We provide a model with overconfident risk neutral investors, and therefore no risk premia, in which a price-based portfolio such as HML earns positive expected returns and loads on fundamental macroeconomic variables. Furthermore, loadings on such portfolios are proxies for mispricing, and therefore forecast cross-sectional returns, even after controlling for characteristics such as book-to-market. Thus, an empirical finding that covariances incrementally predict returns does not distinguish rational factor pricing from a setting with no risk premia. The analysis reconciles the high risk (market betas) of low book-to-market firms with their low expected returns, and offers new empirical implications to distinguish alternative theories.