By relying on artificial intelligence, hedge funds are showing impressive returns on investments. However, only a small number of them entrust a significant part of their portfolio management to AI and machine learning algorithms.
Volt Capital Management AB, a Swedish hedge fund that manages about $30 million, has just announced a 24% return on investment this year. His secret? Artificial intelligence..
Thanks to the combination of about 200 traditional business models and AI algorithms, this fund has been able to accurately predict trends such as the oil crash and currency appreciation in various countries. Its AI model automatically applied the correct weighting to get accurate results.
Volt Capital Management AB is not alone in building on AI’s strengths. The EY’s 2018 Global Alternative Fund Survey shows (see chart below) that more and more hedge funds (including Bridgewater, Citadel, Cerebellum Capital, Taaffeite Capital Management…) are using AI to optimize their investments and improve their performance.
Although AI-focused hedge funds have lagged behind their “classic” competitors (i.e. with “real” analysts) over the past two years, they have performed more well over the past five years.
The human “flair”
All these edge funds have surpassed the Preqin Pro over the past year. This benchmark tracks the performance of 152 hedge funds that exploit artificial intelligence. Based on cumulative returns over three years, these funds achieved a return of 26.96%.
In general, one of the main uses of AI is to transform big data,including unstructured data (satellite images, news, social media messages), into structured data that can be more easily used to generate trends or sharp analytics.
But human experience is still essential, even for these hedge funds. They use quantitative models to develop new buying and selling strategies and to identify new trends. All of this data is then integrated into more “classic” software. But because these often require reprogramming by analysts, they are considered “pre-IA” models.
Leverage and over-adjustment
Second, these funds use “pure” AI models that can adapt to changing markets. As a result, these funds have a decisive advantage over their analyst-only competitors: time. Because they can process large volumes of data quickly, they can instantly adapt to market turmoum.
Machine learning (ML) machine learningalso means that the AI model can be updated automatically as new data is collected, without any human intervention.
The update takes into account, among other things, social media analysis to assess consumer, market and investor sentiment on a particular asset or security. But as we explained in a previous article,predictive models for human behavior are far from perfect…
“All these models make their business profitable, but for me, it’s not very moral, because it’s purespeculation,” says Jean-Gabriel Ganascia, professor at Pierre and Marie Curie University and head of the ACASA (Cognitive Agents and Automatic Symbolic Learning) team at LIP6.
There are also risks inherent in hedge fund strategies that rely on machine learning, warns JP Morgan. In particular, there is “leverage” that can be amplified. This technique is used to invest more money in financial markets than the fund has. It allows you to multiply the winnings with a small starting bet, but it also greatly increases the risks.
Over-adjustment can occur in ML processes when models are so finely adjusted to identify past models that they fail to accurately predict future stock market movements.
Finally, algorithms cannot predict crises that are by definition impossible to anticipate. This is the case with the Covid-19 pandemic…