Stockholm (HedgeNordic) – Artificial intelligence has been a part of the asset management world and the hedge fund industry for some time, with programs increasingly helping fund managers stay ahead of their game by assisting parts of the decision-making process or running portfolios entirely. Artificial intelligence (AI) and machine learning (ML) have been the buzzwords in the hedge fund industry as of late and the adoption of artificial intelligence applications among hedge fund managers has been rising, too. AI and ML applications however may be more suitable for certain trading styles and strategies than for others.
“Eventually every strategy that you can trade can be used in AI or ML concepts,” argues Sebastian Schäfer, managing principal of Leibniz Group. The Swiss investment firm, which Schäfer founded following his tenure as Regional Head at one of the largest alternative investment firms, focuses on developing systematic and machine learning-driven strategies. “In the end, every strategy depends on data,” he explains. “All managers, either systematic or discretionary, use data and systematic tools for data accumulation and aggregation. Every decision-making process is assisted by these tools, it does not work without them. The only question is whether managers are systematizing their entire investment strategy since eventually pretty much every decision-making process can be systematized,” argues Schäfer.
“Eventually every strategy that you can trade can be used in AI or ML concepts. In the end, every strategy depends on data.”
Artificial intelligence, after all, can mimic the decision-making capabilities of the human brain. Martin Källström of Swedish systematic manager Lynx Asset Management corroborates Schäfer’s views. “Virtually all hedge fund strategies would stand to benefit from the use of artificial intelligence in some way,” Källström, Partner and Senior Managing Director at Lynx, tells HedgeNordic.
“Virtually all hedge fund strategies would stand to benefit from the use of artificial intelligence in some way.”
“Although purely systematic strategies and managers – like quant equity and managed futures – are most likely to maximize that value,” emphasizes Källström. Daniel Broby, Director of the Centre for Financial Regulation and Innovation in the United Kingdom, tells HedgeNordic that the applications and benefits of artificial intelligence can vary greatly from one asset manager to another.
“What AI can bring to the table very much depends on the strategy and what you decide you want your AI to actually deliver for you,” explains Broby, who has produced a number of papers on the use of artificial intelligence in banking and fund management. “The specific strategy of a hedge fund manager is very relevant to how you program artificial intelligence and what you want to get out of it. Whoever has the best processing power, the best understanding of the data and the best ability to translate that into a model has the best shot at extracting alpha.”
“What AI can bring to the table very much depends on the strategy and what you decide you want your AI to actually deliver for you.”
Low Signal-to-Noise Ratio
“Managers with ample resources and extensive experience managing quantitative investment programs should have an edge since the biggest challenge in applying AI in finance is how to deal with the low-signal-to-noise ratio and the fact that financial markets are non-stationary,” says Martin Källström of Lynx. “This is something that experienced quant managers have assessed before.” Sebastian Schäfer agrees. “The signal-to-noise ratio is very low in financial data. It does not allow for AI/ML to give you decisions which will be correct 100 percent of the time, contrary to AI applications in image recognition,” says Schäfer. “Letting a machine decide if it is looking at a picture of a cat or a dog is very different to letting a machine decide what the likelihood of an immediate rise in price of a financial asset is, but we believe that a machine is certainly much faster in deciding free-of-emotions what to do next and to accept when it was wrong.”
Per Ivarsson, Head of Investment Management at Swedish CTA specialist RPM Risk & Portfolio Management, shares the same views. “The main problem for systematic investment strategies is that financial data has a very low signal-to-noise data,” says Ivarsson. “Financial markets usually jump between semi-stable regimes, often driven by narratives. The dynamics and feedback mechanisms can be quite different, depending on the current regime,” he continues. However, Ivarsson believes “the AI field includes a large variety of techniques that can help resolve some of those issues, if employed correctly.”
AI Applications for Hedge Funds
Artificial intelligence has grown its presence across the hedge fund space, having the ability to transform many facets of the industry. Different forms of artificial intelligence such as machine learning and natural language processing are being used and can be used across the industry to improve portfolio management, trading, and risk management practices, among many other things. “You can identify how artificial intelligence can benefit things and gain a competitive advantage,” points out Broby, who enjoyed a successful career in the Danish asset management industry prior to joining academia. “Over time, the advantage will disappear because everyone will be employing artificial intelligence, but right now, there is a gap and that gap is there to be exploited.”
“Artificial intelligence can (and is) applied in most areas,” points out Per Ivarsson. “For customer experience, there is already a vast selection of available tools from other industries,” he continues. “Risk management can benefit to some extent. The problems are somewhat different as they often deal with extremes where data is even more scarce. Risk management, therefore, relies more on hard limits and worst-case scenarios,” says Ivarsson. “The most important developments will probably be in the investment area.”
“There are many ways a hedge fund can benefit from AI,” agrees Lynx’s Källström. However, the investment decision-making process is likely to benefit most from the use of artificial intelligence. “Improving technology and increasing computing power has been transforming the hedge fund industry for decades. While AI could accelerate that transformation in various areas, the challenge for all of us will be to increase the forecast accuracy in our investment decisions and to find new and innovative ways to generate uncorrelated returns,” says Källström. “This is where the real transformation can occur, in my view. We have found that these techniques are particularly valuable in modeling complex relationships between markets and the factors that are driving their returns, both known and unknown.”
Skill: The Obstacle and Key to AI Adoption
“The field is developing at a rapid pace and with the democratization of computational- and storage capacity, there is a certain allure in applying techniques that might not be ideally suited for a particular problem,” points out Per Ivarsson from RPM. “When you are holding a hammer, everything looks like a nail. This means that you need quite a deep understanding to identify strengths and weaknesses of the available techniques.”
Sebastian Schäfer of Leibniz Group believes that the biggest obstacle to increasing AI adoption among hedge funds managers is the lack of skill. “Some can produce robust strategies and some cannot,” says Schäfer. “You can systematize a lot of strategies. A new strategy with a strong backtest might work in a given market environment but will only last as long as there are no major market changes or disruptions,” he continues. “You see strategies which work, and then they stop working, either suddenly or slowly, with long track records hiding their decay.”
“Some can produce robust strategies and some cannot. You can systematize a lot of strategies. You see strategies which work, and then they stop working, either suddenly or slowly, with long track records hiding their decay.”
“Our industry is driven by performance. If you generate good performance over a number of years, then building a business around it can eventually lead to success,” says Schäfer, who then also goes on to repeat the often-forgotten investing mantra: “but be aware, past returns are not indicative of future performance.” Especially among systematic managers, “looking at returns and track records of returns going back 10+ years is most often not very useful since so many things were different back then and old times might never come back or the trading strategy has simply been changed too many times,” he argues. “And even shorter track records require proper analysis since you may look at a strategy which improved tremendously since inception. The risk and return statistics might not be representative of what the system can do now, and especially in machine learning strategies, it is possible to never repeat the same mistake twice. Very different to a human, no matter how smart.”
This article featured in HedgeNordic’s “Quant Strategies” publication.