Friday, March 1, 2013
Curry Redux: SportVU cameras dissect an MSG shooting exhibition
NORTHBROOK, Ill. (STATS) -- Over 19,000 fans were in attendance Wednesday night at Madison Square Garden, and they saw one of the best shooting performances ever witnessed at the world's most famous arena.
So did six SportVU cameras.
And while opinions no doubt varied among the crowd as to where Stephen Curry's offensive exhibition stood in the annals of Garden history, the data the SportVU system was instantaneously tabulating lacked any element of subjectivity.
For those unfamiliar with NBA teams' version of next-generation statistics, a quick primer: the technological guts of STATS' SportVU product were originated in Israeli military optical tracking initiatives before finding an unlikely but intriguing fit -- with a few modifications, of course -- in advanced sports analytics.
Half of the NBA's 30 clubs now subscribe to the service, which includes a half-dozen cameras strategically placed above the court, monitoring all the action in three dimensions. Instantaneously, they stream a matrix of data to a dedicated computer just waiting to be deciphered by a front-office analyst.
Moneyball, meet basketball.
About 200 miles north of the Garden, forward-thinking industry professionals will come together at the MIT Sloan Sports Analytics Conference starting Friday to discuss breakthroughs like this and other innovations which will soon be pushing the business of sports information forward.
Sure, you know Curry scored an NBA season-high 54 points and went 11 of 13 from 3-point range in Golden State's 109-105 loss at New York. That's the stuff of any garden variety boxscore. But did you know he ran 3.36 miles over the course of his 48 minutes? How about the fact that he took 486 dribbles or made 54 passes -- seven of which resulted in field goals.
The Warriors had only seven other assists as a team, meaning not only was Curry doing all the scoring, but he was doing mostly everything on his own -- including creating his own shots.
"He put on a clinic. Knocked down shots. Made plays. Carried us. Led us in rebounding. He did it all," said Warriors coach Mark Jackson, himself no stranger to Garden history. "I've seen a lot of great performances in this building and his goes up there. ... That shooting performance was a thing of beauty."
While mileage and dribble metrics are certainly interesting in and of themselves, they lack context in a vacuum and beg follow-up. For one, how far is 3.36 miles relative to other players?
Thus begins the neverending quest for answers in a seemingly infinite pool of data, one query leading to another, the next to another, and so on.
For the record, Curry's trek of 3.36 miles would, in fact, be considered far. He significantly outpaced every other player on the floor. The next-closest was Knicks big man Tyson Chandler, who logged 2.80 miles -- 20 percent less than Curry.
Given that Chandler generally has to go from baseline to baseline while Curry was seemingly doing the bulk of his work between the arcs all night shows just how hard Curry had to labor -- and is a testament to his stamina that he still had the legs to be a deadeye marksman from outside 23 feet, 9 inches, throughout the game.
In fact, Wednesday night's contest proved to be a matchup of two of the league's best long-range shooters, and while there's certainly no confusing the respective all-around games of Curry -- second in the league in 3-point percentage at 46.0 -- and New York's Steve Novak -- fifth at 44.3 -- SportVU can in some ways quantify the difference.
Novak is the definition of a spot-up specialist, reliant on others to break down the defense, while Curry is just as comfortable taking someone off the dribble as he is draining an open jumper. In nine minutes, Novak had 12 touches, taking just four dribbles. Even with the difference in playing time, that's in stark contrast to Curry's 93 touches and 486 dribbles, making his season-long 3-point percentage all the more impressive given the heavy ball-carrying load he takes on as a combo guard who likes to create.
But it's safe to say every NBA coach knows Curry is good. The real question is how to stop him.
Data from previous games can be used to devise strategies. For example, the numbers show Novak is a blatant catch-and-shooter with little threat of putting the ball on the floor. Keeping a body aggressively on top of him without hedging toward lane penetration or doubling anyone else should totally neutralize him.
Granted, that's a relatively obvious assessment; there are hundreds if not thousands of other tracking observations that are significantly more subtle for every player in the league. Unfortunately for the Knicks, the data only confirms that -- for one night anyway -- Curry was somewhat indefensible.
Make him work for a long-range shot? Curry was still 4 of 5 from behind the arc when taking five or more dribbles. He was 5 of 6 when taking none.
Crowd him and try to take away the jumper? Curry drove a game-high seven times -- New York only had 10 drives total -- leading to eight points for himself and nine team points overall.
The best advice coach Mike Woodson could have given may have been the most obvious: stay close and don't lose him. While Curry hit an array of highlight-level shots with players in his face, he went 12 of 15 from the field when the nearest defender was four feet or more away, taking full advantage of lazy transition defense and incredible range.
Curry's performance was clearly the headliner on Broadway this week, but an interesting tidbit lost in the shuffle was Chandler's 28 rebounds. In games tracked by SportVU cameras, he had a league-high 36 rebound chances -- defined as a potential rebound coming within 3½ feet of a player -- and was able to grab 78 percent of them.
But that feat served merely as a footnote to a shooting performance for the ages.
"He's a special young player with a very unique talent," Chandler said of Curry. "We ran everything at him. He just got hot. There were some shots that he couldn't have seen the rim."
That's one area that SportVU doesn't have conclusive data on. Not yet, anyway.