Technologies embedded in computer and console games increase with the high graphics performance, Nowadays game software includes 3d graphics engine, physical simulation and artificial intelligence. Testing the game's levels takes more and more time and becomes a complex task. How can you ensure that your tests are relevant? In this paper we present a software framework that performs automatic tests of first player shooter (FPS) game levels. This framework simulates the player's movements and actions according to the gameplay features. During this simulation, the testing framework captures events, tests visibility of objects, profiles the embedded artificial intelligence (AI). It delivers compact reports with the following information: which events arrive, how are they synchronised, what behaviours of non player characters are used, how can these behaviours can be modelled in finite state machines, what is the emergent game dynamics? These reports give an abstract view of what goes right and what goes wrong in the level. Then the level designer can tune his job and perform new tests.
