Mathematic & computer quant models don’t represent a novelty. For years, quantitative hedge funds have been using computer algorithms to make trading decisions. In recent years, these funds moved to AI-driven algorithms, which became mainstream. Thus, 70 % of new hedge fund launched globally in 2018 will include investment processes supported by AI and machine learning algorithms, as compared with 47% in 2015 [1]. And the amount of money managed by quant hedge funds is not negligible: it rose to more than $940bn by the end of October 2017 [2] — nearly double the level of 2010 — and flows have continued to be strong in the fourth quarter.
Why a move towards AI?
First, the volume data available have increased exponentially. The amount of data stored on IT systems around the world has doubled every two years and it would mushroom from 4.4 zettabytes (a zettabyte equals 10^21 bytes) in 2013 to 44 zettabytes by 2020. It has already reached 25 zettabytes in 2017 [3]. In the same time, the nature of data has changed: it is estimated that 80% of the world’s data are unstructured (video, audio, images, social networks, etc.), so non directly exploitable.
Second, the methods to analyze complex data-set have been tremendously enhanced with the increased computing power and data storage capacity, at reduced cost (especially by cloud computing). Traditional algorithms used by quant funds and hedge funds were driven by static models, developed and managed by data scientists/traders, which weren’t able to deal with the volatility of financial markets. In contrast to these traditional models, artificial intelligence (AI) and machine learning (ML) algorithms work by analyzing huge amounts of data and defining their own rules based on the patterns and connections they find between different data points. They are able to autonomously update themselves as they ingest new data.
The applications in hedge fund’s investment process
In portfolio management, AI and ML tools are being used to identify new signals on price movements and to make more effective use of the vast amount of available data and market research than with current models (see article How AI disrupt research in AM ? in DigiBook #9).
As existing analytical techniques used in systematic investing, the aim is to generate higher and uncorrelated returns. The difference is that ML algorithms are able to achieve flexibility with regards to the model and evolve with the ever-changing markets. They continually learn from its previous forecasts and adapt to new conditions and features quickly. So that, they further enhance their capabilities to be able to make predictions in circumstances not observed before, as a result of its learning experience.
Main hedge funds – Bridgewater, Two Sigma, Point72, Aydia... – are using ML & AI to find investment opportunities, predict price movements in financial markets or spot patterns that offer an edge in investing.
Sentient Technologies, an AI company that also runs a hedge fund, has developed an algorithm to find trading patterns and forecast trends, which enables it to make successful stock trading decisions. Sentient runs trillions of simulated trading scenarios created from the vast amounts of public data available online to identify and blend successful trading patterns and define new strategies. These techniques enable the startup to squeeze 1,800 days of trading into a few minutes. Successful trading strategies are then tested in live trading, where they evolve autonomously as they gain experience.
AI-driven investment process is extending to traditional funds
This technology is not limited to hedge funds ; it’s now on the rise in quantitative mutual funds, and even in fundamental management. AI makes it possible to develop new investment strategies and portfolio construction that differ from other human and machine-guided strategies.
It is estimated that ‘pure’ AI & ML players have about $10 billion in assets under management, and that this figure is growing rapidly [4].
In June 2017, Sanlam Global Investment Solutions (SGIS) has launched the Sanlam Managed Risk UCITS Fund, which is solely managed through AI, making it the first Balanced Fund in the world fully driven by AI & ML. It has currently £57 million AuM.
The AHL Dimension, managed by Man Group (UK), is a multi-strategy quantitative fund which invest in a range of diversified systematic strategies trading liquid instruments. Its AuM has quintupled from 2014 to 2017, thanks – for a part – to its strong performance driven by IA and alternative data. This has led Man Group to deploy AI techniques for an additional four funds managing $12.3 billion [5].
Mirae Asset AI Smart Beta Fund which uses deep learning technology to come up with investment portfolios. The AI learns various indices and other information which affect the investment market. It then organizes a portfolio based on this knowledge and compares it with an optimal portfolio, which reflects the result of investment. By repeating the process, the AI narrows the gap between two portfolios.
AI could be also used by asset management firms to cope with new regulations (MiFID II, UCITS Directive, AIFMD). Firms could potentially leverage machine learning tools to interpret these regulations into a common language. They could then analyze and codify the rules for automation into the integrated risk and reporting systems to help firms comply with the regulations. This could bring down the cost, effort and time needed to interpret and implement new and updated investment constraints for fund managers.
In this regard, BNP Paribas Securities Services is working with around 15 AMs to test the application of ML for compliance in portfolio management to ensure that portfolios are meeting local and international requirements around issues like liquidity and equity holdings.
IBM has launched a Watson product for financial regulation, rolling out an AI tool to help financial institutions, such as JPMorgan or Credit Suisse, to identify requirements that companies might face and to assess whether the company's compliance programs are sufficient to comply with the rules
Are AI better than human?
Some argue that the talk around the potential of AI and machine learning to unlock profitable patterns in large data sets is nothing more than marketing buzz. It is true that track records are already too short to be affirmative and machines are still in learning process. But if AI-driven hedge funds failed to beat the S&P 500 Index since 2013, they outperformed both traditional quant and more generalized hedge funds since 2010.
Thus, Eurekahedge’s AI & ML Hedge Fund Index, which tracks the historic performance of 23 hedge funds, outperformed its peer indexes during this period. The AI-driven hedge funds provided investors with an annual return of 8.44% over this period. This compares favorably with the Eurekahedge CTA / Managed Futures index, which delivered 2.62% over a similar period, the Eurekahedge trend following index, up just 1.62%, and the Eurekahedge hedge fund index, up 4.27% since 2010 [6].
The Fintech I Know First develops, back-tests and offers systematic trading strategies powered by AI. The final product consists in giving trading recommendations for execution, depending on the investment strategy profile chosen. The strategies can be used for any investment vehicle. They affirm that they all generate high positive alpha while keeping beta in the 0.3-0.8 range, yielding overall high risk-adjusted returns. To be followed...
An organizational mutation is required
One view in the industry is that for AI & ML to be effective, both traders and quants need to have good oversight and understanding of the tools used. Many quant funds state that they are not comfortable with fully automating and implementing a model if they cannot understand and explain how a particular prediction is made.
This mutation therefore requires new profiles in the asset management industry, such as data scientists and AI & ML specialists. New positions are already in creation such as Man GLG’s “Head of machine learning” or as experts located in the Innovation / Digital Labs established by several asset managers (RBC Global AM, Nomura, BlackRock...).
For the moment, a small number of managers are making a serious commitment, investing significant resources and capital in the hope of becoming leaders in the AI and machine learning space. But, it could be rapidly mainstream because, as Cerulli states: “Leading asset managers recognize that machine learning and AI represent a solid opportunity. The ability to extract value from large data sets will become a key differentiator in the future.” [7]
Author: Pascal Buisson - March 2018
[1] Deloitte, 2018 Investment Management Outlook - [2] Hedge Fund Research, 2017 - [3] International Data Corporation (IDC), Data Age 2025, April 2017 - [4] Financial Stability Board, Nov. 2017 - [5] Bloomberg, 2017 - [6] Eurekahedge, 2017 - [7] Cerulli, European Alternative Investments 2018
