Will these cameras make box scores obsolete?
DEC 05, 2012 4:00a ET
For nearly every game at the beginning of last season, R.J. Adelman's handwritten notes were detailed, copious. They came not just from his time on the bench but from the hours he logged in his office after games watching video. Win or lose, the Minnesota Timberwolves' director of player personnel would relive it all again, rewinding, pausing, noting the tiniest details that no box score ever would reveal.
What Adelman, the son of Timberwolves head coach Rick Adelman, was doing was nothing new. Many a coach has instructed his assistants to manually track detailed statistics from the bench, keeping them at the ready to provide context for in-game huddles and discussions. For such coaches, stats are the key to improving their teams; they give a greater understanding about what's unfolding in front of them. They are more than just color. They're a necessity.
Now, though, Adelman and the Timberwolves have something the generation before never did: a system that will do it all for them. A system, in fact, that does it all and more, that computes things most coaches never dreamed of knowing, things some coaches still don't care to know.
Look up in the rafters of the Target Center. Up, up, up. You won't spot them, these six cameras barely bigger than the size of two fists. To see them involves a ride in a freight elevator and a walk out onto the precipice of the arena's highest level. You must look down, which is where the cameras' tiny eyes are trained, down onto the court where the players look no bigger than tiny Lego men. It's almost frightening — not only the height but also the scope of what these nondescript cameras can do from so far away.
Ever wondered how far Kevin Love runs during a game? How about how many points Nikola Pekovic scores per touch or how many times J.J. Barea dribbles the ball? Probably not, but that's only because you've been conditioned not to. Conventional basketball statistics tell us to look at scoring, at rebounding, maybe at assists or blocked shots; for many, the 12-stat box score is enough. For the more advanced, there are Association for Professional Basketball Research Metrics (APBRmetrics) stats such as player efficiency rating, true shooting percentage and rebounding rate, derived using complex mathematical equations.
Somewhere in between lie SportVU cameras and their data. It's complicated, yes, and wholly dependent upon detailed algorithms, but what the cameras in the Target Center's rafters reveal through player-tracking images captured 25 times per second is in simpler terms than any of basketball's advanced metrics. No one can see a player efficiency rating or a true shooting percentage on the court, but a contested rebound or a secondary assist — when Player A passes the ball to player B, who passes the ball to Player C, who scores — they're easy to visualize.
SportVU's statistics are a deeper version of the conventional, validating or cheapening simpler metrics. The cameras can reveal whether a player's rebounds are contested or simply by default. They show potential assists, good passes that precede missed shots, which otherwise would disappear into the basketball oblivion. They do all that and so, so much more.
The data exists in three categories: visual, which someone could collect on his own (dribbles, passes, touches); contextual, which augments already-existing data (secondary assists, rebounding chances); and physical (speed, distance, reaction time, closeout speed), which would be impossible to collect even from video. Adelman, who's the middleman between the staff working with the cameras and the coaches, says the information the Timberwolves find most valuable relates to players' individual strengths: productivity off the dribble, sweet spots on the court and hottest areas to score from.
Take one former Timberwolves player, who had a solid field-goal percentage when he took fewer dribbles. The more he dribbled, the more that percentage fell off, and the coaching staff was able to recognize and relay that information. The data can be motivational in that way, with the ability to correct for a less-than-obvious problem or even to back up coaches when they tell players to do certain things — such as dribble less, in this case.
It also can add context to a player's scoring. Pekovic ranked sixth in the league last season in points per touch, with 0.412. ( Kevin Durant led the league with 0.496.) Not only was the Timberwolves center scoring to the tune of 13.9 points per game and shooting 56.4 percent from the field; he also was sticking to his role, simply getting the ball in the paint and scoring.
NBA executive vice president of operations and technology Steve Hellmuth refers to much of what the cameras track as "hustle statistics," and nothing better epitomizes his term than what Kevin Love did last season. The power forward ranked third in the league behind Luol Deng and Rudy Gay in average distance run per game, at 2.6 miles. He also posted the 11th-longest single-game distance of any player last season in a double-overtime loss in Oklahoma City on March 23, logging 3.28 miles. That same night, Luke Ridnour ran 3.33 (eighth-longest on the season).
Those metrics are just a small fraction of what the cameras yield, and each of the 14 teams using SportVU this season — Houston, Oklahoma City, San Antonio, Boston, Golden State, Milwaukee, New York, Toronto, Washington, Minnesota, Dallas, Orlando, Phoenix and Cleveland — tailors the system to its individual needs. As the cameras catch on, the Timberwolves view the technology as a growing part of their operations that can keep them ahead of the curve. In the year and a half the system has been in place, they've learned how much growth and refinement it takes to stay on pace with what the cameras can do.
In March 2011, Timberwolves basketball operations assistant Matt Bollero attended MIT's Sloan Sports Analytics Conference. It's the Mecca of the sports-geek world, and it was a two-day affair for the first time in 2011, complete with a 300-person wait list on top of its 1,500 attendees.
While there, Bollero spoke with STATS LLC, which purchased SportVU in 2008. The cameras, which use missile-defense technology, had gained popularity in Israel for player tracking in soccer, and at the time of the conference, they'd been installed in some NBA arenas for nearly two seasons. The Spurs, Mavericks, Rockets, Thunder, Celtics and Warriors were already on board, and the advanced player tracking and statistics SportVU affords quietly were becoming some of the league's best-kept secrets.
Brian Kopp, STATS' vice president of strategy and development, explained the system to Bollero. He told Bollero how the cameras track players, the ball and referees on a three-dimensional graph, that they could provide the Timberwolves with deeper understanding of the game. Bollero was sold. He returned to Minneapolis and pitched the investment (teams that initially signed up in 2009 paid about $50,000 each) to Timberwolves president of basketball operations David Kahn. By April, Kahn had approved the idea, and the Timberwolves became one of 10 teams using the cameras for the 2011-12 season.
A year after SportVU first began mining data for the Timberwolves, Bollero still has the binder that contains his initial presentation to Kahn. He'll flip through it, revealing charts and information about Timberwolves players past and present. They're like his state secrets, and Bollero believes the process he helped bring about and those six little cameras are going to change the way teams and even fans look at the game.
The cameras spew out information in real time, transmitting their snapshots to a hard drive in each arena, which delivers data that's essentially gibberish. From there, many teams, including the Timberwolves, rely on STATS to synthesize the information and send them detailed reports. Some teams hire their own staffs to do so, but they're in the minority, Kopp said, and STATS' goal is to create a comprehensive system in which it can be the intermediary between the data and the teams. In addition, STATS provides some teams with a scouting tool that combines scouts' input, simpler statistics and camera data into one comprehensive database, further enriching the camera output.
If it sounds like a lot of information, you're right: It is. Too much, even, for some teams and coaching staffs. So much, in fact, that Bollero says right now, he thinks the Timberwolves (and most of the 13 other NBA teams using the cameras) are using just a fraction of the cameras' potential.
When they're getting 72,000 images per game, that's hardly shocking. This is reams of paper and a boatload of data. Even if a team employed a 100-person staff to work with the data, it still likely would have to limit its focus.
A typical night with SportVU at the Target Center begins when players are warming up. Two employees in the basketball press box man the camera system, diagramming the court so the cameras know their boundaries and programming players' numbers into the system. From there, they let the lenses do their work. Then, after the game, Bollero and his team do about an hour's worth of work with the data. They wait until the next morning to do more in-depth analysis of the information they get back from STATS.
From there, Bollero chooses what he wants to share with his point person from the coaching staff, R.J. Adelman. He'll put reports on Adelman's desk when he sees fit, and Adelman will relay what he deems important to whomever on the coaching staff should see it. It's a filtering process, from too much information to simply a lot to eventually just what's absolutely necessary to share.
"If we see something, see a trend that might be helpful, then we put it on (coaches' desks)," Adelman said. "We just keep monitoring and compiling information. After like 10 games, maybe we'll give them a report. And then in the offseason, we show them team trends and league trends. But on a game-to-game basis, we can't inundate them with too much stuff."
Timberwolves coach Rick Adelman has been receptive to the information, and he used some of last year's data this preseason to show the team how it needed to improve. Most players are aware of the cameras, but they don't know how complex the information is. Even coaches are still privy to just a tiny portion of what's being computed. The work lies with Bollero's team and R.J. Adelman, and it's up to them to make sure the cameras are getting coaches what they need.
That means constant communication with Kopp and others at STATS; Bollero emails the company about five or six times a day with everything from questions to ideas to new algorithms the Timberwolves propose be added to the system — which STATS collaborates with the team to execute. It's an experimental process, with the team putting forth an idea, STATS working to build it into the cameras and the team then seeing if it gets what it's looking for.
"We want this to be a whole new level of data that (teams) can integrate into their process of analyzing players," Kopp said. "In order to do that, we need to use language that they're familiar with. . . . It's not going to be us saying, 'Hey, this is the way you have to look at it.' It's very much a back-and-forth (relationship)."
With only 14 teams on board, SportVU's reach remains limited. It's most useful for teams to use only their own personnel data, though Bollero noted that if two teams with SportVU-equipped arenas met in the playoffs, the data might be relevant to game plan for opponents. That's because everything the cameras put out is shared among member teams, so if the Timberwolves wanted to check up on, say, Russell Westbrook's prime scoring spots on the court, they could do so easily.
Teams that have SportVU are perhaps the biggest proponents of the system spreading in order for them to gain more data, both on opponents and from their road games. That's the funny thing about the cameras: Teams are secretive about what they're learning, but at the same time, they want as many opponents as possible to have the cameras and thus access to the data.
"The coaching staffs and the general managers that are using STATS are certainly very happy with the product, and they're very . . . close-mouthed about exactly what they're getting out of it and what they're doing with it," Hellmuth said.
Bollero has a similar take. Teams know Minnesota has the cameras, he said, but they don't know what the Timberwolves are doing with them. Half the battle is keeping the ways they use the system secret, not letting opponents know the algorithms they've found particularly beneficial. But as fervently as Bollero guards his secrets, he's perhaps more passionate about the idea of more teams coming aboard. Both he and Kopp hope to see the cameras on a league-wide level someday, and in the tradeoff between secrecy and innovation, innovation will always win.
Now it's just a matter of convincing teams that innovation is the way to go, or even that this is innovation in the right direction. The NBA is embroiled in a battle between traditionalists and stat geeks, between the Daryl Moreys and the Doug Collinses. (Morey, the Rockets' GM, is a devotee of advanced stats and the chairperson of the Sloan Sports Analytics Conference; Collins, the 76ers coach, said in November that "he'd blow his brains out" if he had to deal with analytics, indicating that the only such information he needed was in his brain and his gut.) SportVU falls squarely in the realm of geekdom, but what people don't understand about the cameras is that they aren't necessarily the stuff of advanced mathematics and theoretical basketball. They capture what we see but struggle to quantify, making the game richer. As more and more young coaches and executives rise through the ranks, as owners from the business world infiltrate the NBA, these ideas will only become more popular. The league is moving in one direction, a direction that SportVU and STATS would seem to enhance and facilitate.
What's most fascinating about SportVU, though, is in what's not yet fully tapped into. Four teams — Phoenix, Dallas, New York and Washington — purchased the system mostly for its physical data. They're using the cameras in conjunction with their training staffs to monitor players' physical performances, tweaking workouts based on how strenuous games have been and gaining a better understanding of how best to use their players.
This data could shift the way coaches think about using their players. Instead of putting someone on a minutes limit, they could limit his miles. Instead of pulling a player when he looks tired, coaches could review how many sprints he's completed during the game, how many times he's reached his peak speed, how much he's run in relation to other games. It's a more nuanced way to look at personnel decisions, and it's in every way an improvement upon a vague minutes restriction.
"It could be the most impactful area of this entire project because it's never been able to be captured before," Kopp said. "One team has told me that if they can keep a guy on the court for one or two more games a year, this thing pays for itself."
In the future, the cameras also could help teams evaluate players returning from injuries. Imagine if the Timberwolves had more than 23 home games' worth of information about Ricky Rubio from last season; if they did, they could look at richer trends to better analyze where he stands when he returns. Or, if Portland had used the cameras during Brandon Roy's time there, the Timberwolves could have better been able to see the areas in which his game fell off as his knees declined and might have gotten a more complete idea of the extent to which his game had devolved.
We're a long way from any of that, 16 teams and several years, but it's tempting to imagine.
There's also another direction that the data can (and likely will) go: monetization. Bollero thinks that it will become a profit-driven enterprise, with STATS and the teams leveraging the information for sponsorships. STATS already wants to use camera output to enhance broadcasts, and Bollero said he can imagine teams putting SportVU information on screens in arenas during games with sponsors attached.
"I can see it on the scoreboard right now. 'Brought to you by Red Bull: How far has Kevin Love run so far?' " Bollero said, with a laugh but dead serious.
So where is this going? It's hard to say. STATS is in discussion with more teams to purchase camera systems, and there's hope that once the number of subscribers reaches a critical mass, the NBA will buy in. The league, of course, won't confirm its plans with the system, but it's a fan of what SportVU is doing and the information age it's ushering in.
Talk to Hellmuth about all this, and he's unquestionably intrigued. Technology is his thing, though, and SportVU is just one tiny part of it. When Hellmuth gets going, he's talking about putting faces on blobs, heart rate monitors and motion-capture suits, and suddenly the future of the NBA sounds a lot like the filming of "Avatar." He has this habit of punctuating his sentences with "Right?" Motion-capture suits, right? Missing data pieces, right? Accelerometers, right? Right, right, right, you find yourself saying, with only a vague clue of what you're agreeing to and a growing sense that all this SportVU stuff is just a tiny part of some big, scary future.
It's big, yes, but it doesn't have to be scary. This is just basketball. These are just men in uniforms, no matter how freakishly athletic they might be. And these cameras, they're just telling teams more about their personnel, how to better use and motivate them. They're not forcing coaches to act one way or do things they don't want to do, and the information can be tailored to fit different philosophies.
Several years ago, Hellmuth attended a conference at which Celtics coach Doc Rivers spoke. During the talk, someone in the crowd asked Rivers about the most important piece of his coaching philosophy; Rivers responded not with some statistic or tenet, like great offense or air-tight defense, but with a bigger-picture philosophy.
Buy-in from players, he said. Commitment to a system.
And Hellmuth, the man who makes his living off statistics and innovation and all the things that are so much more complex than simple relationships, could not have agreed more. That's what all this is about, the cameras and the scouting databases and the advanced metrics. They're about buy-in, about fostering a situation in which teams' complete information leads players to improve and believe.
Information, as complex and detailed as it might be, facilitates Rivers' buy-in. It might be the key to it, in fact. Who'd have thought something so personal, so based upon relationships and the fundamental-but-abstract core of sport, could flow from those tiny, unfeeling cameras?
Follow Joan Niesen on Twitter.