The first test was to simply have two clients, one using the simple memory model and the other using the predictive model, compete for a free ball. The ball and players were independently placed in random positions on the field. The range of positions for the players and ball is shown graphically in figure 2. The distance of players to the ball ranged from about one fifth to a half of the length of the field. After placing the players, there was a few seconds delay to allow the players to face the ball. The players were then released and it was then recorded to which side of the field the ball was kicked. Since both players were trying to kick to opposite goals, this would measure which was more effective at gaining control of the loose ball.
It was expected that the predictive model would outperform the simpler approach, since the predictive model would not have to take the time to turn and face the goal, before shooting the ball. The results of 2500 trials is shown in table 4.4, and corresponds with our expectations.