Comparison of RPM and DRE
Posted: Sun Jun 12, 2016 8:08 pm
Matching up player data for Daily RAPM estimate with RPM for 14-15, I found the average absolute difference to be about 1.75 per 100 possessions. That is a lot of potential value to consider. Over 5 man lineup that is essentially enough to account for first to worst team by net points. The box score data can do that too, but the indirect impacts may be as much as half the story. There are indirect effects that DRE and every box score metric miss, including shot defense but also impacts to every team 4 Factor.
35% of players had a differential between RPM and DRE over 2 absolute points. The differential varied a little by position but not that much. Centers varied the most, SFs the least.
I'd ideally like to make this comparison for several seasons and include PT-PM data too. And add demographic data and do cluster analysis. Having all that plus what else is publicly available, one could analyze specific players for impact, relative value, situational fit, etc. A matter of investing time after the initial data has been produced.
RPM gives LeBron James more than half his value to indirect impacts not covered / found by DRE. Same for Draymond Green and a good number of others including some folks that many wouldn't know that about their impact without this technique or possibly believe (rightly or not, pending further research). Among the biggest net negatives for RPM - DRE: Whiteside (biggest), then Ed Davis, Anthony Bennett, Henson, Derrick Williams, B Knight, valanciunus, Bargnani, T Harris... Looks plausible, worth comparing further to me. Or you can ignore the one metric data, if that is your choice. Not clear cut and subject to noise / error but so is everything. Taking the analysis down to factor level would be warranted, if any of the analysis was going to be used.
35% of players had a differential between RPM and DRE over 2 absolute points. The differential varied a little by position but not that much. Centers varied the most, SFs the least.
I'd ideally like to make this comparison for several seasons and include PT-PM data too. And add demographic data and do cluster analysis. Having all that plus what else is publicly available, one could analyze specific players for impact, relative value, situational fit, etc. A matter of investing time after the initial data has been produced.
RPM gives LeBron James more than half his value to indirect impacts not covered / found by DRE. Same for Draymond Green and a good number of others including some folks that many wouldn't know that about their impact without this technique or possibly believe (rightly or not, pending further research). Among the biggest net negatives for RPM - DRE: Whiteside (biggest), then Ed Davis, Anthony Bennett, Henson, Derrick Williams, B Knight, valanciunus, Bargnani, T Harris... Looks plausible, worth comparing further to me. Or you can ignore the one metric data, if that is your choice. Not clear cut and subject to noise / error but so is everything. Taking the analysis down to factor level would be warranted, if any of the analysis was going to be used.