What does the ability to track how many rebound opportunities Amir Johnson gets in a game have to do with missile defence in Israel?
Well, nothing really. But the technology behind SportVU, the evolving data collection system used by exactly half (and counting) of the 30 teams in the NBA -- including the Toronto Raptors
-- was born out of that military application.
"I always caveat with everyone that we didn't steal any state secrets," jokes Brian Kopp, the VP of strategy and development for Stats LLC, the Chicago-based global sports statistical agency that purchased SportVU from its Israeli creators in 2008.
"The developers that created it came from the Israeli Defence Forces, working on optical image recognition technology and algorithms for the military, a lot of it for missile tracking and other things."
After the soldiers left the IDF, they decided to apply the technology to their favourite sport, soccer, with a system that tracked players on the pitch.
For years now, anyone who pays any attention to the statistical side of sports has had to at least acknowledge the presence of advanced stats -- be it Bill James or Billy Beane and the "sabermetrics" leaders of baseball, or John Hollinger and the other statisticians crunching the numbers behind the numbers of basketball.
And whether you like it or not, whether you are a purist or not, it's here to stay. Advanced numbers don't lie, and there's simply too much money invested in professional athletes for franchises to not do all due diligence possible in evaluating performance.
Yet before some of you dismiss sports stat "nerds" as stereotypical coke-bottle glasses types huddled around algorithms and reruns of The Big Bang Theory
, consider some of the minds and technology involved in the process these days: From former missile defence technologists developing the SportVU camera system to former U.S. Navy officers like Alex Rucker (the Raptors' stats guru) breaking down the data that SportVU spits out, this is more than just math club stuff.
How SportVU works
That said, the actual images that the SportVU cameras currently produce do look like something out of a '90s-era video game. However the story is not the images themselves. The story comes from the stream of information the cameras produce -- six of them positioned up in the rafters of various NBA arenas.
"Basically, there's triangulation on either side of the court," says Kopp of the setup that places three cameras at each end of the floor, with exact placing based on the unique catwalk design of each arena. "We try to get two that are overhead, looking over the basket, and the others spread out."
The raw system is computer vision cameras, each weighing less than a pound, that not only collect video but also data. The algorithm behind the system is telling the camera what to track and what not to track, based on light, movement, colour and a number of different factors built into the system. The camera assigns each player x and y coordinates and tracks his centre of gravity.
As the game unfolds, all of this player data is fed into a workstation at the arena and sent in a constant stream to Stats LLC. Then the information is linked to the wire feed of the game's play-by-play, giving them an optical feed to match up with the NBA's game informational feed.
"It's not as simple as just finding the timestamps and snapping them together," points out Kopp. "The best example is when there's a missed shot, the guy at the scorer's table puts in 'missed shot' [but] that's not the actual point in time the shot was missed. Was it rebounded first? It could be several seconds off. So we have to go back and autocorrect."
Ultimately, they match the play-by-play feed to the exact frame in which the sequence occurred. It adds an entirely new context to statistics.
"Within that frame, we assign a data point," says Kopp. "So every single frame of the game, we have a data point.
"So, for instance, here's DeMar DeRozan's missed shot, we know where he was, we know how he got the ball, who passed it to him, and we know where everybody else on the court was when he had the ball. Once you've done that, there are a lot of queries and programs and reports you can write off the back of that data."
There most certainly are.
One of the biggest successes SportVU has had with teams is tracking player speed, as well as performance when a player is coming off an injury.
For instance: "Amir Johnson has good speed," Rucker told me last week. "He sprained his ankle in Memphis earlier this season. But coming back from the injury, you could definitely see [with what SportVU highlighted] how his velocity had declined."
Kopp elaborates: "Certain players run around a lot, but they're not doing a lot of cutting," he explains. "Whereas other guys are doing a lot cutting, going from slow to fast ... and when you think about it, when you have to go from slow to fast, there's a lot of load on your legs. And that's where, for some of these teams, especially for guys coming back from ACL injuries, that's the kind of data you want to look at."
As Kopp told Grantland.com's Zach Lowe earlier this month, the system has also unearthed a bunch of those I-should-have-thought-about-that tidbits. Like Kevin Love being the hardest runner in the NBA. Or the fact that centres sprint more than any other position on the floor, because they are generally trailing a play offensively after planting themselves down low on defence. Or better yet, New York Knicks players are registering an unconscious 61 per cent effective field goal percentage over the last two seasons off direct passes from Carmelo Anthony.
"It's the future of data collection," says Kopp.
All of that information is summarized from the participating teams around the league and the files are delivered to people like Rucker the next morning. Rucker says working all of this new-age data into the day-to-day practice and execution of basketball has come along with time.
"The challenge we have is you're dealing with [coaching staffs] who have done this for 25 years, so to try and come from left field with this stuff can be a challenge," he said. "Yet with a pretty high level of specificity, we can get good data that gives our coaches an edge ... It's much more important to have an impact than to be right."
While the bulk of Rucker's daily in-depth player data doesn't serve a needed real-time purpose, with information coming in every 60-90 seconds, there are stats from SportVU cameras that can be turned around and released for public consumption alongside traditional stats.
"For instance, what is our shooting percentage in the first half based on what the shot clock is," Kopp offers as an example.
"The only constraints at the moment are that the people letting us in the arena are [the teams]. We had talked about doing it in Toronto, about getting some of this information up at the arena, but with the ownership changes with MLSE it kind of slowed things down. But we have put up some unique stats on scoreboards -- who's run most in the first half, et cetera."
Thanks in part to the large amount of overlap with NBA and NHL buildings in the same city, the Toronto situation may change soon if SportVU gets into the hockey world as Kopp expects.
"We were hoping to start this year, but because of the lockout we didn't. We did some initial testing, and the players themselves aren't very hard to track. There's a lot you can do with player movement and positioning, and I know hockey is at least opening their eyes to analytics," Kopp said. "I know there's been this thought, 'Well how can you analyze it when it's so fluid?'
"But if you can actually measure the fluidity, then you're getting somewhere. Of all the sports that are going to be next, I think hockey would be up there."
Back to accessibility links