We propose experiments in virtual reality (VR) as a new approach to examining behavior in an economic context, e.g., heterogeneity in dynamic tournaments. We simulate a realistic working situation in a highly immersive environment. Implementing a tournament in VR, we are able to mitigate the reflection problem, which usually undermines research on dynamic interaction. Moreover, VR allows us to ceteris paribus control for the performance of the virtual peer (humanoid avatar), and thus to get an understanding of the reaction of the subject to the avatar in a really dynamic setting, as the subject is constantly able to observe the avatars performance. Based on a first experimental phase, we match our subjects with an avatar yielding a specific output. We observe that the subjects' performance is highest in a homogeneous tournament, i.e., when they compete against an avatar achieving the same output as they did in the preceding phase. Interestingly, these results are particularly driven by peer effects rather than by tournament incentives. We extensively track the behavior of subjects and the particular situation and, e.g., examine the role of intermediate score differences and the degree of the subjects' movements.