Implementation of Random Forest Algorithm Based Medal Count Prediction

Shuai Chen, Junjie Wang, Yuhang WuTransactions 2025 · DOI

TL;DR

Researchers have developed a multi-stage machine learning framework to predict the 2028 Los Angeles Olympic medal table, combining Random Forest Regression with association rule mining. Beyond just predicting winners, the study quantifies the "Great Coach" effect and identifies which countries are most likely to win their first-ever Olympic medal.

What Problem Does This Paper Address?

Predicting Olympic success is a classic challenge in sports analytics, usually involving a messy mix of economic data (GDP), demographics (population), and historical performance. However, traditional linear models often fail to capture the complex, non-linear interactions between these variables—such as the "host country effect" or the sudden impact of a world-class coaching hire.

The authors aim to move beyond simple trend lines. They address three specific questions:

  1. Can we accurately forecast the 2028 medal distribution using ensemble learning?
  2. Which sports are most strongly associated with a high total medal count?
  3. Can we statistically "prove" the impact of elite coaching on a nation's performance?

Methods

The paper employs a "hybrid" analytical approach:

  • Random Forest Regression (RFR): This is the workhorse of the study. Unlike a single decision tree, RFR builds a "forest" of trees to handle non-linear feature interactions and reduce noise. It was used to predict gold and total medal counts based on variables like athlete count, historical data, and host status.
  • Apriori Algorithm: Typically used in retail to find "frequently bought together" items, the authors used this to mine association rules between specific sporting events and medal distributions.
  • Mann-Kendall (MK) Mutation Test: This statistical test is usually found in climatology to detect shifts in time-series data. Here, it was repurposed to identify "mutation points" in a country's medal history to see if performance spikes correlate with the tenure of "great coaches."
  • Bootstrap Method: Used to calculate 95% confidence intervals, providing a measure of uncertainty for the 2028 predictions.

Datasets

The study utilizes historical Olympic data, though it does not name a specific public repository (like the Kaggle Olympic dataset). The features extracted include:

  • Predictor Variables: Proportion of winning athletes, total athletes participating, historical medal counts, number of sports events, and a binary "host country" variable.
  • Coaching Data: A binary variable indicating the presence of a "great coach" in a given year for specific countries.

Key Insights and Ideas

  • The "Anchor" Sports: Using the Apriori algorithm, the researchers found that Athletics, Badminton, Table Tennis, and Tennis are highly correlated with overall medal success. For top-tier nations, dominance in these specific "high-yield" areas is a prerequisite for leading the table.
  • Quantifying the "Great Coach" Effect: The study treats elite coaching as a measurable variable. Interestingly, the impact isn't uniform. China and India showed extreme sensitivity to the "positive influence of coaching," whereas France’s performance remained relatively stable regardless of specific coaching hires, suggesting a more decentralized or systemic athletic infrastructure.
  • Predicting the "First-Timers": The model identifies countries currently in a "medal deficit" that are poised for a breakthrough. Kyrgyzstan and Malaysia are flagged as the most likely to secure their first gold in 2028, with 80% and 70% probability respectively.

Key Results

  • Model Accuracy: The Random Forest model achieved an $R^2$ of 0.949 for gold medals and 0.969 for total medals, meaning it explains roughly 95-97% of the variance in the test data. The Root Mean Square Error (RMSE) was impressively low (1.23 for gold).
  • 2028 Forecast: The model predicts the USA (110 total) and China (91 total) will continue to lead the pack in Los Angeles. Great Britain (61) and France (54) are expected to see boosts, while Japan and the Netherlands may see their counts deteriorate compared to previous games.
  • Coaching Validation: The MK mutation test successfully identified performance shifts in Chinese volleyball, American table tennis, and Japanese swimming that aligned with the introduction of elite coaching programs.

Limitations and Future Work

The authors acknowledge that while their model is robust, it is still a "black box" in some respects. It currently does not account for:

  • Athlete Psychology: The mental state and individual readiness of athletes are not captured in the data.
  • Technological Innovation: Sudden leaps in equipment (e.g., "super shoes" in track) or training methodologies are difficult to quantify.
  • Geopolitical Factors: Sudden changes in funding or political stability that affect sports programs were not explicitly modeled.

Future research aims to integrate more dynamic factors, such as real-time training data and psychological assessments, to move toward an even more granular level of prediction.

Summary updated

Abstract

This paper proposes a data trend prediction model based on random forest regression, focusing on the comprehensive use of machine learning and association rule mining in complex data prediction. Firstly, considering various factors, we constructed a medal prediction model based on random forest regression, assessed the accuracy of data, predicted the medal list of the 2028 Olympic Games and the first medal prediction results of the non-winning countries, and at the same time, we used the Apriori algorithm to analyze the relationship between the setting of the competition events and the distribution of the medals. Then, a regression model was constructed with “coaching effect” as the main variable, and the coefficients of “coaching effect” were judged according to the contribution degree of the coefficients. By applying the MK mutation test, we select and evaluate the changes in the number of medals of the three countries after investing in “great coaches”. The model analyzes and validates the impact of great coaches on sports performance through a predictive model with excellent accuracy assessment obtained from Random Forest Regression.
BibTeX
@article{chen2025implementation,
  author = {Shuai Chen and Junjie Wang and Yuhang Wu},
  title = {{Implementation of Random Forest Algorithm Based Medal Count Prediction}},
  journal = {Transactions},
  year = {2025},
  doi = {10.62051/mddr0e54},
}
Figure