Study finds AI trading strategies underperform buy-and-hold investing over 20-year period

A Harvard-led academic study examining AI trading strategies over a 20-year period has found that algorithmic approaches systematically underperform passive buy-and-hold investing strategies, presenting a significant challenge to the widespread institutional embrace of artificial intelligence in portfolio management. The research aligns with broader findings on active management performance, where empirical data consistently demonstrates that most professionally managed funds fail to generate returns exceeding simple index-tracking approaches.

The study reveals that AI systems can predict approximately 71% of active mutual fund trading decisions, yet this predictive capability does not convert into superior market returns. This discrepancy suggests that AI trading algorithms may be replicating human decision patterns without actually improving upon them—a critical distinction for institutions considering substantial capital allocation to algorithmic strategies. The inability to translate predictability into outperformance raises fundamental questions about whether AI trading generates genuine alpha or merely mimics existing market behavior at higher operational cost.

Research examining machine learning applications in equity investment over the past two decades presents a nuanced picture. While some AI-driven fund managers have achieved notable short-term outperformance—including one documented case of an AI analyst outperforming human stock pickers across a 30-year backtest using publicly available information—such results remain exceptional rather than systematic. More critically, academic analysis indicates these profits are "uncertain and fragile," deteriorating rapidly when transaction costs increase, competitive pressures intensify, or market regimes shift. This fragility undermines the case for institutional reliance on AI strategies as a stable return source.

The historical performance record of active management reinforces the case for passive approaches. Data shows that 92-95% of actively managed funds fail to beat their passive index benchmarks over 15-year periods, establishing buy-and-hold strategies as the default superior approach. When AI trading strategies are evaluated against this baseline—and when transaction costs and implementation expenses are fully accounted for—the institutional case for algorithmic trading becomes increasingly difficult to justify. For sophisticated investors evaluating alternative strategies, the evidence suggests that modest, low-cost passive investing continues to outperform both traditional active management and emergent AI-driven trading systems over meaningful investment horizons.