Thanks to SportVu the NBA now tracks 'touches' of the basketball for each player in the league. Touches is sort of interesting on a player level because is forms a new baseline for all sorts of other stats we're interested in measuring such as Assists per touch or turn overs per touch, as well as where certain players tend to get the ball with the location break outs.
In the Regressing column of Dead Spin, Reuben Fischer-Baum took an interesting look at what the aggregate number of touches per possession means for an offense. In short, generally higher touches is not such a good thing, with more touches being correlated to less efficiency on offense.
However, as Fischer-Baum notes, faster/shorter possessions are both generally more efficient and lead to fewer touches per possession. The number of touches per possession was strongly negatively correlated with pace a -.44 correlation, while pace was positively correlated with offensive rating and touches per possession in nearly equal amounts. This leaves open the question of causation, are teams with less ball movement scoring more efficiently or getting more touches because they are not able to get into transition as often.
To dig into it a bit, I decided to try to control for possession length. I looked at both the estimated average possession length and the total player 'possession time' as calculated by SportVu. The SportVu description for Time of Possession is given simply as "how long the player possessed the ball."
The estimated possession length is calculated simply using Pace to divide the number of offensive possessions into 48 minutes to get an Estimated Possession Length. The Estimated Possession Length is, of course, only approximate as pace is symmetrical, including defensive possessions. To date I have not found any public NBA data that tracks possession length on the offensive and defensive end separately. However, Ken Pomeroy released a study earlier this year that indicated that in the NCAA pace for any given team is driven primarily by their offensive style of play because teams have significantly more control over the action on the offensive end.
Then touches per possession is adjusted to seconds per touch by dividing the estimated possession length by the number of touches. Teams with a lower seconds per touch are moving the ball faster while those with higher are moving the ball slower with the ball possibly 'sticking' more. For example, the Brooklyn Nets have an average possession length of 15.18 seconds and get 4.21 touches per possession, for 3.6 seconds per touch.
The stickiest team in the league by a couple of standard deviations according to these figures are the Golden State Warriors with 4.2 seconds per touch, but a slightly above average efficiency. Meanwhile the Philadelphia 76ers move the ball the fastest, but are one of the least efficient teams in the league. In fact, the overall correlation is for teams with slower ball movement to be more efficient, though not statistically significant at the 95% confidence level.
I also looked at Pace and the Seconds per Touch and found that the link between Touches and Pace was indeed broken with no correlation between the two.
Lastly, I added up SportVu's Time of Possession (TOP) for each player on every team and divided that by their number of possessions. The SportVu Player TOP was well correlated with the seconds per touch from average possession length and showed same lack of relationship with offensive efficency.
Below are visualizations of the relationships between ball movement speed and offensive efficiency, as well as pace and SportVu and Seconds per Touch relationship.
Comments