Frustration is an integral part of game experiences (and life in general); however some types of player frustration are just plain bad and can lead to players feeling powerless and even angry. This talk describes a computational model that was able to identify patterns of behavior indicating unwanted player frustration. The model is based on player behavior logged through gameplay metrics from the game Kane & Lynch 2: Dog Days. The framework for the method used to arrive at this particular model is at least as interesting as the results: It was obtained by methodological triangulation, using both qualitative and quantitative methods (e.g. observation, interviews and data mining). Apart from the advantage for game development being able to establish automated frustration detection systems, the model could also theoretically be used as an element in adaptive gameplay design. This presentation is co-authored by Dr. Alessandro Canossa (Information Technology University of Copenhagen).