AI-based hedge funds: a key to success?

© Adobe stock

An impressing field for the use of AI is the capital market, especially the use of AI in hedge funds. As the environment seems to be made for it: high volumes, quantitative aspects and the need for speed and accuracy are promising. But what does the reality look like? Are AI-based hedge funds really more successful? A new study now provides evidence and shows that AI-based hedge funds outperform those that are managed and controlled by human investment managers.


A recent study has examined the relationship between the use of AI and hedge fund performance. It concludes that hedge funds with the highest degree of automation achieve the highest performance compared to hedge funds based on human behavior. According to the study, AI-based hedge funds achieved average returns of approximately 0.75% per month. These results were compared with hedge funds based purely on human investment management. Those achieved average returns of about 0.25% per month. The study examined the period from 2006 to 2021 and is strikingly titled, "Man Versus Machine: On Artificial Intelligence and Hedge Funds Performance" and appeared in the April issue of the journal Applied Economics.

The importance of artificial intelligence in the field of hedge funds

According to a report, total assets managed by hedge funds worldwide reached $3.87 trillion in 2020 and $4.53 trillion in 2021, according to statista. This shows that the importance of hedge funds is high and the capital employed could rather increase due to the still relatively low interest rates. The hedge fund industry is a highly competitive environment. The high competitive pressure forces hedge funds to adopt new technologies and constantly improve them. Computerized and automated trading systems have been around for some time. They also change with time and are constantly improving. The increasing technical progress is especially noticeable in accuracy and speed. This in turn is reflected in the performance.

Increasing automation in the hedge fund industry

The objective of the study was to find out whether transferring actions and control to advanced trading algorithms can lead to better returns. Related to this is the question of whether decisions can be better made by an algorithm. The study considers a sample of 826 hedge funds, 36 of which are AIML funds. The study is based on fund performance for 173 consecutive months from September 2006 to January 2021, and only considers data from hedge funds operating in North American markets. Also included are hedge funds that report their returns in U.S. dollars and are primarily equity-focused.

What is characteristic of "AI-based" hedge funds?

For a better assessment, the hedge funds were divided or categorized into different styles of action. These are subdivided from the least automated to the most automated: discretionary, systematic, combined, or AIML. Discretionary refers to all funds that are mainly planned and executed by human involvement. Here, the emphasis is primarily on human knowledge. These are followed by systematic hedge funds, which are based on statistical methods and operate within the framework of quantitative rules. Combined funds are based on a systematic style of action. However, actions are decided manually by human judgement. The most automated hedge funds are called "AIML funds". This includes artificial intelligence and machine learning funds. They are characterized by specific inputs and a targeted output. The algorithm itself determines the course of action.

Findings of the study: AI-based hedge funds performed best

As a result, the study was able to show that the hedge funds with the highest level of automation were also able to generate the highest returns. This corresponds to the category of hedge funds that use and employ AI and machine learning in the investment process. AIML funds averaged returns of 74 to 79 basis points per month. This represents an average monthly return of 0.74% to 0.79%. In comparison, hedge funds that followed a discretionary style of action, i.e. a mainly manual approach, achieved the lowest average return. This ranged from 0.23 to 0.28 basis points. This corresponds to a difference of about 0.5% per month compared to the AIML funds. However, the authors also conclude that it may be that performance can change over time. As a result, however, AIML funds were shown to have better average returns than hedge funds with a high degree of human involvement.

What are advantages of AIML funds compared to discretionary funds?

Advantages of AIML funds over discretionary ones are mainly the speed of assessing and evaluating the available data. As a result, the speed of action and implementation of strategies is also crucial. They are also able to identify short-term and even relatively small market inefficiencies and exploit them to their advantage. AIML funds are also able to constantly identify new forecasts and advantageous strategies. This means that new trading inefficiencies can again and again be identified and exploited.

What can be deduced from this study?

The benefits of AI-based decision-making are becoming more and more evident in many studies. So the question is rather not whether AI-based decision making will become important or not. AI-based decision-making methods have the potential to become mainstream. Therefore, the study suggests much more that addressing AI-based decision making is becoming more important than ever. This is because organizations that are aware of the advantages and disadvantages can also adapt and change processes in a beneficial way. As a result, organizations can introduce new technologies that are necessary to stay ahead of the curve and respond in an agile manner in the medium and long term. Thus, it is important to deal with the technologies today that could be more or less commonplace in the near future.

21strategies: AI for decision optimization under uncertainty

The core competence of 21strategies lies exactly here. 21strategies uses third-wave AI to optimize decisions under uncertainty. Algorithmic planning and execution strategies are trained for this purpose. Find out more and have a look at our website.

Author: Tanja Zimmermann