Abstract
This article aims to enhance factor investing with reinforcement learning (RL) techniques. The agent learns through sequential random allocations which rely on firms' characteristics. Using Dirichlet distributions as the driving policy, we derive closed forms for the policy gradients and analytical properties of the performance measure. This enables the implementation of REINFORCE methods, which we test on a large dataset of US equities. Across a large range of parametric choices, our result indicates that RL-based portfolios are very close to the equally-weighted (1/N) allocation. This implies that the agent learns to be agnostic with regard to factors, which can partly be explained by cross-sectional regressions showing a strong time variation in the relationship between returns and firm characteristics. All in all, our results contribute to a nascent stream of literature that relativizes the usefulness of mainstream characteristics in asset pricing models.