Practical analysis for investment professionals
17 May 2018

Artificial Intelligence-Driven Investing: High Alpha behind the Buzz

Artificial intelligence (AI) may be among the latest buzzwords in finance, but applying it to investment decision making will disrupt the industry and benefit those investors who harness its power.

If used correctly, AI can add high alpha potential within a more stable modeling framework.

Behind the Buzz: What Is AI?

AI is the basis for a different quantitative investment paradigm. It is a nonlinear, high-dimensional learning approach that typically seeks to replicate human reasoning. One interpretation of this new paradigm can be thought of as learning (and learning to apply) “Graham and Dodd”–style systematic rules.

More generally, AI endeavors to make machines that are capable of intelligent behavior. This distinguishes AI from the less ambitious machine learning, which seeks to build machines that can act without explicit programming.

Whatever the interpretation, AI investing is very different from traditional quant investing, and in a world of crowded factor trades, differentiation is likely a good thing.

At the 2018 Cognitive Computation Symposium in February, leading researchers made two key points about AI.

  1. DeepMind and Imperial College researchers indicated that AI’s results should ideally be readable to humans, rather than confined to a black box.
  2. IBM AI researchers stated that “narrow AI,” or AI applied to specific tasks, is now very much a reality.

Here we offer a specific interpretation of a narrow AI applied to investments in emerging markets equities that makes decisions that are communicated in human terms.

AI Investing: Conceptually Distinct

The chart below highlights some central AI concepts. The pink shaded items are relevant to investments.

A central aspect of AI is Integrated Reasoning, with the scope for Adaptation and Evolution. This means an AI can learn expert investment rationales and adapt or generalize them to future investment environments. This is analogous to generalized Knowledge Discovery of systematic rules, where an AI can scour very large databases, say Bloomberg or Factset, for investments that can lead to objective outcomes consistent with guiding rationales.

For example, targeting defensive, value return characteristics in emerging markets equities can be achieved with “income orientation” as the guiding rationale.

Artificial Intelligence (AI) Concepts

Artificial Intelligence (AI) Concepts

Source: Rothko Investment Strategies, Hiroki Sayama, State University of New York

To be clear, big data, unstructured data, and sentiment scoring are derived from specialized data-mining and machine-learning algorithms. They are not AI.

Learning to Be Different

AI differs in almost every respect from the ubiquitous factor-driven investing of traditional quants that now accounts for around $1.5 trillion in total AUM.

While factor-driven investing views the world through a simplified, linear-constrained lens, AI can retain more information about the world to inform decisions. AI can integrate myriad perspectives into each investment decision through the Collective Behavior of different rules or models, synthesizing the most pertinent information to guide decision making.

This is very different from the vast majority of traditional quants who generally seek to forecast expected returns or tilt exposures towards higher momentum factors (for example, an EDHEC survey  indicates that 74% of respondents forecast expected returns and 65% tilt exposures to high momentum factors or factor reversals).

Traditional Quant vs. Artificial Intelligence (AI)

Traditional Quant vs. Artificial Intelligence (AI)

Source: Rothko Investment Strategies

Compared to a stylized and simple traditional quant strategy, AI has more in common with a human-driven, fundamental approach, as the graphic below demonstrates. An AI successfully applied to investing combines the most appropriate aspects of data-driven modeling techniques with guiding human-like rationales.

There is also a distinction between AI and such purely data-driven methods as raw machine learning. We have found that applying complex data-driven technologies like machine learning to long-term investment models can be dangerous.

Financial data is extremely noisy, with the faintest signals concealed in that noise. Bringing powerful data-mining to bear on noisy data tends to result in overfit. That is, the models fit more to the noise than to the signal. This results in inaccurate trading signals.

A far more pragmatic strategy is to constrain a powerful data-driven methodology using fundamental rationales as a guiding principle.

Differentiated in Principle: AI vs. Factor-Driven Investing

Differentiated in Principle. AI vs. Factor-Driven Investing

Source: Rothko Investment Strategies

Under the Hood: Stock Selection and the AI Decision Boundary

The chart below visualizes how an AI strategy can select stocks based on a higher dimensional view of the world.

In this case, the target is a portfolio of emerging markets equities that exhibits defensive, value characteristics. The chart maps all listed equities in the emerging market equities (MSCI EM IMI) universe. Each security is plotted on the surface based on its coordinates to high dimensional value and earnings features, which are each defined as part of the AI approach.

The vertical axis shows how defensive these feature combinations have been on average between 2013 and 2017. The higher the peak, the more defensive in down markets. The deeper the trough, by contrast, the less defensive. On the right side of the visualization, a two-dimensional slice shows where well-known emerging market names — described by their value and earnings characteristics — ended this period.

This approach defines a complex and evolving decision boundary, illustrated by the maroon contour, within which stocks are selected to construct a portfolio. This decision boundary represents a stable, defensive/value region of the map. It is a nonlinear region that is high dimensional and evolves as market conditions change. Stocks within the decision boundary tend to exhibit defensive characteristics per se, but the method also identifies stocks that tend to have future dividend income growth potential as well.

By fusing a powerful learning strategy with guiding rationales of this sort, return characteristics tend to be more accurate and more stable.

Through AI’s High-Dimensional Lens

Through AI’s High-Dimensional Lens

Source: Rothko Investment Strategies

Because of the nonlinear modeling framework, we would expect a higher proportion of portfolio returns not to correlate well with traditional factors. This would most probably result in a higher (multifactor) alpha. In our experience, turnover tends to be lower relative to traditional quant approaches.

This indicates that all else being equal, an AI modeling framework can produce a more stable outcome than traditional equivalents.

A Future for “Human Intelligence” in Investing?

While a well-conceived and well-implemented “narrow AI” can generate a stable alpha source at lower turnovers, human portfolio managers should still be part of its workflow. Why? Two reasons: exogenous risk and sense checking. For this to be possible, a precondition must be met: Modeling outcomes need to be human readable concepts.

To enable human-machine interaction, AI models must produce simple fundamental concepts, not, for example, vectors of coefficients. Purely data-driven or black box decision making — and that includes factor investing — is not acceptable for the high magnitude decisions of long-term investing.

This idea reads across to a cutting edge area of AI research known as Neural-Symbolic Reasoning which is being led by research teams at the University of London, DeepMind, and Imperial College, among others.

While having an experienced team of investment professionals “sanity checking” the modeling outcomes is critical, the amount of human portfolio manager intervention in a long-term AI-driven investment strategy is, from our experience, likely to be relatively low. As the AI learns and evolves (or is adapted), intervention decreases.

AI-Driven Investing: A New and Powerful Paradigm

AI is far more than a buzzword. It is a different quantitative investment paradigm that seeks to create a repeatable process that is capable of quasi-intelligent behavior. This is achieved by consistently applying human-like investment rationales while avoiding the behavioral pitfalls of true human stock selectors.

Elon Musk and others may fear that an AI singularity will cause a mass extinction event. We believe that scenario is farfetched. But a singularity in narrow investment AI could well presage a mass extinction of low alpha investment managers.

If you liked this post, don’t forget to subscribe to the Enterprising Investor.

Views expressed were current as of the date indicated, are subject to change, and may not reflect current views.

Views should not be considered a recommendation to buy, hold or sell any security and should not be relied on as research or investment advice.

The information was obtained from sources we believe to be reliable, but its accuracy is not guaranteed and it may be incomplete or condensed. All information is subject to change without notice. No representation is being made that any account will or is likely to achieve profits or losses similar to those shown.

This document may include forward-looking statements. All statements other than statements of historical facts are forward- looking statements (including words such as “believe,” “estimate,” “anticipate,” “may,” “will,” “should,” “expect”). Although we believe that the expectations reflected in such forward-looking statements are reasonable, we can give no assurance that such expectations will prove to be correct. Various factors could cause actual results or performance to differ materially from those reflected in such forward-looking statements.

This introductory material is for informational purposes only and is not an offer or solicitation with respect to any securities. Any offer of securities can only be made by written offering materials, which are available solely upon request, on an exclusively private basis and only to qualified financially sophisticated investors.

Neither MSCI nor any other party involved in or related to compiling, computing or creating the MSCI data makes any express or implied warranties or representations with respect to such data (or the results to be obtained by the use thereof), and all such parties hereby expressly disclaim all warranties of originality, accuracy, completeness, merchantability or fitness for a particular purpose with respect to any of such data. Without limiting any of the foregoing, in no event shall MSCI, any of its affiliates or any third party involved in or related to compiling, computing or creating the data have any liability for any direct, indirect, special, punitive, consequential or any other damages (including lost profits) even if notified of the possibility of such damages. No further distribution or dissemination of the MSCI data is permitted without MSCI’s express written consent.

Graphical exhibits in this document are solely owned by and the intellectual property of Rothko Investment Strategies.

Rothko Investment Strategies is a trading name of Mondrian Investment Partners Limited. Mondrian Investment Partners Limited is authorised and regulated by the Financial Conduct Authority.

All posts are the opinion of the author. As such, they should not be construed as investment advice, nor do the opinions expressed necessarily reflect the views of CFA Institute or the author’s employer.

Image credit: ©Getty Images/Photographer is my life.

About the Author(s)
Dan Philps, CFA

Dan Philps is head of Rothko Investment Strategies and is an artificial intelligence (AI) researcher. He has 20 years of quantitative investment experience. Prior to Rothko, he was a senior portfolio manager at Mondrian Investment Partners. Before 1998, Philps worked as an analyst/programmer at a number of investment banks, specializing in trading and risk models. He has a BSc (Hons) from King’s College London, is a CFA charterholder, a member of CFA Society of the UK, holds a post graduate research role at London University, and is a member of the AAAI.

Leave a Reply

Your email address will not be published. Required fields are marked *

By continuing to use the site, you agree to the use of cookies. more information

The cookie settings on this website are set to "allow cookies" to give you the best browsing experience possible. If you continue to use this website without changing your cookie settings or you click "Accept" below then you are consenting to this.