We apply machine learning techniques and use stock characteristics to predict the cross-section of stock returns in 33 international markets. We conduct a stringent out-of-sample test to examine concerns about overfitting: the models are trained with past U.S. data and used to predict...