We propose a linear bi-objective optimization approach to the problem of finding a portfolio that maximizes average excess return with respect to a benchmark index while minimizing underperformance over a learning period. We establish some theoretical results linking classical No Arbitrage...