In a world of limited resources, scarcity and rivalry are central challenges for decision makers. We examine choice behavior in competitive probability learning environments that reinforce one of two strategies. The optimality of a strategy is dependent on the behavior of a computerized opponent: if the opponent mimics participant choices, probability matching is optimal; if the opponent is indifferent, probability maximizing is optimal. We observed accurate asymptotic strategy use in both conditions suggesting participants were sensitive to the differences in opponent behavior. Moreover, the results emphasize that ‘irrational’ probability matching can be adaptive once such competitive pressures are taken into account. The application of reinforcement learning models to the data suggests that computational conceptualizations of opponent behavior are critical to account for the observed divergence in strategy adoption.