The Big Problem With Machine Learning Algorithms
The potential for tapping new data sets is enormous, but the track record is mixed.
Machine learning is enabling investors to tap huge data sets such as social media postings in ways that no mere human could. Yet, despite the enormous potential, its record remains mixed. The Eurekahedge AI Hedge Fund Index, which tracks the returns of 13 hedge funds that use machine learning, has gained only 7 percent a year for the past five years, while the S&P 500 returned 13 percent annually. This year the Eurekahedge benchmark dropped 5 percent through September.
One of the potential pitfalls for machine learning strategies is the extremely low signal-to-noise ratio in financial markets, says Marcos López de Prado, who joined AQR Capital Management as head of machine learning in September and is the author of the 2018 book Advances in Financial Machine Learning. “Machine learning algorithms will always identify a pattern, even if there is none,” he says. In other words, the algorithms can view flukes as patterns and hence are likely to identify false strategies. “It takes a deep knowledge of the markets to apply machine learning successfully to financial series,” López de Prado says.