AI Model Creates Profitable Algo Trading Strategy

Summary

Anthropic's new Fable 5 AI model has been tested for algorithmic trading strategy development. While hyped for its advanced capabilities, the focus here is on its performance in generating and refining trading strategies. The model was tasked with creating a trend-following strategy for ETHUSDT on a 30-minute timeframe, incorporating validation of entry rules, optimization, and Monte Carlo simulations to prevent overfitting. The model successfully generated three strategies, with one achieving a Sharpe ratio of 1.34, exceeding the target of 1.2, and demonstrating a favorable win-to-loss ratio. It identified a strategy with a simple breakout logic using Bollinger Bands and an EMA. However, the developer advises against using Fable 5 for this purpose due to its high cost and suggests that existing models like Opus or Sonnet can achieve similar results more affordably. The generated strategy is not recommended for immediate live trading without further out-of-sample testing.