For a period of time, one of the first things I did every morning, was check the price of eggs in the Southeastern part of the United States on the USDA website. A screenshot of the type of report I would seek out is given below.
I did this because I had a position in Cal-Maine Foods, the largest producer and distributor of fresh eggs in the United States. The company has a totally zen stock symbol of CALM. I closed the position several years ago and occasionally think about the company every time I come across news about how the price of eggs have suddenly spiked or that there is a shortage of eggs at my local Costco.
When reviewing the 13-F filing for Seattle-based quantitative investment firm Euclidean Technologies this week, I was surprised to see that they had started a new position in Cal-Maine Foods in Q1 2023 by allocating 2% of their portfolio to CALM. Euclidean uses AI and specifically deep learning to inform its investment decisions. The firm, which has been a pioneer in using machine learning for investing is generous about sharing its approach and insights with the investment community. They describe part of their approach in the Q4 2022 investor letter:
“By using deep learning techniques to train a model to generate the output sequence given the input sequence, we were, in effect, teaching the model to forecast the next year’s financial information for a company from its own historical financial information.”
More than a decade ago, I became fascinated by systematic or quantitative investing and decided to build a model that quantified my investment experience. The model used 20 factors like growth, valuation metrics, strength of balance sheet, management effectiveness, etc. and assigned various weights to these factors to come up with a numerical score for a company.
For example, if I was a value-oriented investor that preferred a company trading at a low P/E multiple and had adequate liquidity as defined by a high quick ratio, those two factors would have a higher weight assigned to them than something like year-over-year revenue growth.
Experienced investors would recognize both the advantages and limitations of such a model. It could work as an excellent initial filter but there would be additional work needed to weed out the value traps and cyclical stocks at the wrong end of their cycle. I strongly believe in a hybrid approach of using machines for pattern matching and then human insight or ingenuity for decision making as discussed in the book The Second Machine Age.
I met one of the co-founder’s of Euclidean Technologies, John Alberg, in Seattle several years ago. As an experiment for several weeks, we ran both my model and Euclidean’s model through a list of companies to see how each of our models would score the potential investment. It was a fun little exercise and I started to appreciate the work Euclidean Technologies was doing. Their systems and approach to quantitative investing has evolved significantly over the years.
Since the firm adopted a machine learning approach to investing in March 31 2020, its Euclidean Fund I has returned 22.26% compared to 19.31% for the S&P 500 through the end of 2022. For the full year 2022, losses were well below the losses experienced by the S&P 500, the Nasdaq and the Russell 2000 indexes.
From my conversations with John over the years and after reading his letters, it is clear that the firm has a long holding period and some stocks have been in the portfolio for several years. The value of the firm’s 13-F portfolio was $109 million with 75 positions. Concentration in the top 10 was only 25%, implying a fairly diversified portfolio.
Some of us start our investment process by running screens, some by reading our favorite authors/analysts/magazines and yet others like me use event-driven situations. Sourcing ideas from the 13-F portfolio of a fund that holds its positions for long periods of time and that has a investment philosophy that matches yours can be incredibly rewarding. It is not a substitute for doing your own work but provides an excellent starting point.
Reviewing the latest additions to Euclidean’s portfolio, I was surprised to not just find Cal-Maine Foods (CALM) but also home builders like Meritage Homes (MTH) and Lennar (LEN). The algorithm also appeared partial to coal companies like Warrior Met Coal (HCC), Alpha Metallurgical Resources (AMR) and Arch Resources (ARCH) as well as oil & gas companies like Valero Energy (VLO), Civitas Resources (CIVI) and Southwestern Energy (SWN).
It was also good to see that companies that are considered uber cannibals like Autonation (AN) and NVR (NVR) were new additions to the portfolio. There was even a company in the portfolio that last week announced a large buyback representing more than 10% of the market cap of the company. Instead of revealing the name, I will leave you with the hint that the algorithm increased the position by 350% in Q1 2023. Happy hunting.