I am trying to build a bayesian probit regression model to predict strikeouts in Major League Baseball. I have observations for every plate appearance this year with the following columns:

batterName - name of the batter

pitcherName - name of the pitcher

venueName - name of the ballpark

I am trying to set an informative prior using preseason projections from another source for the players strikeout percentage by making a player with a preseason projection of, say, a 25% strikeout rate, have a prior of N(invnormcdf(0.25), sigma), where I fix the value of sigma. The issue that I am running into is that BRMS holds out the earliest value in the alphabet for each variable as the reference and does not let me set a prior. How do I adjust for this, because I want to set an informative prior for all of the players and venues? If BRMS makes me hold out a player P with a preseason prior of N(invnormcdf(x), sigma), should I set the prior for another arbitrary player with a preseason prior of N(invnormcdf(y), sigma) as N(invnormcdf(y) - invnormcdf(x), sigma)? When I want to update my model with new observations as the season progresses, will this sufficiently update the opinion of the reference player? I am worried if the reference playerâ€™s strikeout percentage diverges significantly from my preseason prior it will not be accounted for correctly. Thank you. Let me know if any other information is needed.