Discovering Combos in Fighting Games with Evolutionary Algorithms

In fighting games, players can perform many different actions at each instant of time, leading to an exponential number of possible sequences of actions. Some of these combinations can lead to unexpected behaviors, which can compromise the game design. One example of these unexpected behaviors is the occurrence of long or infinite combos, a long sequence of actions that does not allow any reactions from the opponent. Finding these sequences is essential to ensure fairness in fighting games, but evaluating all possible sequences is a time consuming task. In this paper, we propose the use of an evolutionary algorithm to find combos on a fighting game. The main idea is to use a genetic algorithm to evolve a population composed of sequences of inputs and, using an adequate fitness function, select the ones that are more suitable to be considered combos. We performed a series of experiments and the results show that the proposed approach was not only successful in finding combos, managing to find unexpected sequences, but also superior to previous methods.