Can We Use Machine Learning To Predict Win/Loss Rates in Chess Using Meta-Game Data?

Artificial intelligence research in chess has traditionally centered on analyzing board positions and moves. While effective for engines, this perspective overlooks the wider context that shapes human play. Players face time pressure, emotional, and psychological influences beyond the board. This paper examines whether Machine Learning can predict win, loss, or draw outcomes using only meta-game data such as ELO differences, time management, and activity statistics. A Random Forest classifier trained on a large dataset achieved competitive accuracy, revealing correlations between contextual signals and results.