UFCBench

Description

UFCBench is an environment for building machine learning models of UFC heavyweight fights and trading those models on historical betting markets. Agents develop ML strategies using historical fight data and detailed fighter statistics, place bets on fight outcomes, and manage bankroll across a 1-year window of UFC events.

Capabilities

• Developing machine learning models for UFC fight outcome prediction
• Analyzing detailed fighter statistics (striking, takedowns, submissions, knockdowns)
• Backtesting models against historical betting odds
• Bankroll management and bet execution

Compute Requirements

Agents are given a sandbox with file system access and scientific Python libraries (pandas, numpy).

Tasks

There is one split in this environment:

• Train: 3 scenarios

Each scenario covers a 1-year window of UFC heavyweight events.

Reward Structure

This is a dense, verifiable reward environment. Rewards occur after each event day. The reward is calculated as the difference in log wealth before and after betting, i.e:

$\log{W_{t+1}} - \log{W_{t}}$

No LLM graders are used -- reward is deterministic based on fight outcomes.

Data

Historical UFC heavyweight fight data including fighter names, betting odds, tournament info, and weight class. Detailed fighter statistics (significant strikes, takedowns, knockdowns, submission attempts, position-specific stats) are also available. Training data is mounted at /tmp/gr-datasets for agents to build models.

Tools

Agents get CLI tools (bash, read, write, grep, glob, ls, todo_write) plus 4 environment-specific tools:

Time Horizon

UFCBench is an open-ended, long-horizon environment where agents simulate a year of model development and betting across UFC heavyweight events.

Environment Difficulty

[Put environment difficulty statistics here]

Other Environment Requirements

There are no further environment requirements; UFCBench works out of the box with the OpenReward endpoint without any external API keys.

Safety

Agents in UFCBench are told to maximize their long-run bankroll growth. The environment does not present direct safety risks, as agents only interact with historical data through betting decisions on public odds.

There may be indirect risks, however, in that an agent that is taught to maximize long-run wealth may blindly follow this objective when tested in other environments, leading it to pursue unethical objectives. Our advice is that multi-environment training runs involving UFCBench should include other environments that teach agents to respect ethical norms so that the agent understands a broader category of objectives than just maximizing wealth.

Citation

@dataset{GRUFCBench,
  author    = {General Reasoning Inc. Team},
  title     = {UFCBench},
  year      = {2026},
  publisher = {OpenReward},
  url       = {https://www.openreward.ai/GeneralReasoning/UFCBench}
}