New SportVU Visualizations Released on NBA.com Possibly Tremendous Resource

facebooktwitterreddit

Dec 15, 2014; Atlanta, GA, USA; Atlanta Hawks center Al Horford (15) blocks Chicago Bulls guard Derrick Rose (1) from Hawks guard Jeff Teague (0) during the second half at Philips Arena. The Hawks defeated the Bulls 93-86. Mandatory Credit: Dale Zanine-USA TODAY Sports

As has seemed to become a pattern over the course of the young season, NBA.com and STATS, Inc. quietly released even more SportVU based-data on Friday. With new SportVU visualizations, it’s now possible to watch every play tracked so far this season exactly as seen by the SportVU cameras. Though this does not represent the entirety of the SportVU dataset[1. The visualization plots the coordinates from that data set, but without the underlying detailed location data, the search and query functionality which is the heart of the analytic usefulness of the data is not enabled.], it is an incredibly valuable tool, perhaps on par with the “All-22 cam” offered by the NFL to allow fans to see the game from a more tactical, coaching-friendly perspective.

While not strictly an analytics tool, theses animations have incredible potential for assisting with the communication and interpretation of all forms of analysis, whether metric or scouting based.

For example, demonstrating even fairly common NBA sets has been difficult at times, with the sheer business of the televised game footage sometimes obscuring the bare bones of what has happened. However, it is now as simply as linking to a play such as this Atlanta “floppy” set vs. Cleveland to allow the reader to view a play develop frame by frame. This view can not only demonstrate the offensive set with greater clarity, in can also give a better qualitative indicator of where defensive breakdowns occur. In the play linked above, Kyrie Irving navigates the “single” part of the signature “single-double screens” which characterize a “floppy.” But he loses contact with Dennis Schroeder going around the double staggered screens set by Paul Millsap and Al Horford, requiring Horford’s defender to help into the lane, leaving Korver wide open for a baseline 17-footer.

Similarly, on this play, Anderson Varejao’s man DeMarre Carroll ends up with a wide open corner 3, in large part because Irving allows Kyle Korver to cut backdoor in attempting to deny Korver the ball coming off a pindown screen. However, perhaps this play isn’t really a “breakdown” — denying Korver the catch on the wing at the cost of the occasional back door cut might well be acceptable to Cleveland’s defense. Quantitatively, we can maybe start to better judge a defensive possession than by points allowed. Even if Carroll misses this shot, crediting Cleveland with a “stop” here doesn’t seem especially correct. Up to the moment the shot went in or didn’t[2. To some extent, a random event with an expectation based on the shooter’s proficiency.], not a great result for the defense. While the public data isn’t to the point of allowing systematic study, it’s a bit easier to decouple the process from that result with this view.  To put it another way, from the standpoint of Cleveland’s defense, the end point of the play can be seen as the release:

Indeed, this sort of visualization allows a clearer view of certain schematic problems. Part of the difficulty in scouting defense from the common TV vantage points is perspective. Last week, in discussing Houston’s somewhat illusory early season defensive efficiency, commenters raised questions as to whether shots I was describing as “open” really were. Certainly, from the mid court sideline camera angle, on many of these attempts it looks like Houston is getting a hand up:

But from the SportVU-based overhead view, the distance for effective defensive pressure is more readily apparent. [3. In general terms. It’s extremely likely various players have different “effective contest radii” but to my knowledge no one has formally studied this.]

This different perspective allows for easier demonstration of good and bad execution of various schematic choices. One prime example from this season is Cleveland’s desire to hedge pick-and-rolls. It’s possible to see the lack of effectiveness of this technique for the Cavs personnel on film, but the differences can be seen even more clearly from the overhead view. Compare this play where Kevin Love’s ineffetive attempt to hedge allows Dante Exum to turn the corner for an easy lob to Rudy Gobert and this one where the surprisingly nimble Zach Randolph is much more effective against Steph Curry. Love does not force Exum away from the basket:

whereas Randolph does force Curry to angle towards midcourt:

This serves to both prevent drives, shots or killer passes from the ball-handler, but also creates a lane for Mike Conley to recover back to proper position rather than (as is the case quite often in Cleveland) the hedging defender serving almost as an additional screen on his own point guard.

These new visualizations aren’t perfect: the SportVU system still does not directly capture orientation or facing. Randolph’s defense certainly appears much better if Curry is facing the left sideline (as he was) than if Steph was dribbling towards the right sideline away from the intended trap. And for some people, this may look like just dots on a screen rather than basketball. But the more objective perspective and exact positioning allows for much better analysis of many aspects of play than traditional tape, and can be used as a teaching tool to implement and explain many analytic discoveries, whether to show the distance a defender needs to cover to contest a three pointer, to the illustrate the  effectiveness of Korver’s “gravity” on a defense.