Gymnastics is associated with a high rate of injuries, particularly to the back. This is due to the fact that landing involves strong ground reaction forces as well as important body rotations, hence as a result of repetitive landing, these movement patterns can lead to significant muscle fatigue, which increases the risk of injury (Campbell et al., 2019). Fatigue influences kinematics during repetitive movements such as drop jumps or create lifting by decreasing hip and knee flexion angles and increasing trunk instability through a greater shift in centre of mass (Kazemi et al., 2021). Under laboratory conditions, muscle fatigue is usually measured by EMG (Thrope et al., 2017). However, incorporating EMG measurement into everyday training is difficult, it requires a fair amount of money, proper setup time and data processing. Thus, considering that fatigue affects kinematics, motion capture seems to be another possible way to measure athlete’s fatigue. Coupling this tool with Machine Learning models would also reduce analysis time. This would further reduce the workload of the athlete. The main objective of this study is to predict the fatigue in the lumbar region during landing repetitions only using the kinematic data of the athlete. A second objective of our research is to identify the best machine learning model for our data and the variables that would most predict fatigue during gymnastics landing. In agreement with the literature, we think that fatigue will lead to a decrease in the knee and hip flexion angles and in the displacement of centre of mass.