Live Pick Scorecard
Every other proof in this app is a backtest. This page is the live, out-of-sample record: what actually happened to the names the app put on screen. Each pick is entered at the next close after it was published — the daily jobs run post-close, so a same-day entry would be lookahead — and measured against SPY over the identical window.
The weight badge on each source is the weight it is actually given in the fused Today's Trades ranking. So this is a direct live test of that backtested choice: if the 1.0-weighted screener does not out-earn the 0.3-weighted nudges here over time, the weighting is wrong.
Window
Grading published picks against realized prices…
Why the t-stat is greyed out. A daily list scored at 20 days overlaps 19 of 20 days with the next day's, so these observations are heavily autocorrelated and the t-stat is optimistic. It is shown as a rough magnitude only — do not read |t| > 2 here as significance. The honest signal on this page is the relative ordering of sources measured the same way on the same days, plus whether excess is stable as the window lengthens.