Abstract: The stabile walking gait is the basis of good performance for biped soccer robot. The ZMP trajectory is optimized to improve walking stability by the simulation in Yobotics based on evolutionary algorithms. By the rules of FIRA humanoid game, penalty kick strategy and goal keep strategy are proposed for biped soccer robot based on evolutionary gait algorithm. The penalty kick strategy includes three strategies, such as ball searching, range measurement and path planning. These strategies enable the kicker robot to finish shooting in real-time successfully. To change passive defence of goalkeeper, goal keep strategy is conducted, by which the goalkeeper is able to block the shooting of kicker effectively. Finally, through the experiment of walking gait optimization, ball color calibration and range measurement, it is validated that the evolutionary walking gait improves the walking stability and enhances the walking speed and the effectiveness of those strategies is able to adapt to the FIRA game.