Uncovering the Secret of Olympics Medals

Abstract

This study aims to predict Olympic medal counts and analyze factors influencing national competitiveness in the Olympics. Firstly, a Two-stage Random Forest Model was developed using multiple random forest regressions to forecast the number of medals and gold medals for each country in the 2028 Olympics. The model achieved an average R² of 0.985 on the test set, demonstrating high prediction accuracy. Additionally, an XG-BOOST Classifier was employed to identify countries most likely to achieve a medal breakthrough from zero to one. Secondly, a Cluster-OLS Prediction Model was established to explore the relationship between changes in events and medal counts. The concept of "market share" was applied to represent a country's project advantages, and clustering algorithms were used to group countries. The regression coefficients revealed that events with greater suspense had higher coefficients, indicating their importance for national competitiveness. The Host Effect was also included as a positive indicator in the model. Thirdly, a PSM-DID Model was used to quantify the impact of "great" coaches. The DID value was calculated as 7.9619, confirming the significant effect of coaching. Based on this, China, France, and the United States were recommended to invest in ice hockey, basketball, and volleyball, respectively, to maximize the benefits of effective coaching. Finally, the study suggests that countries should balance training resources for athletes of different genders based on regional culture and project advantages to enhance their Olympic preparations.