Research shows that AI can predict 71% of active fund trading.

NightAirdropper · 2026-02-26T03:35:52+00:00

A study from Harvard Business School found that the behavior of active fund managers mostly follows patterns that can be predicted by machine learning, with approximately 71% of trading decisions being forecastable by models. However, the unpredicted 29% of trades are more closely associated with excess returns, suggesting that non-traditional investment activities may be more valuable.

NightAirdropper

2026-02-26 03:35:52

Abstract generation in progress

Gelonghui February 24 | A recent academic study led by professors from Harvard Business School found that most behaviors of active fund managers follow patterns that machines can learn. Researchers used a machine learning algorithm called “neural networks” to predict about 71% of mutual fund trading decisions, such as whether the fund manager buys, sells, or holds a particular stock within a quarter. The model was trained on five-year rolling window data from 1990 to 2023, extracting information including fund size, investor capital flows, stock characteristics, and broader economic conditions. Based on this, it can forecast most portfolio adjustments. Paradoxically, the limitations of this model may be more insightful than its successes. On average, the portion of trades the system failed to predict (about 29%) is more closely related to the fund’s excess returns. In other words, the trading activities outside of routine, detectable investment patterns seem to be where true value is created.

View Original

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.