Lineup Mining and Balance Analysis of Auto Battler

Auto battler is currently one of the most popular types of video games on the market. In such type of games, players set up their own lineup of characters which combat automatically with other competitors. There are numerous kinds of chess pieces, lineups and bonuses in the game, which make the balance problem of game design extremely critical and challenging. In this paper, we use Chess Rush to investigate the methodology of evaluating the strength standards of lineups in auto battler games, and build a system for implementing such measurement. A neural network model is firstly trained for quick strength evaluation of the lineup. The genetic algorithm is then adopted and modified for the purpose of lineup mining of auto battler. Finally, strong lineups can be obtained under certain constraints. In this way, game developers can complete a balanced analysis before the game goes online. Moreover, the proposed approach can be generalized to balance analysis tasks of other games.