Evaluating the performance of hedge funds is challenging because any benchmark model is unlikely to capture their numerous strategies. To assess the impact of model misspecification we develop a novel approach to formally compare hedge fund models. This comparison sharpens performance evaluation by improving the separation between pure alphas and factor exposures. We find that the standard models deliver the same performance as the simplest benchmark—the CAPM. In contrast a parsimonious model based on economically motivated factors (including carry time-series momentum and variance) tracks alternative hedge fund strategies and achieves a sizable performance reduction relative to the CAPM.