Method

Screener, v5.0

For each name the ledger script reads twelve quarters of diluted EPS from the company's own SEC releases (8-K Exhibit 99.1, adjusted where the company publishes an adjusted figure, GAAP otherwise). Every figure is tied to two documents: it must reappear as the year-ago or prior-quarter comparator in a later release. A cell that cannot be read by script is not filled by hand; the name is marked ungradable instead.

The composite is 0.45 × CV score + 0.35 × Gap score + 0.20 × Slope score, each mapped onto fixed absolute bands so a score means the same thing in every sector and every quarter. CV is the dispersion of year-over-year EPS growth, Gap the latest quarter's growth, Slope the trend of trailing-twelve-month EPS.

Tiers: Top Quality at 78 and above, Strong 64 to 78, Moderate 48 to 64, Weak below 48. Gates cap a name at 50: CV above 1.00 (unless the §5.1 compounder waiver applies), a sign change, two quarters above +300%, or trailing EPS below 85% of its peak.

The roster is the All Stars universe: a name qualifies on six of the last eight quarterly vintages or a trailing run of four; the watchlist is one vintage short of either path.

Larry Williams study

Both forecast reports were digitized chart by chart (the printed cycle lines, blue windows and red reversal areas). Five tests score the windows, the red turns, the long-term-line turns, the index-level calls and the printed 75% claim against cached Yahoo closes. A Larry-grade run scores every chart by his own rubric; a dance test asks whether price turns when the line turns.

Every metric is reported next to a control: the same forecast line dropped at random dates on the same instrument, averaged over many placements. A result counts only where it beats that control.

What is not here

Nothing personal. The hub's household ledger and bank connections are read on a private machine and never reach this site or the repository it is built from. The build reads committed outputs only.