We study whether overconfidence affects the choice over contracts in a real-effort task by using three different measures of overconfidence: miscalibration, illusion of control, and overplacement. We find that overplacement, which is the only measure related to the task, has a significant effect on the choice over contracts. Our design also allows us to observe overconfidence measures at different moments of time. We find that more (less) overconfident subjects remain as more less) overconfident over time.