LLM-enhanced momentum investing combines traditional momentum signals with real-time news interpretation by large language models (LLMs). The idea is straightforward: stocks with strong past returns are candidates for momentum portfolios, but their inclusion and weight are refined by LLM-generated sentiment scores derived from firm-specific news. This hybrid approach improves risk-adjusted returns (Sharpe, Sortino) and is […]

R-bloggers

A tutorial demonstrating how to combine traditional momentum investing with LLM-generated sentiment scores using R. The strategy screens the top 20 S&P 500 stocks by YTD returns, queries Bing News for recent headlines, sends them to a Microsoft Azure AI Foundry-hosted LLM, and uses the resulting sentiment scores (0–1) to re-weight and narrow the portfolio to 10 high-conviction stocks. Key parameters include a 5-day lookback window, 5-day forecast horizon, and weekly rebalancing. Full R code is provided using tidyquant, httr, and gt packages. The resulting portfolio shows heavy energy sector concentration, attributed to geopolitical tensions boosting oil and gas sentiment.

Momentum Investing Enhanced by Microsoft Foundry-Hosted Large Language Model