In this paper, we introduce a generalized version of impulse balance equilibrium. The stationary concept is applied to 3 × 3 games based on the Bailiff and Poacher Game (Selten, 1991) and its predictive success is experimentally tested against the one of Nash equilibrium. Experiments with 26 different games were conducted; 12 games with completely mixed Nash equilibria and 14 games with partially mixed Nash equilibria. In all the games, generalized impulse balance yields predictions that are closer to the data than the ones of Nash equilibrium. Overall, generalized impulse balance has a significantly higher predictive success than Nash equilibrium.