Photo Source: Matheus Bertelli
A team of researchers from Chinese universities and Washington University in St. Louis has found that ChatGPT significantly outperforms DeepSeek in predicting stock market trends, with nearly 30 years of data analysed. The study highlights both the potential of AI in financial forecasting and its significant limitations.
ChatGPT’s Accuracy in Predicting Market Trends
The study’s methodology was simple yet effective, relying on ChatGPT’s analysis of Wall Street Journal headlines and alerts from 1996 to 2022. Each month, the AI classified news into “good” or “bad” categories and compared sentiment shifts with stock market performance.
The findings were noteworthy—ChatGPT’s recognition of positive news led to accurate market predictions, with returns strong enough to offset transaction costs. Additionally, the AI achieved a higher Sharpe ratio, suggesting superior risk-adjusted returns,
“They also found ChatGPT particularly excelled at finding positive economic news during periods of economic downturns and rising economic policy uncertainty,” the researchers stated.
DeepSeek Falls Short in Financial Forecasting
DeepSeek’s performance fell short compared to ChatGPT, with the researchers suggesting that the model’s training on Chinese-language data could be a limiting factor in interpreting English financial news.
“DeepSeek, however, didn’t fare as well, which the researchers surmised was due to it being trained more on Chinese than English,” the study concluded.
Following these findings, the researchers are exploring further tests with DeepSeek on Chinese-language financial data to determine if its predictive power improves when using data in its native language.
Bad News Lacks Predictive Power
A fascinating finding from the study was that negative news did not reliably forecast market downturns. While positive news appeared to align with economic growth, bad news did not consistently signal market declines.
The researchers suggested that investor psychology played a key role, explaining that “Investors tend to pay more attention to bad news than good news, which is supported by the Wall Street adage that ‘markets take the stairs up and the elevator down’ and by index option trading where fund managers focus on hedging downside risk.”
Implications for AI in Financial Markets
The findings suggest that AI models like ChatGPT could be valuable for financial professionals trying to assess market sentiment. However, the study also stresses the importance of training AI in specific languages and financial environments to maximise effectiveness.
Future studies will examine whether ChatGPT’s performance holds in non-English markets, as research into AI-driven market prediction continues. Simultaneously, researchers are testing DeepSeek with Chinese-language financial data.
The study ultimately highlights both the potential and limitations of AI in stock market forecasting, with ChatGPT having a clear advantage over DeepSeek, but ongoing research being crucial to utilising AI’s full potential in global finance.