RAPM prior for rookies.

[AcrossTheCourt]

I was wondering if anyone here has already experimented/tested a way to deal with rookies in a prior-informed RAPM model without anything like box score stats or draft position. Typically, I've seen people use a -2 prior for rookies, but I was wondering if someone has already found a better method. I was thinking of reducing the lambda value for all rookies so they're less constrained. Probably will test it out by reducing it by half.

[colts18]

I posted this on RealGM about the topic. The prior should be closer to -1 or -1.5

I did some research. I looked at the 2006 season using J.E.'s NPI RAPM (I believe that one gives everyone a 0 prior). I looked at all rookies who played 500 MP that season (30 guys):

Average:
Ranking: 263.1 out of 458
Offensive RAPM: -0.60
Defensive RAPM: -0.45
Total RAPM: -1.04

8 out of 30 were 0 or better (26.7%)
17 out of 30 were -1 or better (56.7%)
21 out of 30 were -2 or better (70%)
Only 1 out of 30 was worse than -3 (3.3%)

the 13 players with 1500 MP averaged -0.51 RAPM.

Here is the breakdown by round.

Lottery picks, top 14 picks (including those under 500 MP):
-0.92 average RAPM

Rest of 1st round:
-1.42 average RAPM