Let me point out that you can make very straightforward Bayesian estimates of performance/skill/alpha by starting with, say, zero-alpha prior and then updating on monthly returns.