Many experimental studies implement two versions of one game for which agents’ behavior is fundamentally different even though the Nash prediction is the same. This paper provides a novel explanation of such findings. Starting from the observation that many of the games under consideration satisfy the strategic-complementarity property, I obtain predictions for the direction of adjustment in response to parameter changes which do not require calculation of the quilibrium. I show that these predictions explain the experimental evidence very well. Further, I provide a behavioral justification of the approach, and I explore the relation to alternative explanations based on equilibrium selection theories and the quantal response equilibrium.