Recently, high-speed vision systems, which provide high frame rate visual information in real time, are widely used. Since the high-speed visual information is too fast for humans to recognize directly in real time, these systems are mainly used to control artificial systems such as robots. Contrary to this trend, we attempt to provide the high-speed visual information directly to humans so that they can respond to fast motional phenomena in real time. In this paper, we discuss an air hockey gaming system that displays predictive information of the puck position so that a player is assisted to respond to fast puck motion. We investigate dominant factors that make air hockey difficult and effectiveness of predictive information display by analyzing the plays of six subjects on an air hockey simulator. The results show that players almost cannot predict the puck motion without any assistance and visual feedback control plays an important role in positioning the mallet. It is also shown that displaying predictive information results in better positioning performance mainly because the players' task is transformed into a simpler one.