Rule v3 (November 4, 2026), fixed before any results. Changes are dated and apply only to later weeks.
US-listed companies outside financials and real estate, with a share price of $1 or more, at least $1 million traded a day, clean price data, a full research report less than a week old, and an earnings report due within three months. Companies linked to the owner's own trades are excluded.
A company qualifies only if our research expects a big change in next-quarter earnings (30% or more either way). Small changes are close to a coin flip, so they are left out.
The eligible list is frozen and fingerprinted (SHA-256). The fingerprint is the random seed: one company is drawn from each size group (micro, small, mid, large), at most two per sector, and none featured in the previous 13 weeks.
The claim must agree with the underlying data; every statement must be supported by the research report; there must be no rating, price, target or advice wording; and every risk must be supported by the company's own filings. A company that fails is held back, shown with the reason, and the next one in the random order takes its place.
Risks come from a register that only changes on new evidence, and a shown risk is replaced only by a more serious one. Common, boilerplate flags are not shown as key risks.
Its clarity grade (our model's weekly read of crash risk), what happened next to 100 past stocks with the same grade (how many dropped 20% or more within 3 months, stayed within 20%, or rose 20% or more), and the measured facts behind the grade. The counts come from every stock we scored each week from January 2024 to June 2026, using only data available at the time.
60 trading days after the post, we mark where the share price landed against the post-day close: dropped 20% or more, stayed within 20%, or rose 20% or more. The record sets the number that dropped against the number their grades implied, so you can see whether the grades hold up.
No buy, hold or sell ratings, no price targets. The counts describe past stocks with the same grade, not a forecast for this one. The record only counts picks posted before the result, and misses stay on the record.
Research, not advice; not a recommendation to buy or sell.