Multiplayer online battle arena (MOBA) is currently one of most popular game genres in game artificial intelligence (AI). In this paper, we propose a method for predicting game outcome in MOBA games. Firstly, we extract features that contain the properties of the game outcome from game replay logs, taking into account game time. And then we implement a model for predicting the game outcome and interpret which features are important factors in the model. Experimental results show that the proposed method has high accuracy enough to predict the game outcome at a certain time point. It also analyzes the main play factors of game outcome for each game time zone.