Early-season statistics (Weeks 1--4) are noisy. We apply Bayesian shrinkage, blending current-season averages with a prior based on the previous season's performance. The shrinkage weight decays exponentially as more games are played.