The ship rock data is one of important data which need to be handled real-time during the mission of track and control ship. In order to improve the accuracy of the ship rock data, the paper proposes an improved M-step maximum likelihood ship rock data predictive method. The proposed method uses the stable 41 point-endpoint smoothing algorithms to solve the low accuracy level of the beginning 210-point filtering algorithm, which is also called maximum likelihood method. Besides, the proposed method need less filtering data than existing 210-point filtering algorithm. The paper uses some ship rock data from a real-time mission as a simulation data, and the simulation results show the effectiveness of the proposed algorithm.