Interpreting other’s actions is a very important ability not only in social life, but also in interactive sports. Previous experiments have demonstrated good estimation performances for the weight of lifted objects through point-light displays. The basis for these performances is commonly assigned to the concept of motor simulation regarding observed actions. In this study, we investigated the weak version of the motor simulation hypothesis which claims that the goal of an observed action strongly influences its understanding (Fogassi, Ferrari, Gesierich, Rozzi, Chersi, & Rizzolatti, 2005). Therefore, we conducted a weight judgement task with point-light displays and showed participants videos of a model lifting and lowering three different weights. The experimental manipulation consisted of a goal change of these actions by showing the videos normal and in a time-reversed order of sequence. The results show a systematic overestimation of weights for time-reversed lowering actions (thus looking like lifting actions) while weight estimations for time-reversed lifting actions did not differ from the original playback direction. The results are discussed in terms of motor simulation and different kinematic profiles of the presented actions. © 2020, The Author(s).