Summary

Alpha Arena is an experimental platform enabling AI models, particularly Large Language Models (LLMs), to engage in real-time trading within simulated and live financial markets. It addresses the challenge of evaluating LLMs' ability to develop sustainable, profitable trading strategies in uncertain environments. The platform targets AI researchers and quantitative developers, offering a reproducible, comparative framework to assess and evolve AI trading agents' performance, risk control, and decision-making capabilities.

How It Works

The system employs an orchestrator that periodically fetches market data (K-lines, order book) to construct unified prompts for multiple LLMs. Each LLM generates structured trading decisions in a standardized JSON format. A dedicated gateway manages LLM API interactions, while an exchange adapter facilitates trade execution, starting with paper trading and expandable to live exchanges. A portfolio module tracks individual accounts, PnL, and risk metrics, with a dashboard visualizing performance. This approach uniquely positions LLMs directly within a live trading loop, emphasizing comparability and reproducibility across different models.

Quick Start & Requirements

• Installation: Clone the repository, install dependencies via pip install -r requirements.txt, configure API keys in .env, and run python main.py.
• Prerequisites: Python 3.11, PostgreSQL, Redis, FastAPI, pandas, asyncio, LLM API access (OpenAI, DeepSeek, Anthropic, Google, Qwen), and Exchange API access (Bitget, OKX, CCXT).
• Links: Repository: https://github.com/AmadeusGB/alpha-arena.git.

Highlighted Details

• Real-time Market Simulation: Leverages live market data (60-min K-lines, order book) for AI decision-making.
• Multi-LLM Comparative Analysis: Supports simultaneous trading and performance comparison of 2-6 LLMs under identical conditions.
• Reproducible Trading Loop: Ensures full traceability of prompts, market snapshots, decisions, and execution outcomes.
• Flexible Execution: Starts with paper trading (simulated with fixed slippage) and supports integration with live exchanges.
• Strict Decision Schema: LLMs must adhere to a precise JSON format for trading instructions.

Maintenance & Community

No specific details regarding maintainers, community channels (e.g., Discord, Slack), or sponsorship are provided in the README. The project outlines a clear 5-7 day implementation plan for its MVP.

Licensing & Compatibility

The README does not specify a software license. This omission prevents an assessment of its terms for use, modification, and distribution, particularly concerning commercial applications or integration into closed-source systems.

Limitations & Caveats

The MVP is limited to spot trading, long-only positions, market orders with fixed slippage, and a 5-minute decision cadence. The project is in an experimental alpha stage (v0.1.0). Reliance on LLM APIs introduces potential timeouts and schema adherence issues. The absence of a defined license is a critical adoption blocker.