Lessons from Testing an Evolutionary Automated Game Balancer in Industry

Game balancing has been an important area of academic research in the past few years, with various methods of approaching the task being proposed. At this point in time, however, industry impact has been minimal, with these approaches appearing overwhelming or expensive to game designers and developers. The work presented in this paper takes one of the algorithms and approaches previously researched and, in cooperation with a commercial games studio, defines a specification language and tests its applicability in a real world scenario. The game being balanced in this paper is of a different genre to those previously targeted by the algorithm, with vastly different mechanics and expectations of what is considered a balanced state. Results indicate both great potential of the research area in general, but also highlight challenges to achieving mainstream use in the real world of balancing techniques.