Nylon Calculus: PT-PM at the season’s halfway point

Los Angeles Clippers guard Chris Paul (3) makes tonight's DraftKings daily picks. Mandatory Credit: Jayne Kamin-Oncea-USA TODAY Sports
Los Angeles Clippers guard Chris Paul (3) makes tonight's DraftKings daily picks. Mandatory Credit: Jayne Kamin-Oncea-USA TODAY Sports /
facebooktwitterreddit

A few years ago I created my version of a statistical plus-minus model, called Player Tracking Plus Minus (PT-PM). The basic idea was to explore the relationship of some of the new SportVU player tracking statistics along with traditional box score stats, with player impact as estimated by a regularized adjusted plus-minus. I have put up a couple of posts in the past on both the offense measure and defense measure if anyone is interested in reading more of the specifics.

I updated the numbers for this year for the first time this weekend, given that we are approximately at the half-way point of the season. The results are in the visualization and table below. The visualization is has a scatter of the offensive and defensive PT-PM, color coded by overall score. The table has the offensive PT-PM, defensive PT-PM, overall PT-PM, and a regressed version that adjusts for the minutes played, which is probably a somewhat better prediction of the player’s performance going forward. Both can be filtered by minutes played (preset at a minimum of 300 minutes) or teams.

A couple of highlights:

  • Top three PT-PM so far: Kevn Durant, Chris Paul, and Steph Curry. Chris Paul continues to be perpetually under rated.
  • Bottom three PT-PM so far: Marquese Chriss, Emmanuel Mudiay, and Isaiah Whitehead. Mudiay moves up one spot from last year!
  • Top three offense so far: Durant, Isaiah Thomas and Chris Paul. Isaiah Thomas, everybody, Isaiah effin Thomas.
  • Top three defense so far: Manu Ginobili (???), Andrew Bogut, and Dewayne Dedmon

The Ginobili thing probably needs more explanation and investigation, basically Ginobili is getting steals at a very high rate, fourth highest in the NBA, he is drawing offensive fouls at a high rate, he has gotten very lucky as a defender around the rim, something the PT-PM model factors in along with how often the defender protested the rim, and, in addition, the Spurs defense has been stellar when Ginobili is on the court (along with all of the Spurs second unit). This is where, however, I caution against interpreting that he has performed very well as a defender over this time as, the fact that he is a great defender or will continue to perform at this level.

Next: James Harden is proud of his defensive stats

In addition, I also compared my numbers to ESPN’s RPM. For players with 300 minutes or more the R^2 was .73 to the minute regressed version of PT-PM. As I had found in previous years the relationship on offense between the two metrics is much higher than on defense, with a .78 and a .49 R^2 respectively. Giving some indication of how much easier it is to capture offensive contributions with than on defense.