Perceiving and reacting to external events in real time requires our brain to continuously generate predictions about upcoming events. These predictions can be examined more closely using a dynamic extension of Representation Similarity Analysis (dRSA). This approach uses temporally variable models of representational similarity to characterize the representational content at each time point of a temporally extended, unfolding event. dRSA allows testing if stimulus features such as position or motion of a moving object are represented in a predictive or delayed manner. Results of a previous Magnetencephalography (MEG)-based dRSA study that focused on human body movements demonstrated lagged and predictive representations of body part position and motion direction, respectively (de Vries & Wurm, Nature Communications, 2023). However, the precise influence of stimulus feature parameters, in particular the predictability of observed movements, on the timing and duration of lagged and predictive neural representations remains unclear. Therefore, in a pilot MEG study using more controllable stimuli, we aim at investigating how predictability influences at what time point the brain represents dynamic events: Participants watch 30-second videos of a moving dot while the likelihood of the dot maintaining its current direction at each time point is systematically varied. Applying dRSA, we capture the neural representations of stimulus features such as dot position and direction. We expect that these features will be represented before their actual occurrence for highly predictable dot motions whereas more unpredictable movements will be represented in a delayed manner.