Unpredictability creates alpha: Fund picking

Yesterday, I wrote about a new paper by Lauren Cohen and his colleagues that trained an AI to forecast the investment decisions of fund managers. I talked about the insights for stock picking then. Today, I am going to talk about what we can learn from this research on how to pick the right fund manager.

As I mentioned yesterday, the AI was able to correctly predict on average 71% of fund managers’ buy, hold, and sell decisions of the next one to four quarters. But the power to forecast fund manager actions wasn’t equal across managers.

Funds with a longer track record were more easily predicted, which shouldn’t surprise anyone. The machine has more data to identify the preferences of individual fund managers and thus can better forecast future actions based on past trades.

However, being predictable is really not what you want in a fund manager. As I discussed yesterday, the stocks that are most likely to outperform are the stocks that don’t neatly fit into a style like value or growth. And something similar seems to happen with fund managers. Fund managers who are more predictable in their investments tend to underperform.

Fund performance and the predictability of fund managers

Source: Cohen et al. (2026)

What seems to be going on here is that fund managers who follow a rigorous process tend to end up in similar stocks as other fund managers. The less freedom there is in a fund manager’s process to deviate from the standard approach, the more a fund manager will end up buying and selling clear-cut ‘crowded stocks’ that fit neatly into a specific investment style.

Meanwhile, when fund managers are allowed to be creative and stray from a rigorous process, when they can buy and sell stocks that don’t neatly fit into a process but where the fund manager ‘sees something’ in the company, or ‘believes’ in the company or its management, that’s when they are more likely to generate alpha.

Investment management may be a quantitative job, but the quantitative part can increasingly be replaced by machines. What cannot be replaced is the creative part of the job. And fund managers that are more creative and have more ‘out-of-the-box’ ideas have a better chance of outperforming. Alpha increasingly is where AI can’t go.