We present a novel approach using reinforcement learning to create balanced game levels for players with different abilities.
Our work was accepted for publication at the International Conference on the Foundations of Digital Games (FDG 2025). The paper, titled “Level the Level: Balancing Game Levels for Asymmetric Player Archetypes With Reinforcement Learning,” addresses a challenging problem in game development: creating balanced gameplay experiences for players with vastly different abilities.
The Challenge of Asymmetric Game Balance
Game balancing traditionally requires extensive manual effort and countless hours of human playtesting, especially when dealing with asymmetric multiplayer content where different player types have distinct advantages and disadvantages. We tackle this challenge by framing game balancing as a procedural content generation problem.
Innovative Approach Using AI
The paper builds upon and extends the recently introduced PCGRL (Procedural Content Generation via Reinforcement Learning) framework to automatically generate balanced levels tailored to asymmetric player archetypes. The key innovation is that the disparity in player abilities is balanced entirely through intelligent level design rather than modifying the players themselves. This means that while one archetype may inherently have advantages over another, the level design ensures both players have equal chances of winning.
Evaluation and Results
We evaluated our method across four different player archetypes and demonstrated its superior ability to balance a larger proportion of levels compared to two baseline approaches. Interestingly, the results reveal an important insight: as the disparity between player archetypes increases, the required number of training steps grows, while the model’s accuracy in achieving balance may decrease.
This work represents a step forward in automated game development tools, potentially reducing the significant manual effort and extensive playtesting currently required to create engaging, balanced gameplay experiences.
Citation: Florian Rupp, Kai Eckert (2025): Level the Level: Balancing Game Levels for Asymmetric Player Archetypes With Reinforcement Learning. In Proceedings of the 20th International Conference on the Foundations of Digital Games, FDG ‘25.