Artificial intelligence can predict whether fund managers will buy, sell, or hold stocks with an accuracy rate exceeding 70%, according to new research.
A team led by Professor Lauren Cohen at Harvard Business School recently published findings based on analysis of 33 years of trading data, Bloomberg reported on the 24th (local time).
The researchers trained an AI model on actual fund trading records and economic conditions from 1990 to 2023, then evaluated how well it predicted fund managers' trading decisions. The AI achieved a 71% accuracy rate. Bloomberg noted that "for some managers, the model correctly predicted most of their trading decisions in certain quarters."
However, this does not mean AI has completely "decoded" the market, Bloomberg reported. The AI showed particular weakness in predicting stocks that defied expectations and delivered outsized returns. The researchers explained that the remaining 29% of incorrect predictions mainly involved decisions surrounding "outperform" stocks that broke from market conventions and patterns.