Multi-Agent LLM Trading Framework Outperforms Buy & Hold in BTC Backtesting

A new study has shed light on the potential of Multi-Agent LLM trading frameworks in cryptocurrency markets. Researchers at an exchange research institute developed a framework that simulates collaboration among analysts, researchers, traders, and risk control teams. This approach allows for more transparent decision-making and enhanced risk control.

The framework, based on the TradingAgents framework, was backtested on one-hour BTC/USDT data and achieved a +20.25% total return in the testing period. In contrast, the Buy & Hold strategy resulted in a -7.89% loss over the same period. The maximum drawdown for the Multi-Agent framework was also significantly lower at -17.41%, compared to -27.06% for Buy & Hold.

The study's findings suggest that the Multi-Agent LLM framework has potential applications in crypto trading scenarios. However, further validation is needed to ensure the strategy's stability and generalizability across different market conditions and asset classes.