ChatGPT Micro-Cap Experiment

Welcome to the repo behind my 6-month live trading experiment where ChatGPT manages a real-money micro-cap portfolio.

The Concept

Everyday, I kept seeing the same ad about having an some A.I. pick undervalued stocks. It was obvious it was trying to get me to subscribe to some garbage, so I just rolled my eyes. Then I started wondering, "How well would that actually work?".

So, starting with just $100, I wanted to answer a simple but powerful question:

Can powerful large language models like ChatGPT actually generate alpha (or at least make smart trading decisions) using real-time data?

Each trading day:

• I provide it trading data on the stocks in it's portfolio.

• Strict stop-loss rules apply.

• Everyweek I allow it to use deep research to reevaluate it's account.

• I track and publish performance data weekly on my blog. (https://substack.com/@nathanbsmith?utm_source=edit-profile-page)

  Research & Documentation

• Research Index

• Markdown Research Summaries (MD)

• Weekly Deep Research Reports (PDF)

Performance Example (6/30 – 7/25)

---

---

• Currently stomping on the Russell 2K.

Features of This Repo

Live trading scripts — Used to evaluate prices and update holdings daily

LLM-powered decision engine — ChatGPT picks the trades

Performance tracking — CSVs with daily PnL, total equity, and trade history

Visualization tools — Matplotlib graphs comparing ChatGPT vs Index

Logs & trade data — Auto-saved logs for transparency

Price data caching — Avoids unnecessary yfinance requests by storing daily quotes under cache/

Why This Matters

AI is being hyped across every industry, but can it really manage money without guidance?

This project is an attempt to find out, with transparency, data, and a real budget.

Tech Stack

Basic Python

Pandas + yFinance for data & logic

Matplotlib for visualizations

ChatGPT 4o for decision-making

Setup

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
# If you see errors about missing packages such as `schedule`, make sure this
# step completed successfully.

Set up access to your brokerage API by either exporting credentials or creating a .env file. Copy the provided .env.example to .env and add your Alpaca API credentials:

BROKER_API_KEY=your_api_key
BROKER_SECRET_KEY=your_secret_key
BROKER_BASE_URL=https://paper-api.alpaca.markets

Environment variables take precedence over values in the file. This project uses the paper trading API only, so no real money is automatically traded.

The src.broker module exposes helper functions to interact with the Alpaca paper API. Use place_order to submit trades, get_account to fetch account details and list_positions to inspect open positions. Additional helpers let you manage orders (list_orders, cancel_order, cancel_all_orders), view tradable assets via list_assets, and close positions with close_position or close_all_positions.

Follow Along

The experiment runs June 2025 to December 2025. Every trading day I will update the portfolio CSV file. If you feel inspired to do something simiar, feel free to use this as a blueprint.

Updates are posted weekly on my blog — more coming soon!

One final shameless plug: (https://substack.com/@nathanbsmith?utm_source=edit-profile-page)

Find a mistake in the logs or have advice? Please Reach out here: [email protected]

Example Usage

Run the trading script with your portfolio CSV and starting cash:

python main.py trade --portfolio my_portfolio.csv --cash 100

The trading script also saves a PNG graph under the graphs/ directory each time it runs. Open the generated file with any image viewer to see the latest performance chart. If a ticker's price history can't be retrieved (for example if yfinance has no data), the program prints a warning and skips that symbol. Skipped tickers are not written to the daily portfolio CSV and are ignored when calculating totals.

Cached price files are saved under cache/ as pickles. The script checks this folder each day and only queries yfinance when the file for that ticker and date doesn't exist.

Configuration File

You can store common settings in a config.yaml (or .json) file at the project root. Parameters defined here are used as defaults when running the trading script.

Example config.yaml:

default_cash: 100.0
default_stop_loss: 0.05
extra_tickers:
  - "^RUT"
  - "IWO"
  - "XBI"

Run the script using the config (values provided on the command line override the file):

python main.py trade --portfolio my_portfolio.csv --config config.yaml

Dashboard

Start the Flask dashboard to view the portfolio, trade log, performance graph, and a quick summary of the latest totals:

python main.py dashboard

Visit http://localhost:5000/ in your browser. Use the navigation links at the top of the page to switch between the Portfolio, Trade Log, Graph, and Summary views.

Automating Daily Runs

Use the scheduler to run the trading script every day.

python main.py schedule --portfolio my_portfolio.csv --cash 100 --time 09:00

Background execution

Keep the scheduler running even after you close the terminal:

nohup python main.py schedule --portfolio my_portfolio.csv --cash 100 --time 09:00 &

Cron example

Add this to your crontab to start at boot:

@reboot /usr/bin/python /path/to/main.py schedule --portfolio /path/to/my_portfolio.csv --cash 100 --time 09:00 >> /path/to/trade.log 2>&1

Disclaimer

All content in this repository is provided for educational purposes only and does not constitute financial advice. See DISCLAIMER.md for more information.