Euclidean Technologies Turns from Machine Learning to Deep Learning to Objectively Analyze Investments


Based in New York and Seattle, Euclidean Technologies uses machine learning to evaluate individual companies as potential long-term investments. The company currently has about $100 million under management.  Before Michael Seckler and John Alberg created Euclidean, they founded one of the first software-as-a-service (SaaS) companies.  In 2006, their company was acquired by ADP for $160 million. They started Euclidean because they faced hard questions about how to manage their money.  They have stated they were not comfortable with the short-term focus, leverage, and reliance on gut instinct that characterized much of the asset management industry.  Instead, they wanted a way to invest that would be process-driven, protected from human behavioral swings, and would enable the evaluation of public companies mathematically as long-term investments. In other words, programmed value investing.

Value investors seek to own stocks they believe the market has undervalued. Investors who use this strategy believe the market overreacts to good and bad news, resulting in stock price movements that do not correspond with a company’s long-term prospects, giving them the opportunity to buy when the price is deflated.

Undervalued stocks come about through investor irrationality. Typical value investors seek to profit from this illogicalness by selecting stocks with common metrics: lower than average price-to-book ratios, lower than average price-to-earnings ratios, and/or high dividend yields. When the market value of the stock is compared with the company’s book value, a value investor invests if the difference is enough to warrant the risk.

Good in theory. However, the practical challenge of value investing lies in the fact that estimating the value of a stock is difficult. Two investors can be given the exact same information and place a different value on a company. What makes Euclidean different is something Seckler and Alberg call systematic value investing, which initially involved machine learning but is now turning to deep learning.

Euclidean’s challenge was to determine which metrics would be most useful in comparing current opportunities with ones from the past. Would earnings yields prove better than price-to-earnings ratios when evaluating whether a company is inexpensive? What about price-to-book or price-to-sales ratios? How has a company’s historical rate of growth tended to relate to its intrinsic value? What would prove to be the best way to consider one-time charges when evaluating earnings? What do measures such as debt-to-equity, return on capital, and gross profitability say about companies’ relative quality? Should one look at just the last 12 months of data on a given company, or should one look instead at how it has evolved over the last several years, or even since the company’s inception? Which of these measures are best, and which are redundant? Investing has been compared with having a lifetime of unfinished algebra homework.

Traditional machine learning techniques require the company put a great deal of effort into trying to answer these types of questions. If the input factors are not right, machine learning doesn’t work.

This ability to work with raw information, instead of information that is extensively pre-processed, is one of the advantages of deep learning. It is part of the reason deep learning-powered image recognition programs can outperform humans by some measures. So deep learning has the potential to find measures that are more meaningful than the input machine learning relies on today, and also to limit potential human biases that go into choosing inputs.

Will this effort be a success? As the founders themselves admit, “We won’t know until we try.”

Innovators like Seckler and Alberg are part of the reason we feel the investment sector is one of the industries best positioned to leverage artificial intelligence. In Tractica’s Artificial Intelligence for Enterprise Applications report, we forecast that spending on AI software in the investment industry will grow from $32.24 million in 2015 to $2.4 billion by 2024.

AIE-15 chart Investment

Comments are closed.