Why I built it

Risk management taught me to decide what would change my mind before looking at the answer, so here the failures get published with their reasons and their numbers, and the gates never move to fit a result. Along the way I felt compelled to write a few theses down, as well as a set of research notes that highlight some broader themes I am following closely.

How I work

Thresholds are fixed in code before a test runs, every statistic comes out of deterministic Python rather than a language model, each extra look at the same data raises the bar for the next one, and I assume the data is wrong until I have checked it.

Where to look if you want to poke at it

Every number traces back to the code that produced it, and corrections stay visible on the paper rather than being quietly folded in.

What this is not

Not a track record. No capital is at risk here, and a backtest cannot supply what a live book does. Not investment advice, and nothing here is a recommendation to buy or sell. The stock universe is survivorship-biased, so every raw performance figure is an upper bound.