AI Hedge Fund Index Underperformance
The Eurekahedge AI Hedge Fund Index tracks hedge-fund managers who “utilize artificial intelligence and machine learning theory in their trading processes” — making it the closest available aggregate, live-money readout on whether ML-driven trading strategies actually pay off. The readout is negative. From December 2009 to July 2024 the index returned 9.8% annualised versus 13.7% for the S&P 500 (IG Prime / Hulbert Ratings), and Buczynski, Cuzzolin and Sahakian’s review found that from January 2011 to January 2020 it underperformed both the S&P 500 and MSCI World cumulatively (115% vs 210% and 133%). Tellingly, the index’s relative performance was better in the first half of the sample than the second — evidence against the marketing claim that ML strategies “learn and improve” once live.
This note exists because it is one of the few hard data points on Live Trading Evidence for the model family the vault studies. The index has caveats — only ~13 equally weighted constituents, high outlier sensitivity, and a universe definition that differs from Preqin’s broader ML-fund set — and it does not isolate Markov methods specifically. But the direction is unambiguous and consistent with the high-profile liquidations of ML funds (Aidyia, Sentient Technologies) and with active funds generally lagging trackers once fees and costs are paid. Set against a literature of ML/Markov backtests claiming forecasting accuracy “oftentimes exceeding 90%”, the gap between those claims and this live aggregate is exactly the Backtest-to-Live Performance Gap. For the vault it is corroborating negative evidence: the one place where ML trading is measured with real capital at scale, it does not beat a passive benchmark.
Connections
- Backtest-to-Live Performance Gap — supports, the live aggregate readout of the gap, source: https://www.ig.com/za/prime/insights/articles/has-artificial-intelligences-impact-on-hedge-funds-been-overhype-241121
- Live Trading Evidence — reports_underperformance, the rare real-money data point, and it is negative, source: https://pmc.ncbi.nlm.nih.gov/articles/PMC8019690/
- Reinforcement Learning Trading Policy — reports_underperformance, ML-driven funds lag benchmarks live, source: https://pmc.ncbi.nlm.nih.gov/articles/PMC8019690/
- Buy-and-Hold Benchmark — compares_benchmark, underperforms a passive S&P 500 hold, source: https://www.ig.com/za/prime/insights/articles/has-artificial-intelligences-impact-on-hedge-funds-been-overhype-241121
AI Hedge Fund Index Underperformance [supports] Backtest-to-Live Performance Gap AI Hedge Fund Index Underperformance [contradicts] Live Trading Evidence
Sources
- IG Prime (2024). “Has artificial intelligence’s impact on hedge funds been overhyped?” (citing Eurekahedge AI Hedge Fund Index, Hulbert Ratings, MarketWatch). https://www.ig.com/za/prime/insights/articles/has-artificial-intelligences-impact-on-hedge-funds-been-overhype-241121
- Buczynski, W., Cuzzolin, F., & Sahakian, B. (2021). “A review of machine learning experiments in equity investment decision-making.” International Journal of Data Science and Analytics, 11(3), 221–242. https://pmc.ncbi.nlm.nih.gov/articles/PMC8019690/