EFG Private Bank’s head of fund research, Jerome Berset, and senior portfolio manager Alex Allen unearth the science behind systematic strategies and what the future holds for funds focusing on this function.
Hedge fund strategies can be discretionary or systematic. Systematic managers employ two distinct approaches. The first group of systematic managers uses fundamental data to evaluate the current economic situation and trigger buy or sell signals according to the identified opportunities, which is macro systematic.
The second group typically uses only technical, price-based data in their decision making process. Within this second cluster, we can split the strategies between ones applying their models to futures instruments (i.e. CTAs) and a second, which invest only in equities, which is statistical arbitrage and quantitative market neutral.
With the technical (r)evolution of the late 1990s and the changes in the regulatory environment in the US, a new avenue for statistical arbitrageurs opened up. Some systematic managers quickly took advantage of this evolution in order to trade more efficiently and more rapidly.
It enabled them to shorten the trading time frame of their strategy. From this point onwards, we observed the emergence of new strategies taking advantage of very short term price movements and High Frequency Trading became a reality. These managers use algorithm-based processes to generate their edge.
Diving into data
More recently, other quantitative strategies have emerged that incorporate Big Data and other novel data sources - such as satellite, credit card and social media - as the key inputs into process and decision-making using machine learning techniques.
Evolved from the study of pattern recognition and computational learning theory in artificial intelligence (AI), machine learning (ML) explores the study and construction of algorithms that can learn from and make predictions on these types of data.
Some investing titans have begun to dedicate significant resources on the practical appliance of machine learning. Several entrepreneurial thought leaders are also in the process of launching or backing pure AI hedge funds.
All this has been made possible by three trends: an explosion in the amount and range of data available; an increase in computational power and storage capacity along with a drop in cost (helped by cloud-computing); the emergence of new machine learning techniques that are able to sort and make sense of complex raw data.
Therefore, the investment universe in the quantitative space is rapidly increasing and gives allocators more opportunity to create portfolios with interesting performance characteristics.
Indeed, all the above mentioned strategies (CTAs, Macro Systematic, Stat Arb, HFT and AI/ML) will offer up different streams of returns from their orthogonal performance drivers and potentially provide significant diversification benefits within both conventional and alternative portfolios.
In the mix
We have always viewed these quantitative types of approach as being an integral part of the exposure to alternatives for investors and we will continue to use them alongside other more traditional/discretionary managers.
In recent times, our research focus has tilted to the very new branch of machine learning funds that use techniques such as supervised learning, unsupervised learning, deep learning and reinforcement learning algorithms.
While we have only identified a handful of funds that predominantly use these techniques, performance results have been, for the most part, very encouraging and we are currently evaluating our next move.
While some have called AI/ML the Third Wave of Investing, it is our belief that certain quarters of the more traditional quantitative investing have become commoditised, have reached the limits of what they can achieve and have seen their original edges arbitraged away.
These comments originally appeared in a shorter format in the September edition of Citywire Selector magazine.