FighterDDA: A Simulation Testbed for Evaluating Director-Based Dynamic Balancing

Abstract

While AI Directors have been widely studied, less attention has been paid to how different director styles can coexist in a single game environment with divergent design goals. This paper presents FighterDDA, a simulation testbed designed to evaluate dynamic balancing strategies aimed at different audiences using varied AI Directors in turn-based role-playing games. The testbed can rapidly simulate thousands of games, output formatted data, and provide visualizations of play traces for designers to evaluate whether a director strategy was successful. In an initial evaluation with the system, we found that a director made to even out win rates between differently-skilled players as well as a director to tune the overall difficulty for equally skilled players both showed success. The system demonstrates how a simulated testbed with data reporting supports prototyping and evaluating varied director systems.