Buttons are the most commonly used input devices. So far the goal of the designers was to provide a passive button that can accept user input as easily as possible. Therefore, based on Fitts' law, they maximize the size of the button and make the distance closer. This paper proposes Button++, a novel method to design smart buttons that actively judge user's movement risk and selectively trigger input. Based on the latest model of moving target selection, Button++ tracks the user's submovement just before the click and infers the expected error rate that can occur if the user repeatedly clicks with the same movement. This allows designers to make buttons that actively respond to the amount of risk in the user's input movement.