Stock_Data
Stock_Data is the trading workbench I'm building for myself. It's a Windows desktop app written in Python. It pulls market data from my brokers' APIs (read-only, no orders), stores it in PostgreSQL, and puts it all in one place: price charts with my indicators, an options heatmap with expected moves, trade history, and cost basis.
I wanted to see what I actually did on the chart where I did it, so trades, option rolls, and events are drawn on the price action. Double-click a trade and the app jumps to that day in the history.
Where it stands: it runs every day on my desk. It has more than 2,600 automated tests, and it's built one small pull request at a time with AI coding agents doing the heavy lifting while I handle the logic and design. A companion app records option chains all day so I have clean history to learn from.
What's next is a neural-network model. I learned the ground rules from a mentor who ran market-beating models for about twenty years:
No look-ahead data. The model only sees what was known at the time.
Clean the data hard before it goes anywhere near training.
Normalize every input.
Run regressions between all the inputs so redundant ones don't fool the model.
After that comes testing it honestly, and only then any thought of automating.
See also: Trading TSLA
Comments