ChatGPT-enhanced Stock Momentum

Can insights inferred from real-time financial news by large language models (LLM) such as ChatGPT 4.0 mini enhance a conventional stock momentum strategy? In their October 2025 paper entitled “ChatGPT in Systematic Investing – Enhancing Risk-Adjusted Returns with LLMs”, Nikolas Anic, Andrea Barbon, Ralf Seiz and Carlo Zarattini investigate whether ChatGPT can improve a conventional momentum strategy applied to S&P 500 stocks by extracting predictive signals from minute-level Stock News API news articles. Specifically, they each month:

1. Rank stocks based on returns from 12 months ago to one month ago.
2. Construct an equal-weighted or value-weighted long-only momentum portfolio by buying stocks in the top 20% of rankings (top two deciles).
3. Ask ChatGPT to quantify the potential of each stock in the momentum portfolio to increase Sharpe ratio and suppress maximum drawdown (MaxDD), and weight each stock according to this signal.

They apply 0.02% trading frictions to portfolio changes to test net performance. Using daily total returns for S&P 500 stocks, relevant high-frequency Stock News API articles and the daily U.S. risk-free rate to perform model validation during October 2019 through December 2023 and out-of-sample testing during January 2024 through March 2025, they find that: (more…)