How game theory fails to explain man

We analyze human decision making experimentally in a simple air campaign model (Campaign), in order to test game theory's explanatory power. Game theory predicts that Campaign-playing subjects should play randomly according to a given probability distribution. Our experiments indicate that this theory fails to explain both how humans think, and how they act. With experience, their actions even tend to move away from the theory's predictions. However, our subjects consistently improve their ability to avoid dominated actions, so it is the failure to randomize that separates human practice from game theory, rather than failure to identify reasonable candidate actions.