Territory Paint Wars: Diagnosing and Mitigating Failure Modes in Competitive Multi-Agent PPO
Abstract
We present Territory Paint Wars, a minimal competitive multi-agent reinforcement learning environment implemented in Unity, and use it to systematically investigate failure modes of Proximal Policy Optimisation (PPO) under self-play. A first agent trained for episodes achieves only win rate against a uniformly-random opponent in a symmetric zero-sum game. Through controlled ablations we identify five implementation-level failure modes -- reward-scale imbalance, missing terminal signal, ineffective long-horizon credit assignment, unnormalised observations, and incorrect win detection -- each of which contributes critically to this failure in this setting. After correcting these issues, we uncover a distinct emergent pathology: competitive overfitting, where co-adapting agents maintain stable self-play performance while generalisation win rate collapses from to . Critically, this failure is undetectable via standard self-play metrics: both agents co-adapt equally, so the self-play win rate remains near throughout the collapse. We propose a minimal intervention -- opponent mixing, where of training episodes substitute a fixed uniformly-random policy for the co-adaptive opponent -- which mitigates competitive overfitting and restores generalisation to (, seeds) without population-based training or additional infrastructure. We open-source Territory Paint Wars to provide a reproducible benchmark for studying competitive MARL failure modes.
Keywords
Cite
@article{arxiv.2604.04983,
title = {Territory Paint Wars: Diagnosing and Mitigating Failure Modes in Competitive Multi-Agent PPO},
author = {Diyansha Singh},
journal= {arXiv preprint arXiv:2604.04983},
year = {2026}
}
Comments
16 pages, 5 figures