This paper studies the strategies used in StarCraft II, a real-time strategy game (RTS) wherein two sides fight against each other in a battlefield context. We propose an approach which automatically classifies StarCraft II game-log collections into rush and non-rush strategies using a support vector machine (SVM). To achieve this, three types of features are evaluated: (i) the upper bound of variance in time series for the numbers of workers, (ii) the upper bound of the numbers of workers at a specific time, and (iii) the lower bound of the start time for building the second base. Thus, by evaluating these features, we obtain the optimal parameters combinations.