We propose a unified set of distance-based performance metrics that address the power and extreme-error problems inherent in traditional measures for asset-pricing tests. From a Bayesian perspective, the distance metrics coherently incorporate both pricing errors and their standard errors....