Predicting Olympic Medal Tally with XGBoost: A Novel Model for First-Time Medalist Probabilities and the Great Coach Effect

We developed an eXtreme Gradient Boosting (XGBoost) prediction model based on historical medal data, host country factors, participant numbers, and other non-economic variables. The model successfully forecasted Olympic medal totals from 1960 to 2024, achieving an outstanding performance with an NRMSE of 0.05 and R²of 0.9, outperforming Random Forest (RF) and Support Vector Machine (SVM) models. Using this model, we predicted the 2028 Olympic medal table and optimized it for countries that have never won medals, achieving an NRMSE of 0.0589 by incorporating the medal share of key nations. Our model also suggests a high probability (0.2655) that three countries will secure their first Olympic medal in 2028. Furthermore, we analyzed the Great Coach Effect with Entropy weight method(EWM) and XGBoost, identifying it as the most influential factor in medal success. Based on this, we highlighted three promising coaching investment opportunities. This model provides a novel perspective to study Olympic-related data and can also be utilized for calculating Olympic medal odds and guiding future investments in sports development.