Ranking Before Prediction

Mine started with selection.
Years ago, working with a PHP-based stock selection system long before today’s ML ecosystem, I found myself repeatedly returning to the same uncomfortable realization:
I didn’t actually need accurate return forecasts — I needed a way to order stocks under uncertainty.
That distinction sounds minor. It isn’t.
This post is the first in a series of exploratory notes meant to reinterpret an old z-ranking–based stock selection framework using modern mathematical ideas: order statistics, ranking geometry, stability, and noise.
There is no code here. No backtests. No claims of optimality.
This is about conceptual foundations — and about understanding why certain simple systems keep reappearing across finance, language, and decision-making, even when theory lags behind practice.
Please read the full note in pdf