Scaling Strategy, Not Compute: A Stand-Alone, Open-Source StarCraft II Benchmark for Accessible Reinforcement Learning Research

2026-02-19

arXiv.org

AI Summary

Existing reinforcement learning research in StarCraft II lacks intermediate-complexity benchmarks that lie between full games and minigames, resulting in high training costs and difficulty isolating core tactical capabilities. To address this gap, this work proposes Two-Bridge, a lightweight intermediate benchmark that disables economic mechanics such as resource gathering, base construction, and fog of war, thereby focusing exclusively on long-range navigation and micromanagement combat. Built on PySC2, the benchmark provides a Gym-compatible interface and open-sources its maps, environment wrappers, and reference training scripts, facilitating curriculum learning and substantially reducing computational overhead. Preliminary experiments demonstrate that agents can learn coherent maneuvering and engagement strategies at significantly lower cost, confirming the benchmark’s effectiveness and practical utility.

Abstract

The research community lacks a middle ground between StarCraft IIs full game and its mini-games. The full-games sprawling state-action space renders reward signals sparse and noisy, but in mini-games simple agents saturate performance. This complexity gap hinders steady curriculum design and prevents researchers from experimenting with modern Reinforcement Learning algorithms in RTS environments under realistic compute budgets. To fill this gap, we present the Two-Bridge Map Suite, the first entry in an open-source benchmark series we purposely engineered as an intermediate benchmark to sit between these extremes. By disabling economy mechanics such as resource collection, base building, and fog-of-war, the environment isolates two core tactical skills: long-range navigation and micro-combat. Preliminary experiments show that agents learn coherent maneuvering and engagement behaviors without imposing full-game computational costs. Two-Bridge is released as a lightweight, Gym-compatible wrapper on top of PySC2, with maps, wrappers, and reference scripts fully open-sourced to encourage broad adoption as a standard benchmark.

Problem

• StarCraft II
• Reinforcement Learning
• benchmark
• real-time strategy
• curriculum design

Innovation

• StarCraft II
• intermediate benchmark
• reinforcement learning
• micro-combat
• long-range navigation