Under the BlackScholes type financial market, a dynamic portfolio decisionmaking model is proposed, where the expected relative terminal wealth is maximized under a constraint on the shortfall probability below a benchmark defined by a stochastic process. Stochastic analysis method and nonlinear programming theory are applied to obtain the explicit solutions of the optimal strategies and the efficient frontiers. The results exhibit three-fund separation theorem which include the riskless asset, revised market portfolio and benchmark portfolio. Numerical examples are presented.