Page 1 of 1

Reworking the Four Factors to highlight shot selection

Posted: Thu Feb 21, 2013 1:47 pm
by Levy2725
Not sure if people have seen it, but earlier this year I built a metric for evaluating shot-selection. I've been calling it Expected Points Per Shot (XPPS), and it's built around expected point values for shots from different locations. It's a variation on TS% and the formula boils down to (Expected Points/True Shot Attempts). I've been tracking these numbers all season long for players and teams (teams I have both offensively and defensively). Here are links to the Tableau Visualizations - http://www.hickory-high.com/?page_id=6154

At the suggestion of DSMok1 I did some regression analysis to see how my XPPS numbers related to TS% and ORTG. With plenty of help from him I was able to pull the whole project through. We arrived at these figures. 24.1% of the variation in TS% can be explained by variation in shot selection (as expressed by XPPS), and 19.1% of the variation in a team's ORTG can be explained by the variation in shot selection (as expressed by XPPS). Here's a link to the whole analysis: http://www.hickory-high.com/?p=6259

The crux of the analysis was the revelation that you could explain ORTG just as well as the Four Factors do by using just TS%, ORB% and TO%. Since TS% can be explained entirely by XPPS and Shot-Making Difference (the difference between PPS and XPPS) we can simply replace TS% with those two elements in the regression equation. So essentially we reworked the Four Factors, obscuring shooting accuracy slightly, or at least presenting it in a different format than we're used to seeing it, and instead highlighted shot selection.

Would love thoughts, feedback and constructive criticism!

Re: Reworking the Four Factors to highlight shot selection

Posted: Tue Feb 26, 2013 11:29 pm
by balkihowser
I think this is really good stuff. That's a significant impact for shot selection.

One idea for future exploration would be to not use the court segments defined by the NBA but define them yourself based on where the shooting percentage differences are most meaningful. Probably not a huge impact to the result, but more improving on the margin.