We experimentally test the hypothesis that players' valuations of a game coincide with their Nash equilibrium earnings. Our results offer significantly less support for this hypothesis than for the prediction of dominant strategy play.