Thanks for the response!
Apologies as I am new to this, I may have formulated the model incorrectly, but my goal is to get an estimated distribution of residuals per player, based on that players individual residuals, along with the overall distribution of all players residuals.
I thought the samples from pm.sample_posterior_predictive stored in posterior_pred would give me sampled residuals and if I grab all samples for an individual player, that is how I could get the estimated residual distribution for a player, is my understanding incorrect?