Optimal arbitrage under model uncertainty (Q657697)

The authors consider an equity market model with immeasurable uncertainty regarding relative risk and covariance structure of its assets. The papers provides answers to the following questions: (1) what is the highest return on an investment relative to the market that can be achieved using non-anticipative investment rules over a given horizon under any admissible configuration of model parameters, and (2), what are the weights corresponding to the assets of an investment rule that realizes this maximum return? The highest return is obtained here via the smallest positive supersolution to a nonlinear PDE of Hamilton-Jacobi-Bellman type which, under appropriate conditions, is the value function of an associated stochastic control problem. This value function can also be interpreted from the perspective of stochastic game theory.