The final test run was an attempt to simulate a real-world host behavior scenario. The maze was laid out, populated with hosts, and a single paramor node. The paramor node was given oreders to try and reach the cell in the top-right-most corner. The hosts were programmed to move forward along their current path until reaching a point of decision. This intended to simulate people in an office building or other environment familiar to the host.

Three different paramor behaviors were implemented and compared. The first was the as-the-crow-flies distance calculation used in the previous simulations. The second was the basic-maze-solving algorithm. The third assumed that the node had knowledge of the complete map. As expected, the omniscient behavior performed best.