We demonstrate that the results of inverse dynamics depend sensitively on the model used to simulate the human body. For this reason we produce "ideal" input data of the ground reaction force and the heel acceleration for a computer simulated drop jump with a wobbling mass model. Then we determine the internal torques and forces in the knee and hip joint via inverse dynamics by employing a rigid body model. It turns out that during the impact phase the analysis with the rigid body model yields completely incorrect internal torques and forces.