The human input

Current news stories about artificial intelligence (AI) seem to focus either on its ability to take over people's jobs or "machines running amok." As long-standing adopters of AI in quantitative investing, we are aware of AI's power and limitations. What we have seen is that computers cannot yet match human insight, and humans can't process data as quickly and comprehensively as computers. Our investment process, which relies on machine learning, gives us a great deal of control over the quality of the input and output to our portfolio construction process.

Quant goes mainstream

Quantitative, systematic or other investment approaches that depend heavily on computing power for research and portfolio construction were once out of reach or not well-known by smaller investors, being chiefly associated with larger institutional investors. However, they have become increasingly accessible to smaller investors in the last 20-plus years through ETFs and mutual funds. These strategies typically rely on mathematical models, statistical techniques and computer algorithms to digest large datasets and identify patterns that help managers make investment decisions.

What often gets lost as these strategies become more common is that human input is, or should be, critical to quant. This is certainly the case at MDT, where the expertise that each team member brings to the process is consequential.

MDT has implemented machine learning for several decades to analyze the vast dataset that comprises our stock selection universe, as we seek equity opportunities through an unbiased lens. On the surface, this may appear "cold," but the team designs, oversees and continuously reviews ways to improve the tools that drive this process.

Our portfolio construction is not a black box. We review trades before implementation to clarify the reasoning behind investment decisions. Our process is clearly defined, testable and in many ways more transparent than those of traditional investors.

Assessing active investment strategies involves many performance variables that shift over time. Comparing traditional bottom-up portfolio managers — discerning skill from luck — is difficult.

Portfolios built to weather the storm

We believe a key advantage of quantitative investing is testable models, using historical market data. This allows us and our investors to understand how a particular strategy may have worked over a specific period and what the return and risk outcomes might have been.

Another potential advantage of quant investing is greater diversification across company types, even within a specific index. Managers of more traditional fundamental strategies often incorporate a particular stock selection focus that aligns with their investment philosophy: quality, value, earnings growth, return on equity, capitalization and yield, to name a few. These characteristics, or factors, can be used to narrow the field of potential investments. However, returns may suffer if the investing environment for the chosen "factor tilt" is out of favor.

By contrast, we believe a flexible quant strategy can adapt to market environments. MDT aims to build portfolios that invest in many different types of companies, with varying return drivers, so that the portfolio has the potential not just to outperform over time but also deliver more consistent performance in many market environments.

Avenues to risk management

MDT does not have to have a view on which sectors or factors will outperform at any given time; thus, we set sector limits and use risk models to avoid unintended risk exposures, focusing instead on stock selection to generate excess return potential.

Leveraging our computer capabilities, we update our stock forecasts — predicted alpha — daily for the entire universe of stocks. However, before we place trades, our team will review these to ensure we incorporate the most recent information. In addition, we review portfolio positioning daily, enabling us to adapt our strategies to take advantage of timely market opportunities. This active approach is designed to help ensure our clients' portfolios always reflect our best, most current ideas.

MDT has long integrated machine learning into its investment modeling, prioritizing transparency and accountability. Using regression trees — supported by proprietary technology — enables us to understand and explain every trade clearly. By maintaining high-quality historical data, overseen by our team of investment professionals, we can build portfolios designed to demonstrate the complementary strengths of computer technology and human insight.

For more information on MDT's development of factors please read Bringing economic moats to quantitative investing.