Quantitative Analysis of Player Win Rate Prediction in Magic Chess Using the Random Forest and XGBoost Algorithms

Apriliano Ernando Benyamin Boimau (Universitas Amikom Yogyakarta)
Nafiatun Sholihah (Universitas Amikom Yogyakarta)
Ferian Fauzi Abdulloh (Universitas Amikom Yogyakarta)
Majid Rahardi (Universitas Amikom Yogyakarta)
Afrig Aminuddin (Universitas Amikom Yogyakarta)

Formosa Journal of Computer and Information Science
Vol. 5 No. 2 (2026): August 2026

This condition makes players rely on intuition and choose speculative tactics without a measurable basis. This study applied quantitative analysis to predict the probability of winning by comparing Random Forest and XGBoost. A total of 517 match data for season 4 was collected semi-manually, then through cleaning, binary class labeling, encoding, and 80:20 training and test data sharing. The model was optimized using Grid Search and evaluated in 30 iterations based on Accuracy, Precision, Recall, and F1-Score. The results show that Random Forest is optimal and stable, with an Accuracy of 53.27% and a Precision of 57.12%. Feature Importance shows that Commander and Go-Go Commander are the most dominant in winning over synergy compositions, and are implemented in web prototypes.