If the out-of-sample residuals of all of your highest-projected guys are positive and the out-of-sample residuals of all your lowest-projected guys are negative, wouldn't that imply that the prior you're using is too sticky relative to what would predict best out of sample?
I'm kind of curious if within position there is variation by height. Like are small centers better? I think we tend to over-index on Shaq, but KAT is really good at FTs (it might be the thing he's best at [shots fired]).
If the out-of-sample residuals of all of your highest-projected guys are positive and the out-of-sample residuals of all your lowest-projected guys are negative, wouldn't that imply that the prior you're using is too sticky relative to what would predict best out of sample?
Yeah! This is something I’ve been working on. Keen eye though
I'm kind of curious if within position there is variation by height. Like are small centers better? I think we tend to over-index on Shaq, but KAT is really good at FTs (it might be the thing he's best at [shots fired]).
That’s exactly what I wanted to do next!
Yeah! These priors matter more and more as you have less data