Abstract: Virtual characters in computer games sometimes need to find a path from point A to point B while minimizing the risk of being spotted by an enemy. Visibility calculations of the environment are needed to accomplish this. While previous methods have focused on either general visibility calculations or calculations based only on current enemy positions, we suggest a method to incorporate the agent's knowledge of previous enemy positions. By creating a probability distribution of the positions of the enemies and using this in the visibility calculation, we create a more accurate visibility map.