# 195M steps of 10-agent CS:GO state → action data (with damage labels), free for research — retained record

> Retained reddit record from reddit:reinforcementlearning.

Canonical: https://highsignal.app/data/records/ca4b724bde713512

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:reinforcementlearning)  published 2026-10-03 15:02 ingested 2026-10-04 02:38

[original source ↗](https://www.reddit.com/r/reinforcementlearning/comments/1wwpe09/195m_steps_of_10agent_csgo_state_action_data_with/)

## retained text



```
If you work on imitation learning or multi-agent behaviour: we just released 9,515 bot-played CS:GO matches. Per 16 Hz step for each of ten agents: actions (buttons + mouse), state (pose, health, weapon…) and who-hit-whom damage. 218,768 rounds, ≈3,958 h of play. 14% of the matches also have the ten first-person videos, aligned frame by frame.

Expert bots on one map with pistols, so it is a clean, narrow testbed rather than human play. Parquet + webdataset videos, CC BY-NC 4.0: https://huggingface.co/datasets/v0rt3ch/cs-5v5-bots-aim\_usp-3k · spec and a round you can watch from all ten POVs: https://vor.tech/?utm\_source=reddit&utm\_medium=social&utm\_campaign=launch-2026-10&utm\_content=r-reinforcementlearning
```



document fetched 2026-10-04 00:34 · canonical https://reddit.com/r/reinforcementlearning/comments/1wwpe09/195m_steps_of_10agent_csgo_state_action_data_with
