’m a big proponent of information,” the sixth-year head coach said. “But…I’m also very leery of information overload. I don’t like to force a lot of information on players, but I like to make that information available to them.”
Stotts suggested that analytics, while helpful, are also of limited utility to coaches. There’s a limit, he said, to what data can tell us.
“The one thing about analytics is they’re cold numbers, and they don’t take into account the dynamics of the team,” Stotts said, citing chemistry, locker room relationships and practice as aspects of the game the quants overlook. “To make decisions strictly on raw numbers would be very difficult.”
Well, maybe not for much longer. According to Gary McCoy, a sports scientist with Catapult, the shift away from man and toward data-hungry machine has long since shaken up other leagues around the globe. McCoy said that, in every sport he’s studied, once the “raw numbers” demonstrated they could determine practice schedules and playing rotations better than the coaches could, that’s precisely what happened. The coaches were asked to move aside.
McCoy pointed to Australia, where Aussie rules football has been leaning on Catapult data to figure out who to play and when for the past seven years—a period in which the league has seen a 30 percent drop in soft tissue injuries. The big surprise, McCoy said, isn’t that the NBA is coming around to this sort of analysis, but that it’s taken so long.
“The NBA is very far behind the rest of the [sports leagues in the] world when it comes to the application of data,” he said from Toronto, where he was on business with the Raptors.
This is unfortunate, McCoy went on, because the NBA is uniquely positioned to benefit from it. With 82 regular-season games, plus erratic practice and travel schedules, player fitness alone is far too complex of an equation to be solved by people. When asked to explain the association’s Luddite impulse, McCoy said that culture can be a hard thing to change: In athletics in general, and North American team sports in particular, it can be difficult to convince decision-makers to abandon the old for the new. Even, maybe especially, when the stakes are high.
“A lot of people look at technology as a threat,” McCoy added. “Those who are good will use it as an opportunity to get better.”
What’s Left for Coaches? Coaching
The transfer of power away from the bench appears to be well underway, and may have been expedited, in part, by a controversial notion that’s gained traction in the last several years: Head coaches never really did much to begin with.
The argument goes something like this: NBA players are relatively fixed quantities—their production is, compared to athletes in other sports, remarkably consistent from season to season—and so the most a coach can do to improve his team’s chances is play the right guys. In 2009, sports economist and Wages of Wins author Dave Berri published a study suggesting just this: The vast majority of NBA head coaches don’t have a statistically significant impact on the teams they helm. (Phil Jackson and Gregg Popovich were predictable exceptions.)
While it’s unlikely a team executive would publicly endorse this unconventional wisdom, the hiring and firing decisions of NBA front offices have come to reflect it. Of the seven teams that won 50 or more games in 2012-13, three made an offseason coaching change—four if we include the 49-win Brooklyn Nets. Front offices, it seems, have started behaving as though there’s not much of a relationship between coaching and wins. Or at least coaching as it’s currently practiced.
So a new set of skills will be elevated, if they haven’t already. “[The rise of analytics] puts a premium on computer programmers,” prescient Heat forward Shane Battier said in an interview with NBA.com. “It’s a premium on guys, engineers, computer scientists, guys who can evaluate data and make it adjustable for scouts and make it adjustable for directors of personnel.”
Which isn't to imply coaches will become extinct. Smart basketball men won’t find themselves in unemployment lines because they can’t write code. Their jobs will simply change. While, if trends persist, in-game decisions will soon be best left mostly to algorithms, there are aspects of coaching that computers are ill-equipped to carry out: managing personalities, massaging egos and, given the youth of many NBA rosters, chaperoning players in strange new cities.
There’s also basic teaching. Schemes that are dreamed up by artificial intelligence still need to be taught to people, by people. Practices have to be overseen. Motivational inducements delivered. Coaches will still, in other words, coach.
Consider the Philadelphia 76ers. Even in an organization that, under general manager Sam Hinkie, is unusually data-driven, the unlikely success of point guard Michael Carter-Williams is widely attributed to the influence and steady hand of assistant coach Lloyd Pierce, who works closely with the rookie in practice. Likewise, the Sixers staff is in the process of painstakingly rebuilding Nerlens Noel’s jump shot in the hope of making the young big man at least a credible threat from the perimeter.
These are labor intensive endeavors. Human labor intensive. They require that men with hard-earned hardwood expertise watch, judge and gently nudge two much less experienced athletes toward the realization of their potential. Toward a kind of greatness. This is important, consequential work that no machine or computer program can carry out.
Not yet, anyway.