We implement a recursive out-of-sample method to examine anomalies-based ex-ante predictability in the cross-section of stock returns. We obtain a series of simulated out-of-sample returns, consistent with investors using only prior information when choosing predictor variables. We find that, by commonly used performance criteria, real-time trading strategies based on size, value and momentum effects would not consistently outperform a passive index of South African stocks – despite consistent in-sample excess returns. Our results suggest that the empirical relationship between the anomalous factors and cross-sectional average returns is unstable.