Game development has recently gained a lot of momentum and is now a major software development industry. Platform games are being revived with both their 2D and 3D versions being developed. A major challenge faced by the industry is a lack of automated system-level approaches for game testing. Currently in most game development organizations, games are tested manually or using semi-automated techniques. Such testing techniques do not scale to the industry requirements where more systematic and repeatable approaches are required. In this paper we propose a model-based testing approach for automated black box functional testing of platform games. The paper provides a detailed modeling methodology to support automated system-level game testing. As part of the methodology, we provide guidelines for modeling the platform games for testing using our proposed game test modeling profile. We use domain modeling for representing the game structure and UML state machines for behavioral modeling. We present the details related to automated test case generation, execution, and oracle generation. We demonstrate our model-based testing approach by applying it on two cases studies, a widely referenced and open source implementation of Mario brothers game and an industrial case study of an endless runner game. The proposed approach was able to identify major faults in the open source game implementation. Our results showed that the proposed approach is practical and can be applied successfully on industrial games.