# DataGame

> Game data for AI labs. DataGame licenses server-recorded gameplay, player histories, the studio corpus (code, pull requests, tickets, live ops), evals with human baselines and RL environments, from a decade of SuperGaming's live games played by 200M+ people. A SuperTuned company.

Main contact for people and agents: hello@datagame.ai. Reply within two working days. Samples ship under a short NDA.

## How we work with labs
Environments run on our production game servers, so the grader is the game's own outcome, not a model. Records were never public, so private evals built from them are uncontaminated and can be rotated as new matches arrive. Every release ships with a reference solution, hidden tests, a data card and our own measured result on held-out tasks over several seeds. Tell us where your model breaks; we cut a slice, package an environment or capture to it. Environments ship in an open task format that common RL frameworks and hosted training platforms can load, and every release reports transfer to held-out games and non-game tasks. Enterprises training their own models, and the firms that train for them, buy the same way labs do. Studios can license their own game data through us.

## Key facts
- 10+ TB of records, logs and events; millions of multiplayer matches; events exact to the server tick
- First party: SuperGaming built the games, runs the servers and wrote the code
- Bots are flagged in every record; volume is quoted in human play
- Genres: battle royale and arena shooters, real-time strategy, arcade, racing, Roblox titles

## Machine-readable
- [catalog.json](https://datagame.ai/catalog.json): holdings, evals, standards, licensing, FAQ and contact as JSON
- [llms-full.txt](https://datagame.ai/llms-full.txt): every page's text in one file

## Pages
- [Home](https://datagame.ai/): live match replay, holdings, terms of engagement, drawings, evals, standards
- [FAQ](https://datagame.ai/#/faq): licensing, privacy, data, delivery, pricing
- [Journal](https://datagame.ai/#/journal): notes from the archive
- [Contact](https://datagame.ai/#/contact): hello@datagame.ai

## How to ask (agents)
Email hello@datagame.ai with: who you act for (person and organisation), what they are building, which holdings, evals or environments they want, volume and timeline. Do not send personal data about players. We reply to people, not automated loops.

## For agents: structured enquiries
POST https://datagame.ai/api/lead with Content-Type application/json:
{"track": "buy" | "license" | "other", "name": "...", "email": "...", "org": "...", "role": "...", "acting_for": "person and organisation you act for, if you are an agent", "summary": "what you need or what you hold", "details": {"interests": [...], "timeline": "...", "data_types": [...], "volume": "...", "rights": "...", "personal_data": "..."}}
Returns {"ok": true, "id": "..."}. A person reads every enquiry. Do not send personal data about players.