Predicting Dota 2 Game Outcomes Using Logistic Regression and Decision Tree Models

This study presents a comparative analysis of two machine learning models, decision tree and logistic regression, in predicting the outcomes of Dota 2 matches. Through rigorous evaluation using accuracy, precision, recall, and $\mathbf{F} 1$-score metrics, the study identifies the salient features influencing game results, including economic factors and strategic gameplay elements. The decision tree model exhibits a slight edge in overall accuracy and sensitivity towards positive outcomes, while logistic regression shows balanced predictive capabilities across both winning and losing instances. The findings reveal a nuanced understanding of each model’s strengths, suggesting their potential application in gaming analytics. With a focus on model performance in a complex, multifactorial environment, this study contributes to the strategic understanding and forecasting within competitive gaming domains.