In RoboCup simulated 2D football match, defensive model is one of key techniques, and strengthening the defense in the key area of the opponent's attack can effectively prevent the opponent from scoring. According to the problem in the difficulty in optimizing the defensive strategies for RoboCup simulated 2D teams, this paper proposes a method to optimize the defensive strategies based on the team's attack key areas. This method explains the rationality for the emphasis on some offenses by analyzing the log files in the process of the games and by using data mining method to extract motion chain and the thermodynamic chart of the passing, makes more reasonable dynamic discrete division for the field by ball position distribution in the process of the game, uses density clustering algorithm with adaptive parameters to calculate the key areas of the attack of the teams in every discrete area, and designs the defensive strategies based on the key players in these areas and in the process of the attacks. The underlying code is optimized according to the defensive strategy in experiment part. Results show that the key defense of the offensive key areas can effectively prevent the attack of the other side, and at the same time, the correctness of the calculation of the offensive key areas is verified.