Assisted-On Three Pointers By Zone and By Player

Partially inspired by Ken Pomeroy’s blog post on assisted-on shots, I wanted to use my shot chart/play-by-play dataset and look at the occurrence of assists on three pointers, and see if I could find any interesting nuggets around this topic.

To my surprise, I found that the vast majority of three pointers occur directly as a result of an assist.  Of the dataset I analyzed, 83.7% of all three pointers were assisted-on.  If this is common knowledge, then I must have been out of the loop.

When you break this down by shot zone, you see that the corner three is the most assisted-on shot, the straightaway three the least, and the wing three somewhere in-between.

This makes a lot of sense, but might not seem all that enlightening.  If you’re shooting from the corner, chances are that you initially got the ball from a pass.  The other alternatives are to either 1) dribble into the corner and shoot or 2) perform an isolation play prior to shooting.  Neither of those possibilities seem as pragmatic as being passed the ball in the corner.

When you combine this finding with the knowledge that the corner three is one of the two most efficient places to shoot from (note: I will have an extensive post about this point later), I begin to wonder if the relative efficiency of the corner three is more attributable to the shorter distance, or to the fact that more shots come directly from a pass.  In other words, is it possible that teams score more from the corner because a higher percentage these shots are being set up by good team passing, and ostensibly result in more less-contested shots?

Unfortunately, only SportVU data (and maybe Synergy data) can answer that question.  But it’s something to think about.

Also unfortunately, I don’t have the data set to get field goal percentages on assisted-on shots, because you would have to know how many potential assists resulted in missed shots, which again could be answered by SportVU’s computer vision-aided data.

But what I could do with my dataset is to plot individual players’ three point field goal percentages against the percentage of their made three pointers resulting from an assist.  As a result, I created a d3-powered visualization, plotting these data points against each other.

Below is a static image of the visualization, and here is a link to the interactive assisted-on visualization, which will allow you to toggle which metrics are plotted against the X- and Y-axes.  Each circle represents a different player, where their position represents their 3pt FG% and 3pt makes assisted-on%, and the size of the bubble represents volume of shot attempts.

This data comes from the (lockout-shortened) 2011-2012 regular season:

The first thing you’ll notice is that those players with the lowest assisted-on three point percentages are point guards.  In this visualization, the two lowest assisted-on percentages are Phoenix Suns guards Sebastian Telfair and Steve Nash.  Green circles represent point guards, where you can see most point guards having an assisted-on percentage lower than the average of 83.7%.

One possible explanation is that a higher proportion of point guard shot opportunities come more naturally from dribble drives and ball screens than passes.  After all, who else on the floor would pass to a point guard for a three?  They’re usually the best passers, and usually the ones setting up other people for threes.

What I find interesting here are the non-point guard gray circles that have low assisted-on percentages.  These players tend to be those who have reputations for one-on-one play.  If you hover over the circles in the interactive visualization, you’ll see usual suspects like LeBron James and Carmelo Anthony, and some non-obvious, but-after-second-thought-that-makes-perfect-sense players like Hedo Turkoglu and Jodie Meeks.

Here I’ve appended the names of a few particularly interesting players.

Poor Kobe Bryant and Jordan Crawford, two players known for chucking up threes at will, appear in the bottom-left quadrant with very large circles, meaning their made threes aren’t usually assisted, they shoot low three point percentages, and they shoot a lot.  Not a great combination.

Two more chuckers, Deron Williams and Brandon Jennings, have decent field goal percentages but low assisted-on percentages, suggesting their shots may be contested.  For Williams, this may be understandable considering the talent around him, but for Jennings, this could be more questionable, given that he has Ersan Ilyasova and Mike Dunleavy, who shoot 45% and 40% from three, respectively.

Two of the more renowned three point bombers, Ryan Anderson and Steve Novak, appear in the top right quadrant, meaning they shoot a great percentage, but are dependent upon the pass to make their threes.

Finally, you have a few surprising guards in the lower right quadrant who shoot high percentages without as much dependence on the assist: Stephen Curry (in an injury-riddled season), Ramon Sessions, and Gary Neal (!).

I’m sure there are more interesting insights to glean from this dataset, so please e-mail, tweet, or comment to let me know your thoughts.

Assisted-on Data Visualization on VORPed