Lately I have been receiving emails from unknown sources that purport to have the power (using AI) to spot a stock’s future price movements before the fact that human eyes and minds cannot see. As proof of their claim, the sender shows a graph of a stock’s actual price movements, including a predicted path  presumbly predicted by an AI algorithm. The second graph shows the old path and the predicted path, overlaid by the actual data to the end of the prediction.  The new path and the predicted parh appear to the  very close to the same. Some gains are reported to be huge.

Did he get lucky, or was there some shady operation at work to draw his mail recipients in for the kill?

Background. I published a paper back in 1995 that showed stock price movements were random. That wasn’t news. Nearly all stock price movements were and still are random. Over the years thousands of students of markets have confirmed that bit of activity. What was new was this: When many stock prices were combined (as in an asset class of many real estate stocks or in a market index), the aggregate was not random. The asset class generally consisted of many stock prices moving somewhat in unison, but not entirely. I recall one University of Chicago professor and two others who called me seeking more information about the nonrandom assets classes and how to detect them. I explained how to detect a non-random time-series that all their students probably knew how to do.

How to use that knowledge profitably was a different matter. In practice I used a formula that captured the excess profits between the asset class and the SP500, plus factors outside the market (alpha), and the correlation between the asset class and the SP500 (R^2), with other cost savings coming from eliminating unnecessary fees. This algorithm produced annual returns of 15 percent, or 5 percent above the markets. Our fee was 1 percent for the company, and an annual average of 4 percent bonus for our clients over a period of 11 years. The larger algorithm also considered costs, alpha, beta, the Capital Asset Pricing Model, and optimization in arriving at the predicted path for guidance throughout the entire year. That algorithm is still valid, but to my knowledge it is still not in use anywhere to this day. It took me 20 years to test and perfect it. When an honest philanthropist comes along, he or she will get my algorithm for free. I still smile when I hear about the University of Chicago School of Finance as the world’s leading authority on market efficiency, random walk, and why nobody can beat the markets.

But 4 or 5 percent per year in excess returns is a long way from the claims of AI practitioners today. A few are claiming annual returns of 4,000 percent per year. Here is what I suspect is going on behind the scene:

They find a stock that is undervalued by conventional methods (earnings, profits, price-to-earnings ratios, etc.). That knowledge, alone, does not give them an edge. It is what 95 percent of all market analysts do. Next they buy enough stock of that company to get positioned for gains. Then they alert their client base that company XYZ is positioned to move up, and they give extra credence to their discovery by crediting their AI model. Their clients rush out to buy up shares. And the money inflows drive up the price of that stock. When they have profited, they sell and take their profits. And their clients? They have to watch as the price falls back to normal. AI giveth and AI taketh away. It’s all a matter of timing. Or so they say. Somewhere in the fine print of their Terms of Use is the obscure statement “the Company and its employees may from time to time have a position in the stocks we recommend.”

It is an illegal scam to own stock, then recommend it to others for a fee knowing that their money inflows will push the stock price up. All the scammer needs to do is to sell after making a profit. AI has nothing to do with it.

Caveat emptor! Artificial Intelligence has its place, but buyer beware. It can also be part of a scam. It cannot predict something that moves randomly, nor can it convert a small percentage gain into a 4,000 percent gain.

*          *           *           *

P.S. Please do not contact me for my complete methodology. Sometimes 4 or 5 percent above a falling market will nevertheless produce a loss. Besides, any advantage can be squeezed by a dishonest person into an exaggerated claim. They say Diogenes never found the honest person he searched for as he wandered through the streets of ancient Athens. In the investment business, neither have I.