Monopoly Bench is a lightweight, text-based simulation of the classic board game Monopoly, designed for benchmarking and analyzing agent-based strategies. This project allows LLMs to play Monopoly against each other, providing a controlled environment to test strategic decision-making capabilities.

• LLM Agent Support: Integrates with OpenAI API (GPT-4, o3, etc.) for AI players
• Condensed Board: Optimized 19-tile board to reduce game length and API costs
• Multiple Agent Types: Support for LLM agents, random agents, and custom strategies
• Detailed Logging: Comprehensive game logs for analysis and debugging
• Auction System: Full property auction mechanics when players skip purchases
• Trading & Building: Complete property trading and house-building mechanics

1. Setup Environment

   # Install UV package manager
   curl -LsSf https://astral.sh/uv/install.sh | sh
   
   # Clone and setup project
   git clone <repository-url>
   cd monopoly-bench
   uv sync

2. Configure OpenAI API

   # Create .env file with your API key
   echo "OPENAI_API_KEY=your_api_key_here" > .env

3. Run a Game

   uv run src/run_match.py

monopoly-bench/
├── src/
│   ├── run_match.py    # Main game runner
│   ├── engine.py       # Core game logic and rules
│   ├── agents.py       # AI agent implementations
│   ├── apis.py         # OpenAI API integration
│   ├── tools.py        # LLM function definitions
│   ├── config.py       # Game configuration
│   └── logger.py       # Game logging utilities
├── results/            # Game logs and results
└── README.md

Edit src/config.py to customize:

• Agents: Choose between LLM, Random, or custom agents
• Starting Cash: Default $1000 (optimized for faster games)
• Max Turns: Default 30 turns to prevent infinite games
• Board Layout: Condensed 19-tile board with core Monopoly mechanics

The simulation includes core Monopoly features:

• Property Management: Buy, mortgage, and trade properties
• House Building: Build houses on monopolies for increased rent
• Auctions: Properties go to auction when skipped
• Debt Resolution: Automatic bankruptcy handling
• Strategic Trading: LLM agents can propose and negotiate trades

=== MONOPOLY GAME LOG ===
--- TURN 1: PLAYER 0'S TURN ---
Player 0 | Cash: $1000 | Position: 0 | Properties: None
Phase: start_management_phase
Action: proceed
Player 0 rolled 8, moved to Baltic Avenue
Bought Baltic Avenue for $60

To add custom agents, extend the base Agent class in agents.py:

class CustomAgent(Agent):
    def act(self, observation):
        # Implement your strategy
        return {"type": "proceed"}

• Python 3.8+
• OpenAI API key (for LLM agents)
• UV package manager for dependencies