After reviewing the original RSPM formula, I’ve decided to tweak it a little bit. The reason was that the original RSPM formula was based on a straight linear regression between box-score stats and RAPM, and the result was that it was better to be a bad free throw shooter than to be a good free throw shooter. That might seem counter-intuitive, but believe it or not, there’s a negative correlation between free throw accuracy and RAPM. I attribute this to players such as Shaquille O’Neal, Ben Wallace, Dwight Howard and others producing outstanding RAPM ratings despite their poor free throw shooting.
I have fixed this glitch by simply removing free throw makes from the regression analysis. Here is the new formula:
RSPM = (PTS/36 * 1.002) – (FG/36 * .981) – (FGA/36 * .566) – (FTA/36 * .141) + (ORB/36 * .546) + (DRB/36 * .462) + (AST/36 * .649) + (STL/36 * 1.526) + (BLK/36 * 1.501) – (TOV/36 * 1.144) – (PF/36 * .224) + (MIN/G * .086) – (9.219)
I really like this formula because it doesn’t reward players for high-volume, low-efficiency shooting. Let’s say a player’s per-36 minute averages are 8 field goal makes, 20 field goal attempts, and 16 points. Those averages would actually hurt a player’s RSPM by -3.1 points.
As I mentioned before, this formula is based on box-score statistics and as a result will struggle to accurately measure either defense or the “hidden aspects” of basketball.
Examples of players the formula will struggle with: Luol Deng, Tony Allen, Nick Collison, Shane Battier, and even Kevin Garnett. That’s because these are all players who play great individual defense and are fantastic team players.
But I do think RSPM is a pretty good estimate for the vast majority of NBA players out there. I’ll keep researching and seeing if any more tweaks would be a good idea.