We conduct an experiment to uncover the reasons behind the typically large behavioral variation and low explanatory power of Nash equilibrium observed in Tullock contests. In our standard contest treatment, only 7% of choices are consistent with Nash equilibrium which is in line with the...