Posted by: Andy Larsen on March 6th, 2011The author's views are entirely his or her own and may not reflect the views of the Utah Jazz.
This weekend, I attended the 2011 MIT Sloan Sports Analytics Conference in Boston. The conference is designed for geeks like me, who are interested in learning about sports analytically. However, it wasn’t just geeks who showed up. Sportswriters like ESPN’s Bill Simmons, Ric Bucher and Marc Stein made an appearance. Owners came too, such as Mark Cuban, Jonathan Kraft, and Joe Lacob (the new owner of the Warriors). General managers came, like Daryl Morey of the Houston Rockets, or Mike Zarren, assistant general manager of the Celtics. It was a remarkable collection of remarkably intelligent people.
The conference was formatted such that there were up to 5 panels and speakers occurring at the same time in different rooms. As a result, I only know of some of what the conference had to offer. Luckily, the conference will post video of every session on their website (http://www.sloansportsconference.com).
This is the first in a series of posts on the conference.
The “Holy Grail” of Basketball Data
Basketball is difficult to analyze because there aren’t easy ways of keeping track of the events that happened on the floor. For example, while it’s easy to see who to give an assist to, it’s more difficult to determine who should get credit for a play’s success: was it an easy entry pass, or a drive-and-dish situation? We don’t know with our current amount of data.
STATS, LLC, is trying to fix this problem with Optical Tracking Technology. Basically, they place cameras in NBA arenas in order to track where the players, referees, and the ball is at every 1/25 of a second. Then, they record the X and Y values for each of those objects in a massive table: each game has about a million records for analysis. Each of these records consists of the time, and the location of everyone on the court. Currently, these cameras are in only 3 NBA arenas, and only 6 NBA teams are paying for the data.
What statistician Sandy Weil has done is interpreted that data. His program breaks that data up into events that we can understand: a pass, a shot, a rebound, and so on. For this study, Weil chose to answer one question:
What affects shot percentages?
He found that really, only 4 characteristics matter.
- How good of a shooter is the player?
Weil found that a player’s TS%(True Shooting percentage) was a remarkably good indicator of how likely a player was to make a point-blank jumper.
- How close is the shot?
Weil found that, adjusting for all other factors, the field goal percentage of a given shot goes down by 1% for every 1.5 feet further away it’s taken from.
- Is the shot guarded?
A tightly guarded shot (with a defender within 3 feet) is about 12 percentage points less likely to go in than a loosely defended shot. It also matters how close a second defender is.
- Did a pass immediately proceed the shot?
Interestingly, Weil found that a shot was about 8% more likely to go in if a pass immediately proceeded it, regardless of how close a defender was. Therefore, there may be something to the idea of “shooting in rhythm”.
This becomes really interesting when you combine aspects of his findings together. For example: a 24 foot 3 pointer which is tightly contested should give you approximately the same number of points a loosely contested(defender within 3-5 feet) 13 footer would. The Jazz, believe it or not, should probably take more guarded 3s (though, of course, open 3s are even better).
Weil also studied stepback jumpers. If you gain 12% by making yourself open, and you only lose 1 percent for jumping back 1.5 feet, shouldn’t players take stepback jumpers more often? Well, no. The problem with this idea is that it forgets that the defender can follow the shooter backwards. The defender makes up most of the distance lost. In the end, stepping back only gives the shooter 0.1% more efficiency, which is probably not worth the time spent and risk of traveling.
This is only one possible study that can be done with the STATS data. Because the data tracks literally everything on a basketball court, everything is possible to study. For example, finding the impact of good ball movement on possession efficiency is now possible to study, when it would have been impossible before.
Because only 6 NBA teams have this data, those 6 are going to get a leg up on the competition. I’m making this up, but say Raja Bell shoots 15% better when he catch-and-shoots, but Gordon Hayward only shoots 3% better. Which player should you send on the Harpring Curl?
What if you knew Rajon Rondo is only 3 percent better when he’s open compared to when he’s tightly contested? Wouldn’t it make sense to then leave Rondo open, and use his defender to break up passes and double-team other players? Using this data, you could become a drastically more efficient defensive team.
Statistical analysis can’t improve everything. But if it improves one personnel decision, or allows for one more win (say, against the Celtics), then the cost of statistical research will have been worth it.
Leave a Reply
You must be logged in to post a comment.