A Quantitative Study of Medal Predictions and the Impact of Good Coaching at the Los Angeles Olympics

Abstract

Focusing on the 2028 Los Angeles Summer Olympic Games, this study aims to build an accurate and effective medal prediction model and deeply explore the key factors affecting the distribution of MEDALS to provide strong support for countries to formulate scientific and reasonable Olympic strategies. First, we cleaned the raw data, dealt with outliers and introduced key variables, improved the data quality through normalization, and built a solid data foundation. Then, we adopted an integrated learning approach, combining multiple machine learning algorithms such as Logistic Regression, random forest, support vector machine, ridge regression, and XGBoost, to construct a stacked model for predicting whether a country will win medals and the number of medals. After cross-validation and grid search tuning, the model performance was evaluated in detail, and the combined prediction accuracy was 97.49%, with good stability of the prediction results. We analyzed each country’s performance and found that most countries had shown positive development while predicting that the 69 countries that had not yet won a medal were expected to make a breakthrough. In addition, we established a time series regression model. The influence of coaches is prominent when they are first transferred to the team, and the improvement is mostly more than 35% when they are stabilized. Finally, we performed sensitivity analysis on the model using Monte Carlo simulations visualized and presented to indicate that the model performs well. This research provides a scientific basis for the National Olympic Committees to formulate Olympic strategies, plan athlete training, and optimize training programs, helping countries take advantage of their strengths and improve their performance in the Olympic Games.