Machine Learning Forecasting of Esports Prize Earnings: Evidence from 500 High-Participation Players (1998–2025)

In esports, cumulative prize earnings represent a key indicator of competitive success and economic value; however, most previous studies have primarily focused on match-outcome prediction, leaving economic forecasting relatively underexplored. This study investigates the feasibility of artificial intelligence–based models for predicting esports prize earnings using participation-related and prize-derived indicators. We analyzed a public EsportsEarnings dataset covering international competitions from 1998 to September 2025 and aggregated records for the 500 players with the highest number of international tournament entries. Per-event participation intensity and average prize features were constructed to represent long-term competitive engagement and unit performance. Traditional correlation and regression approaches were first examined as baseline models, followed by machine-learning methods, including support vector–based and ensemble-based algorithms, implemented for both regression and binary classification tasks. The findings indicate that simple single-variable relationships provide limited explanatory power for forecasting prize earnings, whereas multivariate machine-learning models substantially enhance predictive capability by capturing nonlinear interactions among participation and performance indicators. These results suggest that AI-driven approaches can offer a scalable and reproducible framework for forecasting economic performance in esports and may support strategic decision making for players, teams, and industry stakeholders.