Enhancing Trading Performance Through Sentiment Analysis with Large Language Models: Evidence from the S&P 500

2025-07-13

AI Summary

This paper addresses the limited adaptability of S&P 500 trading strategies under market volatility by proposing a dynamic enhancement framework that integrates news sentiment signals with traditional quantitative factors. Methodologically, it systematically combines FinBERT and GPT-2 to perform fine-grained sentiment modeling on real-time financial news, then fuses these sentiment embeddings with ARIMA/ETS time-series forecasts and momentum- and trend-based technical indicators—yielding a multi-source, heterogeneous feature set. An adaptive ensemble trading model is constructed, wherein weights are dynamically adjusted. Its key innovation lies in treating large language model–generated sentiment embeddings as an independent alpha source, complementary—not merely additive—to classical technical analysis. Empirical backtesting (2018–2023) shows the strategy achieves a 23.6% higher annualized return and an 18.4% reduction in maximum drawdown versus both buy-and-hold and pure technical strategies, with notably enhanced robustness during high-volatility regimes.

Technology Category

• Machine Learning: Ensemble Methods
• Natural Language Processing: Safety and Robustness
• Game Theory and Economic Paradigms: Adversarial Learning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

This study integrates real-time sentiment analysis from financial news, GPT-2 and FinBERT, with technical indicators and time-series models like ARIMA and ETS to optimize S&P 500 trading strategies. By merging sentiment data with momentum and trend-based metrics, including a benchmark buy-and-hold and sentiment-based approach, is evaluated through assets values and returns. Results show that combining sentiment-driven insights with traditional models improves trading performance, offering a more dynamic approach to stock trading that adapts to market changes in volatile environments.

Problem

• Optimizing S&P 500 trading using sentiment analysis and technical indicators
• Combining sentiment data with traditional models for better performance
• Adapting trading strategies to volatile market conditions dynamically

Innovation

• Combines sentiment analysis with technical indicators
• Integrates GPT-2 and FinBERT for real-time analysis
• Enhances trading with ARIMA and ETS models