Profit-Triggered Exit Strategy Integrating Financial News Sentiment with RSI and Moving Average Filters: Evidence from NASDAQ-100

Abstract

Predicting stock price movements remains a fundamental challenge due to the non-linear and dynamic nature of financial markets. While transformer-based sentiment models show strong performance in financial text classification, their direct utility as trading signal generators integrated with technical indicators remains underexplored. This study proposes an algorithmic trading framework integrating daily financial news sentiment from four pre-trained transformer models with RSI and moving average filters to generate buy signals on 30 NASDAQ-100 stocks over a 2-year simulation. News sentiment was extracted from 57,434 articles (February 2024–February 2026) and aggregated daily via majority voting. Buy signals were triggered when RSI fell below 40, price pulled back 0.5–5% below MA20, and close remained within 10% above MA50, with entry at the open of day i+1 and exit at the first profitable close. Four models were compared: ProsusAI/finbert, nlptown/bert-base-multilingual-uncased-sentiment, yiyanghkust/finbert-tone, and soleimanian/financial-roberta-large-sentiment, evaluated as trading signal generators. All four produced positive cumulative returns ranging from 125.22% to 188.09%,  and sensitivity analysis confirms profitability under transaction costs of up to 0.3% per trade. yiyanghkust/finbert-tone achieved the highest final return (188.09%) with 159 trades, while nlptown/bert-base-multilingual-uncased-sentiment achieved the highest Sharpe Ratio (2.78) and Average Daily Return (0.28%). Maximum Drawdown ranged from -22.19% to -25.03%. Notably, the generic multilingual nlptown outperformed domain-specific models in risk-adjusted metrics, suggesting that broad linguistic patterns in financial news can be captured without domain-specific pre-training. These findings demonstrate that integrating news sentiment with technical filters yields profitable, risk-efficient trading signals, with model selection depending on investor priorities.