He entrusted $ 100 to Chatgpt to play on the stock market, his bet flees in a few weeks

Chatgpt

Chatgpt is the Openai chatbot, based on the GPT artificial intelligence model, making it possible to answer all kinds of questions or requests. Available in free online version.

• License:
  Free license
• Author :
  OPENAI
• Operating systems:
  Windows 10/11, Apple Silicon macOS, online service, Android, iOS iPhone / iPad
• Category :
  IA

Nathan Smith began his experience four weeks ago with a simple strategy: giving total control to Chatgpt on his small portfolio of micro-capitalization action. The figures speak for themselves with 23.8 % yield against 3.9 % for the Russell 2000 and 3.5 % for the ETF Biotech Hbi. Clearly, the high school student has done much better than Wall Street algorithms.

He entrusts his portfolio of shares to Chatgpt

“When I looked at YouTube, I constantly received an advertisement with an AI action selector who claimed to feed his algorithm with all NYSE actions”explains Nathan Smith at the microphone of Decrypt. “I then did some research and I was surprised that no one tried a portfolio entirely directed by an LLM before”.

Nathan Smith gave Chatgpt a clear instruction: building the most solid action portfolios using only positions in equity of American micro-capital companies of less than $ 300 million. The goal is to generate maximum yields from June 27 to December 27, 2025.

Where Nathan Smith is distinguished by his approach, it is via the autonomy granted to AI unlike other trading algorithms. “The AI takes care of absolutely everything. The size of the positions, the stop-loss, etc. It is none of my decisions”explains the high school student. The only human intervention is when Chatgpt contradicts, false recurring that is called “hallucinations”.

Nathan Smith has calculated risk metrics that often concern professional traders. The Sharpe ratio is 0.9413, which is a high risk. The Sortino ratio is 2,0021, or strong rising gains with limited losses. A Sharpe ratio greater than 1.0 is generally considered good.

The chatbot is open source and available on the Github by Nathan Smith. AI remains objective and does not affect in its decisions. Chatgpt notably sold frame, an action that generated around 50 % profit in the portfolio. “I think Chatgpt knew that in micro-spaces, all gains can be erased instantly. Few hedge funds could make such a sharp decision”.

A technique that requires investment

Obviously, it is a question of using a chatbot, but the technique however requires prior knowledge and a little dedication. Nathan Smith has built five main features: purchase and sale to humans for new choices, treatment of the portfolio to follow transactions, generation of daily results using data from Yahoo Financegeneration of graphics to view performance compared to the S&P 500.

“Honestly, the configuration is quite simple”explains Nathan Smith speaking of a technique that extracts the reference prices from the API of Yahoo Finance in pandas data for analysis. Chatgpt chooses the shares once a week and still respects the capitalization limit of $ 300 million, while the high school student performs transactions and records the results.

It was by accident that this high school student in a small town in Oklahoma discovered his passion. “I have coded a little in the past by working on Harvard’s online CS50 course, but using C did not seem stimulating because of segmentation errors”he says.

“During the summer, I discovered the quantitative finance and the beauty of Python and I fell in love”. Nathan Smith now plans to extend the experience to a full year once the current one will end on December 27, 2025.

However, avoid getting started too. For what ? The test horizon is too short. Four weeks prove nothing: the markets are cyclical. Micro-capitalizations are volatile and sensitive, variations in one day are frequent.

On a small capital and with few trades, one or two good decisions sometimes give the illusion of a “Edge” nonexistent. In short, the tweezers are put if you want to try it.