Player Tracking Plus Minus (PT-PM) Visualizations

With the NBA season hurtling into mid-season I have updated my statistical plus/minus metric, Player Tracking Plus Minus numbers (PT-PM)[1. More about PT-PM, here and here]. Mid-season enough to get a decent measure of performance for most players, with three hundred minutes played as a rough cut off as where the variance in performance starts to stabilize. I have the total PT-PM offense and defense scores in this linked google doc. But in this post I am going to use those three hundred plus minute players for a few visualizations of the data, to, I hope, provide some insight on the metric and play so far this season.

First is a scatter of the offensive and defensive measures. There is a modest positive correlation with the offensive and defensive PT-PM as well as will minutes played. I have included interactive filters for position (via ESPN) and team.

 
Next are a couple of interesting measures, I think, where I break down the components of PT-PM by position in a couple of box plots, the tabs allow you to look at different breakdowns within the PT-PM calculations:
 

Finally, a couple of scatter plots using the category breakdowns, again with filters for teams and positions. Here I am using the two main categories for offense and defense to visualize the strengths of different players across the league or by position or teams.
 

 
In both of the above scatters, the better offensive or defensive players are in the upper right and the worse ones are in the lower left.

All of these measures are estimations of how a player has performed this year to date, and while that’s related to talent level overall, it is not the same thing. Based on prior research I would expect these to be decent indicators of how players might perform in the second half but certainly not a lock in any sense.