The development of the gaming industry into digital formats has become a rapidly growing trend. E-Sports, particularly in Indonesia, has shown significant growth alongside technological advancements. The increased interest in E-Sports is evidenced by the higher quality tournaments organized by local game developers, such as Moonton, a subsidiary of ByteDance, hosting tournaments for Mobile Legends: Bang Bang. This article aims to analyze the probability of winning heroes in the Mobile Legends Professional League Season 12 using the Naive Bayes algorithm. The results of calculating the probabilities for various hero roles show varying levels of winning potential. By utilizing this method, it becomes possible to predict hero victories or losses more systematically, aiding players in developing more effective strategies during matches. The results obtained from predicting hero victories and losses indicate that for the jungler role, the win rate is 0.145 and the loss rate is 0.088. For midlaners, the victory rate reaches 0.492 and 0.661 for losses. As for roamers, the win rate is 0.120 and the loss rate is 0.102. For goldlaners and explaners, they achieve win rates of 0.528 and 0.177, respectively, while their loss rates are 0.339 and 0.132. Furthermore, after testing the data, the accuracy obtained for the roles is as follows: jungler role 67.61%, midlaner role 67.5%, roamer 67.65%, goldlaner 67.29%, and explaner 67.71%.