mlsanityData to 7 Oct 2026

Balancing of competitive two-player Game Levels with Reinforcement Learning

arXiv 2306.04429 ↗First posted 7 Jun 2023cs.LG cs.GTcited by 4

Not on the map. Why?

Abstract (arXiv)

The balancing process for game levels in a competitive two-player context involves a lot of manual work and testing, particularly in non-symmetrical game levels. In this paper, we propose an architecture for automated balancing of tile-based levels within the recently introduced PCGRL framework (procedural content generation via reinforcement learning). Our architecture is divided into three parts: (1) a level generator, (2) a balancing agent and, (3) a reward modeling simulation. By playing the level in a simulation repeatedly, the balancing agent is rewarded for modifying it towards the same win rates for all players. To this end, we introduce a novel family of swap-based representations to increase robustness towards playability. We show that this approach is capable to teach an agent how to alter a level for balancing better and faster than plain PCGRL. In addition, by analyzing the agent's swapping behavior, we can draw conclusions about which tile types influence the balancing most. We test and show our results using the Neural MMO (NMMO) environment in a competitive two-player setting.

Key facts. Cited by 4 papers in the map (data through 7 Oct 2026). Builds on 15 of its 30 references that are in the map. First posted on arXiv 7 Jun 2023.

Builds on

15 of 30 references.

References · 15

  1. 2017Proximal Policy Optimization Algorithmscited by 14,441
  2. 2017Procedural Content Generation via Machine Learning (PCGML)cited by 136
  3. 2020PCGRL: Procedural Content Generation via Reinforcement Learningcited by 90
  4. 2020Deep Learning for Procedural Content Generationcited by 73
  5. 2019Neural MMO: A Massively Multiagent Game Environment for Training and Evaluating Intelligent Agentscited by 42
  6. 2021Learning Controllable Content Generatorscited by 40
  7. 2021Adversarial Reinforcement Learning for Procedural Content Generationcited by 36
  8. 2020TOAD-GAN: Coherent Style Level Generation from a Single Examplecited by 22
  9. 2022Learning Controllable 3D Level Generatorscited by 19
  10. 2020Learning to Generate Levels From Nothingcited by 17
  11. 2018DOOM Level Generation using Generative Adversarial Networkscited by 15
  12. 2021World-GAN: a Generative Model for Minecraft Worldscited by 14
3 more
  1. 2022Path of Destruction: Learning an Iterative Level Generator Using a Small Datasetcited by 14
  2. 2022Mutation Models: Learning to Generate Levels by Imitating Evolutioncited by 9
  3. 2022Generating Real-Time Strategy Game Units Using Search-Based Procedural Content Generation and Monte Carlo Tree Searchcited by 5

Built on by

Cited by year

Citing papers per year: 2024: 2; 2025: 2.

4 papers.

  1. 2024G-PCGRL: Procedural Graph Data Generation via Reinforcement Learningcited by 4
  2. 2024Identifying and Clustering Counter Relationships of Team Compositions in PvP Games for Efficient Balance Analysiscited by 0
  3. 2025Simulation-Driven Balancing of Competitive Game Levels with Reinforcement Learningcited by 0
  4. 2025Level the Level: Balancing Game Levels for Asymmetric Player Archetypes With Reinforcement Learningcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024Personalized Dynamic Difficulty Adjustment -- Imitation Learning Meets Reinforcement Learning
  2. 2026RuleSmith: Multi-Agent LLMs for Automated Game Balancing
  3. 2026Play Like Champions: Counterfactual Feedback Generation in Latent Space
  4. 2025Enhancing Player Enjoyment with a Two-Tier DRL and LLM-Based Agent System for Fighting Games
  5. 2023AlphaStar Unplugged: Large-Scale Offline Reinforcement Learning
  6. 2025Self-correcting Reward Shaping via Language Models for Reinforcement Learning Agents in Games
  7. 2021Policy Fusion for Adaptive and Customizable Reinforcement Learning Agents
  8. 2025SPIRAL: Self-Play on Zero-Sum Games Incentivizes Reasoning via Multi-Agent Multi-Turn Reinforcement Learning