This study aims at developing an entertainment air hockey robot with which the human player can enjoy his/her play using tactics. In order to provide such highly-offensive tactics with the robot, the authors pay attention to the human player's eye movement, which is deeply related to decision-making, recognize it and etc., as a kind of motion skills. That is, we start it from observation, i.e., analyzing the measured human player's eye movement and giving findings from the results to the robot. In this paper, we introduce a time-series of gaze point likelihood distribution for statistical analysis. Experimental results by this approach have shown tendencies of how the players use their eyes. From discussion of the results, it has indicated that we could generate tactics as attacking weak point of the human eye movement mechanism.