Oklahoma Schoolboy’s ChatGPT Bot Outperforms Russell 2000

A schoolboy from rural Oklahoma entrusted ChatGPT with managing $100, achieving a return that significantly outpaced the market, reports Decrypt.

Nathan Smith’s experiment began four weeks ago with a simple idea: to give artificial intelligence full control over a portfolio of small-cap stocks and observe the outcome.

The return was 23.8%, while during the same period, the Russell 2000 and XBI indices rose by 3.9% and 3.5%, respectively.

“I constantly saw ads for AI stock pickers on YouTube, claiming ‘our algorithm analyzes all stocks on the NYSE, etc.’ I started digging and was surprised that no one had yet attempted to fully hand over portfolio management to a LLM,” Smith said.

The task was simple: to assemble the strongest possible portfolio of American small-cap stocks (less than $300 million). The goal was to maximize returns from June 27 to July 27.

The algorithm was given full autonomy.

“The AI literally manages everything. Position sizes, stop-losses, etc. — I don’t interfere,” Smith noted.

He calculated the risk levels. The Sharpe Ratio was 0.9413, indicating high risk, while the Sortino Ratio was 2.0021, suggesting good growth potential with limited drawdowns.

One of ChatGPT’s successful decisions was purchasing CADL, which contributed about 50% of the portfolio’s total profit.

“It sold CADL without hesitation. I think the AI correctly understood that in microcaps, profits can vanish in an instant. Not every hedge fund would dare to take such a step,” Smith said.

The young man created five main functions:

• manual buying/selling of new stocks;
• portfolio processing for tracking trades;
• daily result calculations via Yahoo Finance;
• generating comparison charts with the S&P 500 index.

“Honestly, it’s all quite simple. The system pulls prices through the API of Yahoo Finance and analyzes data using Pandas,” he noted.

ChatGPT selects stocks once a week, adhering to the capitalization limit, while the young man executes trades and records results. He plans to extend the experiment until December.

“It seems I’ve found my calling in life and I really hope to make it my profession,” Smith concluded.

Back in February 2024, institutional investors predicted more active use of AI in trading.