What today looks like
Every number on this page is measured by the walk-forward engine and published in site/data/buyfinder.json. Nothing here is a forecast — a setup either beat simply owning SPY for the same number of days, or it did not.
Today's buy candidates
Live BUY signals from the last 5 trading days that are still open. Each shows how its exact setup (strategy × stock trend × market trend) did in the past, next to what a random entry into the same stock for the same number of days would have made (the tick mark). Edge is measured against SPY over the same days — a high win rate that still trails SPY is not worth taking.
The evidence behind this
Reference material. Nothing here changes what the verdict says — it is how the verdict was reached, and what it is not allowed to claim.
Does the rule work? out-of-sample validation, the test that cannot be fitted to history
Momentum leaders a pre-registered factor test, run once
All setups, full history every BUY setup, long-only, with costs and intervals
Every BUY setup, long-only, next-day-open entry, strategy's own exit, costs included. "Edge" = average return minus SPY over the same days. "Before" / "After" split the history at the out-of-sample date; a setup is confirmed only if it was positive before and its 95% interval stayed above zero after. Intervals resample whole months, because signals bunch up on market-wide selloff days. Click a column to sort.
Forward record signals written down when published, scored after 25 trading days
Every backtest above looks at the past. This is the only test that cannot be fitted to it: each signal the site publishes is written down the day it appears, then scored after 25 trading days. It tracks the signals of the strategies this build licenses for action — read from the build at runtime — not the Buy Finder rule.