APBRmetrics

I stumbled upon an interesting article which is using SportVU data to create its own RAPM. Here is the description of the stat:

-4.123 + .594 * Points + -.559 * FGA + Passing Efficiency * 21.809 + -.587 * TOV per 100 touches + 4.241 * Contested Rebound Pct + .043 * Minutes per Game + -.247 * FTA + .138 * Rebs * Three Pt Rate

Here a couple of brief definitions of some of the new measures:

Passing Efficiency is a SportVu derived metric that I detailed here, it measures the points created per pass attempts, it could also be thought of as Passing Effectiveness.
TOV per 100 touches (TOVtouch100) is another SportVu derived metric that measures the number of turn overs committed per 100 touches of the ball. I went into that one in more detail here and here for the Celtics last year.
Contested Rebound Pct, also via SportVu, this is the percentage of rebounds a player gets that are up for grabs with the opposing team

http://counting-the-baskets.typepad.com ... minus.html

Its SportVU + Box score regressed onto RAPM. I think the future of RAPM could be something like J.E.'s xRAPM combined with SportVU or Vantage Sports tracking stats. With the player tracking stats, you can acount for a lot of the things that the box score is missing. We could add Blockouts, Screens, Hockey Assists, Deflections, etc. to make RAPM more accurate.

BPM was created with the goal of quantifying SPM so it could be used for all players in the database. This is a very noble and worthy goal. However, I feel as if the next generation of SPM will include sportsvu data though as it allows for us to receive a more clear picture of whats going on. The leading pick in the prediction contest I believe is already doing something similar to this.

While I'm here, I might as well ask. In the prediction contest, blends of SPM and RPM perform better than RPM alone; why does this not carry over on the individual level as well? Wouldn't it also make sense then than individual player RPM blended with SPM is more predictive than RPM alone? I understand why people would want to keep RPM separate (and how the use of priors already includes SPM), but if it offers a new perspective, it should be calculated, no?

The player tracking plus minus mentioned here is Andrew Johnson's. A version of this is the current leader in the team win projection test.


Yes it would be good to see metric blend scores for players. I should have done so already. I was hoping the recently discussion metric comparison test would add further insight on the blend components and optimal blend. Player tracking PM should be part of it. Or maybe it is good enough on its own. Further testing would help.

Mike G has used some metric blends in recent threads, on a one off basis.

Is someone was looking to add something to RAPM, two things come to mind: average RAPM for a player's substitute teammates (to give clues about collinearity issues) and average RAPM for counterpart opponents (gives information about starter / sub environments, east / west, stronger / weaker positions in NBA today). Think roll-up one number metrics are scary or bad? Bring 2 more along to help interpret.