A method and an apparatus for impact position detection of impact sport equipment are provided. The method includes following steps: retrieving a vibration signal generated by a vibration sensor detecting vibration caused by impact of impact sport equipment and a ball; performing a spectrum analysis on the vibration signal to obtain eigenfrequencies of the vibration signal in a frequency domain; and calculating at least one piece of characteristic information by using an amplitude of each eigenfrequency and inputting the same into a prediction model established in advance by using machine learning, so as to estimate an impact position of the ball on the impact sport equipment, in which the prediction model is trained by using characteristic information of multiple vibration signals and corresponding impact positions.