Google Research Football
Physics-based 11-vs-11 football RL environment for training and benchmarking game-playing agents, from the Google Brain team.
Google Research Football
Google Research Football (GFootball) is a reinforcement-learning environment where agents learn to play a full 11-versus-11 game of football (soccer) in an advanced, physics-based 3D simulator. Released by the Google Brain team, it provides a challenging benchmark for research on sample-efficient RL, multi-agent coordination, and long-horizon planning.
Key features
- Physics-based 3D football simulator with configurable difficulty and stochasticity
- Football Academy: a set of progressively harder scenarios (run-to-score, corner, 3-vs-1) for curriculum learning
- OpenAI Gym-compatible API with multiple observation representations (pixels, super-mini-map, floats)
- Built-in reward shaping (scoring plus checkpoints) and single- or multi-agent control
- Reproducible baselines (IMPALA, PPO) with support for self-play and league training
Because a match demands cooperation, opponent modeling, and emergent strategy, GFootball is a rich testbed for NPC behavior and for playtesting AI opponents in an adversarial game setting.
Curated mirror of the open-source Google Research Football (Apache-2.0). Get it from the source.
Compromise signals — malicious or tampered code (leaked secrets, backdoors, a dropped executable) — reduce the score, and known dependency CVEs carry a bounded penalty (they warrant review but never QUARANTINE — update the dependency to clear). Other dangerous-by-capability traits are risk surface, expected for some capabilities. Every finding is mapped to its OWASP control below.
Findings mapped to the OWASP Top 10 for LLM Applications (2025) and the OWASP Machine Learning Security Top 10. Expand any flagged control for the exact findings — compromise reduces the score; expected/risk-surface do not, except a known CVE, which carries a small bounded penalty (high/critical → Review).
The same gate an agent runs before installing (POST /api/v1/trust/google-research-football-rl/check). Click a policy:
Consume Google Research Football programmatically. Authenticate with an API key or session — see Authorize an agent.
# Agents: CHECK BEFORE YOU INSTALL (no auth) — score, grade, level, capability manifest
curl https://ai-supply.store/api/v1/trust/google-research-football-rl
# Gate against your org policy (returns { pass, violations })
curl -X POST https://ai-supply.store/api/v1/trust/google-research-football-rl/check \
-H "Content-Type: application/json" \
-d '{"minGrade":"B","denyPermissions":["shell"],"denyUnknownEgress":true}'
# CLI
npx ai-supply add google-research-football-rl
# REST (install → download)
curl -X POST https://ai-supply.store/api/v1/listings/google-research-football-rl/install \
-H "Authorization: Bearer $AIM_KEY"
# MCP tool
install_listing({ "slug": "google-research-football-rl" })OpenAPI spec →Curated mirror — latest upstream source. See the repository for tagged releases.