<![CDATA[This study proposes a machine learning-based framework for mid-season Grand Final prediction in the Mobile Legends Professional League Indonesia (MPL ID) using historical regular-season performance data. Unlike conventional esports predictions that rely on subjective analysis or post-playoff evaluation, this research formulates Grand Final qualification as a binary classification problem based solely on pre-playoff statistical indicators. Team-season-level data from Seasons 10 to 16 were aggregated, with Seasons 10–15 used for training and Season 16 reserved for testing to simulate realistic future-season forecasting. Four machine learning models were evaluated: Logistic Regression, Random Forest, Support Vector Machine (SVM), and XGBoost. Although SVM and XGBoost achieved higher accuracy (88.88%), Logistic Regression demonstrated superior discriminative capability (AUC 92.85%) and the highest cross-validation stability (88.36%). Feature importance analysis identified Regular Season Rank, Clutch Factor, and Match Loss as the most influential predictors. The results indicate that structured historical aggregation combined with interpretable probabilistic modeling enables reliable estimation of Grand Final qualification before playoff brackets are formed, advancing data-driven esports analytics through an early-stage predictive framework grounded in measurable competitive indicators.]]>
Copyrights © 2026
