This paper proposes an approach to automatically adapt game AI to the environment of the game (i.e., the socalled map). In the approach, a particular map is first analysed for specific features. Subsequently, an automatically established decision tree is applied to adapt the game AI according to the features of the map. Experiments that test our approach are performed in the RTS game Spring. From our results we may conclude that the approach can be used to automatically establish effective strategies dependent on the map of a game.