This paper considers an empty container repositioning policy with flexible destination ports. The policy only specifies the direction of the empty flows, whereas ports of destinations are not determined in advance and empty containers are unloaded from vessels as needed. Interviews with industries show its application in practice, but little research has been reported on its effectiveness. The purpose of the paper is to formulate this policy mathematically and evaluate its effectiveness via simulation. Numerical experiments demonstrate that the new policy outperforms the conventional policy significantly in situations where trade demands are imbalanced and container fleet sizes are within a reasonable range.