NBA Consensus Plus/Minus Data
As mentioned in the "Can The Suns Be Great With Bradley Beal?" story
(Update - 6/27/2023: I added a new metric to the consensus! See below for details.)
As mentioned in this post, here is the file with the new Consensus Plus/Minus data I calculated, including a breakdown of the rescaled categories that went into the overall CPM:
NBA Consensus Plus/Minus (1977-2023)
The basic framework is that I collected regular-season data for the following metrics:
Player Efficiency Rating
Win Shares per 48 Minutes
Box Plus/Minus
“Basketball on Paper” Plus/Minus (Net Offensive Rating minus Defensive Rating, adjusted for Usage using the trade-off method described here)
On-Court Plus/Minus
On-minus-Off Plus/Minus
A ✨new metric✨ called RAPTOR Game Score (RGS) %. What’s this? It’s a stat similar to the NBA’s PIE% stat, which measures a player’s personal share of all “good things” that happened on the court in that player’s minutes (a framework I have always liked, because it seamlessly accounts for pace, league environment and even a lil’ bit of defense in there too), but it uses a Game Score-like linear weights formula derived from FiveThirtyEight’s RAPTOR metric, rather than more archaic weightings.1
Each metric was scaled to a Plus/Minus-style format of Net Points per 100 possessions, and set so that the positional average in each season was 0.0. (I have complicated feelings about positional adjustments in basketball, but Nate used one in his original version of CPM, so I included it here.) Then they were blended together with the following weights, which were set with an eye towards balancing correlation with RAPTOR but also ensuring each category had a value between 10% and 20% to get a true sense of “consensus”:
Player Efficiency Rating: 10%
Win Shares per 48 Minutes: 10%
Box Plus/Minus: 20%
“Basketball on Paper” Plus/Minus: 10%
On-Court Plus/Minus: 15%
On-minus-Off Plus/Minus: 15%
RAPTOR Game Score %: 20%
The resulting metric was then adjusted for team effects such that (4 * the minute-weighted average of a team’s roster ratings) is equal to the team’s efficiency margin. This CPM had a correlation of 0.85 with RAPTOR among all players with at least 1,000 minutes last season.
Also included in the file is a Wins Above Replacement (WAR) for CPM, calculated by adding a replacement-level of 2.75 points/100 to each rating, and multiplying:
WAR = (CPM+2.75)*(Share of team’s total minutes)*(Team games/82)*10.1 + Minutes*N
where N is a residual term to make leaguewide WAR add up to a target value. (N was 0.000008 last season.)
Filed under: NBA
The formula for RGS is: 1.0*PTS-0.9*(FGA + 0.44 * FTA)+0.4*ORB+0.2*DRB+0.6*AST+2.4*STL+0.8*BLK-1.3*TOV-0.7*PF