However, though AWS is clearly a better metric then PER or Game-Score, it still has significant limitations. I wanted a new/better box-score metric, and so I decided to derive one. Using 8-year-long-sample RAPM data from Jeremiah Englemann (posted by Daniel Myers), I created a new SPM-metric I'm calling HDR. For details on what SPM is and how I basically derived it, see this page.
For those who are already privy to this stats-stuff, here's the lowdown. My HDR-SPM is constructed like AWS, except Offensive/Defensive rebounds are replaced by percentages, as are Blocks with BLK%. I don't use points scored, but shots made/missed, broken down by 3's, 2's, and FT's. Defensive Rating and Minutes-per-Game are added on as well, and the whole thing is per-48 minutes/per-100 possessions. The weights were determined by simple multi-linear-regression, and tested for accuracy via weighted-correlation with the RAPM sample. Here's a table for reference.
I'll have more on HDR later, but for now, let's get to the NBA-Finals Ratings. The idea is the same as last time. Scores are minute/pace adjusted, and adjusted for strength of opponent. I previously used Opponent-AWS-Allowed as the basis for opponent-adjustment, but I am now using the regular-season SRS to determine "strength" of opponent. Ratings include the 2000-2013 Finals, and rank the Top 30 by Per-Game production.
- I'm still tweaking HDR, both in terms of the weights, and what type of metric I want it to be (should it include Defensive Rating?) but I plan on posting the formula ASAP.