Machine learning techniques are used extensively in sports to help optimize athletes’ performance and maximize their chances of winning. Weightlifting requires quick and calculated decision-making by the coach and athlete to determine optimal load selection when progressing through lifting attempts during competition. When an optimal load-selection strategy is employed, the probability of success in competition is maximized. An XGBOOST machine learning model was developed to predict the success of weightlifting attempts in competition. The model was trained on an extensive dataset, including the performances of more than 14,000 athletes observed throughout 11 years of competition. We also included data from the 2024 Summer Olympic Games as a test dataset. The data contained information about the athletes’ sex, age, competition weight class, previous lifting performance, and previous competition placings. The accuracy of our predictions varied among the different weight classes, with a maximum accuracy of 89% for female athletes in the clean and jerk event. The outcome of our work is that, in general, the performance of female lifters is more predictable than males. Furthermore, the performance of the athletes competing in the lightest and heaviest weight classes for each sex category is the most predictable. Interestingly, prediction accuracy differs between lift attempts and lifting events performed during competition, with the initial attempt of the clean and jerk being the most predictable.