Player tracking transforming NBA analytics
You ever wonder what NBA assistant coaches are scribbling on their clipboards during games?
Theyגre taking down stats. But not the stats youגll read in any box score. Theyגre logging numbers like touches in the paint, passes per possession, three-pointers off kick-out passes, secondary assists, fouls drawn ג- information central to a gameגs outcome but not found anywhere near a traditional box score.
This is how itגs been for years.
But the stats being tracked by these blazer-wearing NBA lifers -- both during the game and in the film room -- are nothing compared to whatגs being done by tiny cameras in the rafters of a number of NBA arenas.
Those cameras are part of a system called SportVU, and it has the potential to change everything we know about analyzing NBA basketball.
גThis is everything weגve been charting, all-encompassing, and so much more, and itגs all sortable,ג says one Eastern Conference executive. גThis isnגt something I ever thought possible.ג
It was used by ten teams this season -- up from six last year and four in 2009-10 -- and with a third of the league now using SportVU and sharing data with each other, we can begin to draw conclusions about areas of the game previously left up to conventional wisdom.
SportVU tells us, with relative certainty, which player has the fastest top speed in the NBA. It tells us not who scores the most, but who scores the most per touch. It tells us who dribbles the most per game, and who dribbles the most compared to how many shots they take. And thatגs just the surface.
If youגre wondering, the leaders in those stats -- and many more -- are dispersed throughout the words below.
Some of this dataגs useful. Some of it isnגt. Some confirms conventional wisdom, some challenges conventional wisdom. But itגs all fascinating.HOW IT WORKS
If youגre wondering about how serious this system is, know this: This is technology that was originally made for military use. SportVU was created in 2005 by Israeli scientist Miky Tamir, whose background is in missile tracking and advanced optical recognition. He used some of that same science to track soccer matches in Israel, spitting out similar fitness and movement stats now being tracked in the NBA.
The American company STATS purchased SportVU in 2008 and turned its focus to basketball. Soccerגs a second-tier sport in America, baseball was already in a statistical revolution of its own and football had too many players and not enough flow (although the leaders at SportVU havenגt ruled out pursuing the NFL).
With constant movement, a controlled environment and only 11 moving parts, NBA basketball was perfect.
So STATS, led by vice president of strategy and development Brian Kopp, sought out the most tech-savvy NBA teams for a test run during the 2009-10 season. He convinced the Dallas Mavericks, Houston Rockets, Oklahoma City Thunder and San Antonio Spurs to be the systemsג guinea pigs. The next year, the Boston Celtics and Golden State Warriors jumped in.
This year, ten teamsג arenas were fitted with SportVU: Boston, Golden State, Houston, Milwaukee, Minnesota, New York, Oklahoma City, San Antonio, Toronto and Washington. And that number is expected to grow into the teens next year.How does it work?
There are six computer vision cameras set up along the catwalk of the arena -- three per half court. These cameras are synched with complex algorithms extracting x, y and z positioning data for all objects on the court, capturing 25 pictures per second.
Each picture is time-stamped and automatically processed by a computer, which connects the data to the play-by-play feed and delivers a report within 90 seconds of a play. This is the part of the process the STATS people are so proud of ג the proprietary algorithms in the software, which they call the ICE Platform.
Almost instantly, coaches and stat guys have this information at their disposal on their computer or iPad.
They donגt always know what to do with the information -- yet -- but they have it.
ON THE REBOUNDWHAT IT DOES
Not all players get the same number of rebounding chances, so total rebounding numbers can be misleading. Here are the NBA leaders this season in rebounding percentage, defined as percentage of rebounds gathered when within 3.5 feet of the ball.
Player Rebounds Chances Rebounding percentage
Kevin Durant, Thunder 308 422 73%
Carlos Delfino, Bucks 119 170 70%
Kevin Garnett, Celtics 297 450 66%
Chase Budinger, Rockets 124 189 66%
Marcus Camby, Rockets 157 240 65%
Think of a stat within the boundaries of a game. Seriously. Any stat. Doc Rivers tried. He asked the Celticsג stat guy, Mike Zarren, what the Celticsג offensive efficiency was when Rajon Rondo held the ball for more than five seconds on a possession. At the time, Zarren didnגt know. Now he knows.
SportVU tracks every player movement, every pass, every shot, every touch -- everything. At this point, it tracks more information than teams know what to do with. Every executive interviewed for this article agreed they werenגt even using 10 percent of the information this system could provide. And they all agreed this is the future of advanced basketball analytics.
Here are a couple facts we learned from SportVU from this season:
* Paul Pierce averaged 4.5 assists this season, which is pretty good for a scoring wing. But that number doesnגt tell the whole story. According to SportVU, Pierceגs teammates shot a higher percentage after his passes than any other player in the NBA. This shows Pierce is passing at the right time -- heגs giving his teammates mostly layups and open shots.
* Nikola Pekovicגs breakout season was largely helped by Ricky Rubio. Pekovic made 76 percent of his field goals off Rubio passes, compared to 56.4 percent overall.
* The NBA-wide shooting percentage is significantly higher when the shooter doesnגt take any dribbles. This confirms what any basketball observer suspected: ball movement equals offensive success.
How do teams use this information? Well, that depends on the team, depends who you ask and depends what theyגll tell you.
גA lot of this stuff, because itגs brand new, they donגt want other people to figure out how theyגre using it,ג Kopp says. גWe know theyגre using it, but it varies by team. Everyoneגs still figuring out the best way to use it.ג
Itגs competition. The executives interviewed for this story spoke slowly and cautiously, careful not to share anything with the teams trying to beat them.
A Western Conference executive said, frankly, גWe know what weגre doing with it, but we donגt want other teams to know what that is.ג
The potential benefits to coaches and observers is apparent, simply by taking a look at what the system can tell us about the basic areas of basketball.
Shooting: Shoulders square. Elbow in. Flick the wrist. Donגt force it. Thatגs what shooters have always heard from their coaches. With SportVU, coaches can track things like a playerגs shooting arc on made shots vs. missed shots, or shooting percentages off the dribble or off the pass (hint: itגs always better off the pass). They can learn Shane Battier should only shoot threes from the corner, while Dirk Nowitzki shoots better from the wing and top of the key.
Rebounding: Throughout the history of basketball, the players considered the best rebounders were the players who averaged the most rebounds per game. But that doesnגt tell the whole story. What if thereגs another elite rebounder on a playerגs team hogging all the rebounds? Or what if a guy plays for a bad defensive team that doesnגt produce as many missed shots? SportVU allows teams to deeply analyze rebounding by generating never-before-seen stats such as rebounding chances (described as when a player is within 3.5 feet of the ball and, yes, that measurement is exact) and rebounding in traffic (when opponents are within that 3.5 foot circle).
Passing: Itגs tough to analyze passers from team-to-team because two players can be in such different situations. Maybe Steve Nash was a better passer than Rajon Rondo in 2012 -- although thatגs like picking between the cheerleading captain and dance team captain -- but Rondo has better teammates who hit a higher percentage of their jumpshots, which leads to more Rondo assists. But SportVU provides statistics that give context to assists, such as total passes, secondary assists (if Derrick Rose passes to Joakim Noah who quickly finds a wide-open Carlos Boozer under the hoop for a layup, Rose would receive a secondary assist), and shooting percentage after a particular playerגs pass.
Athleticism: Perhaps the most revolutionary part of SportVU is its ability to measure playersג speed and leaping ability within an actual game. It can tell how quickly a player closes out on a shooter, or if heגs running full speed down the court. Pace is traditionally measured by the amount of a teamגs possessions, but now teams can find out literally how quickly theyגre getting the ball up the floor and learn exact time of possession stats.
Fitness: Body language usually tells coaches what they need to know about a playerגs fatigue level, but that can be misleading. These days Tim Duncan looks like heגs rushing to the front of a nursing home cafeteria line while getting back on defense, but heגs still putting up double-doubles and controlling the paint. With SportVU, teams can learn a playerגs average speed, so they can judge his fatigue level based on movement, not if there's color in his cheeks. And say a player is coming off a knee injury. Doctors could limit him to running two miles during a game, rather than giving a minute limit, which is a bit ambiguous in terms of how much stress is being put on a knee.
And, again, all of this is in the early stages.
Like when Harvard students were toying with home computers in the early 1980s, thereגs no telling where this player tracking technology might take basketball analytics.
SECOND IS SWEET
A great point guard will often set his teammates up for assists, but the point guard doesn't get credit in the box score because his teammate made the extra pass. Not anymore.
Player GP Secondary assists per game Total assists per game
Rajon Rondo, Celtics 32 1.3 12.4
Jose Calderon, Raptors 31 1.0 10.3
Steve Nash, Suns 11 1.5 10.2
Derrick Rose, Bulls 11 1.7 10.2
Ricky Rubio, T'Wolves 26 0.6 8.6
WHERE ITגS GOING
Many teams have turned Kopp down when he made his pitch.
Some teams donגt see a use for it right now. Other teams think itגs too expensive (SportVU costs גmid-five figures to low-six figures per year,ג Kopp says, or about $8.8 million less than what Lamar Odom made this year). One team, the Mavericks, stopped using SportVU this year to use their own system (which many believe tracks similar data to SportVU, but Dallas wonגt share).
But STATS is still meeting and working with the teams it doesnגt have.
גThe best thing for me to do is show them data of their own players, and theyגll ask, גHow do you have this?גג Kopp says. גBesides Dallas this year, there hasnגt been anybody that said no. Theyגve all said something like, גWeגre not ready yet.גג
The ultimate goal -- and wide belief -- is the NBA will purchase the technology and the data will be available to all teams and broadcast networks.
And that could lead to even more changes in how we view basketball.
For instance, the box score. Right now a box score isnגt of much use to NBA statheads. Besides the addition of plus-minus and shot attempts blocked, the NBA box score hasnגt changed for decades. And simple totals of points, assists, rebounds and the like donגt tell the whole story.
Thatגs why those assistant coaches are scribbling on their clipboards during games.
When this information becomes widely available, thatגs when things will really start changing from the box score to advanced scouting to MVP voting.
גWe need all 30 teams to get a true gauge on some of these statistics,ג says Matt Bollero, a basketball operations assistant who works with SportVU for the Timberwolves. גUntil we have 30 teams, we still think weגre in the 5-10 percent range of what we could be doing with this. Itגs a really neat system and itגs really going to change the game. But until we get to that point, itגs just a great thing to envision.
At the very least, itגs a lot more than what fits on an assistant coachגs notepad.
GOING THE DISTANCE
SportVU allows the first-ever tracking of how many miles players run over the course of a game. Here are the players who traveled the farthest per game this season.
Player GP Distance Distance per game
Luol Deng, Bulls 13 35.35 2.72
Rudy Gay, Grizzlies 12 31.63 2.64
Kevin Love, T'Wolves 32 83.23 2.60
Brandon Jennings, Bucks 41 103.68 2.53
David Lee, Warriors 35 88.41 2.53