Based on a Bayesian time-varying beta model, we explore how the systematic risk exposures of hedge funds vary over time conditional on some exogenous variables that managers are assumed to use in changing their trading strategies. Using data from CSFB/Tremont indices over the period January 1994-September 2008, we found that (1) volatility, changes in T-bill, term spread and shocks in liquidity significantly impact the time variation of hedge fund betas; (2) when mean reversion and instruments in beta become predominant, hedge funds tend to be more risky, more dynamic and less dependent by their own style benchmark; (3) if risk exposure is assumed to be constant while it is time-varying, performance appraisal can be seriously distorted and overestimated.