tl;dr
Nathan Smith, a high school student from rural Oklahoma, gave ChatGPT $100 to manage a portfolio of U.S. micro-cap stocks under $300 million market cap. Over four weeks, ChatGPT generated a 23.8% return, outperforming benchmarks like the Russell 2000 and biotech ETF XBI. The AI had full control over...
A high school student from rural Oklahoma, Nathan Smith, undertook an intriguing experiment by handing over $100 to ChatGPT to manage a portfolio of micro-cap stocks. Over four weeks, ChatGPT achieved a striking 23.8% return, vastly outperforming benchmarks like the Russell 2000 and the biotech ETF XBI, which gained just 3.9% and 3.5% respectively.
Smith's premise was straightforward yet bold: allow ChatGPT complete control over building and managing a portfolio exclusively composed of U.S.-listed micro-cap stocks with market capitalizations under $300 million. Unlike traditional algorithms, this experiment granted the AI full autonomy on decisions like position sizing and stop losses, with human intervention only occurring when ChatGPT contradicted itself.
The risk metrics paint a fascinating picture. With a Sharpe ratio of 0.9413, the portfolio carries a notable risk level, but the Sortino ratio of 2.0021 highlights strong upside potential with limited downside. Notably, ChatGPT demonstrated pragmatic decision-making by selling a high-performing stock, CADL, which contributed about 50% of profits, showing a lack of emotional attachment often absent even in some hedge funds.
Technically, Smith designed a practical yet elegant system featuring five core components: manual trade execution, portfolio tracking, daily results generation through Yahoo Finance data, and performance visualization against the S&P 500. ChatGPT selected stocks weekly while respecting the market cap constraints, and trades were manually logged by Smith.
Smith’s journey into quantitative finance began somewhat serendipitously, transitioning from coding challenges in Harvard’s CS50 course to discovering the compelling world of Python-driven finance. With nearly 1,000 GitHub commits in the past year and growing interest in his newsletter, he is deeply invested in this field. His aspirations include continuing this experiment for a full year and potentially pursuing a career in quantitative finance despite the demands of balancing schoolwork and standardized tests.