Simple Estimated RAPTOR Projections and Aging Curves
Here are the particulars for the Estimated RAPTOR version of the Simple Projection System…
The following parameters detail how much weight should be placed on each of the previous three seasons’ RAPTOR ratings (multiplied by the player’s minutes those seasons) along with a regression-to-the-mean term:
As an example, if a player was a +3 on offense last year in 2,000 minutes, +2 in 1,500 minutes the year before that and +1 in 1,000 minutes the year before that, we’d project his offensive RAPTOR next season to be:
(3*2000*5.985 + 2*1500*2.508 + 1*1000*1.508 −0.601*2344.144)/(2000*5.985 + 1500*2.508 + 1000*1.508 +2344.144) = 2.2
One other important note is that, while we simplified the example above to use the raw RAPTOR values from previous seasons, the actual projections should age-adjust each previous season’s ratings to bring them to their equivalent for the player’s age the following season. Here are those aging curves, represented as the expected year-over-year change in RAPTOR rating on either offense or defense based on the age the player will be in the season being projected:
The final aspect of these simple projections is to forecast rookies. Perhaps the two most important factors in performance for an NBA rookie are his draft position and his rookie-year age. Here are the parameters for a regression-based estimate of offensive and defensive RAPTOR for players as rookies:
Filed under: Statgeekery