Verification
Audit log
The methodology page states a standing promise: patterns get re-checked, not just checked once and left alone. This log is where we keep that promise honest. Every time we re-run our discipline against a pattern we’ve already shipped, whether the result is “still holds,” “found something and fixed it,” or “retired,” we say so here, in public, at the same altitude as the methodology page: what kind of check ran and what the outcome was, never the check’s internal content. A page that only ever says “trust our process” is a claim. A dated log of actual re-checks, including the ones that found something, is evidence.
Entry #001
- Date
- July 2, 2026
- Pattern (public name)
- Stop-Widening
- What was re-checked
- A newly adopted guard-check for independence from any single market’s mechanics, the ninth part of our verification discipline, added specifically to catch patterns that quietly depend on the specifics of one asset class (see “How we verify” for what this check protects against). We ran it against every pattern already locked before it existed, including Stop-Widening.
- Verdict
- Re-verified, revised.
- What changed
- The re-check found that one part of Stop-Widening’s definition (the threshold for how much widening has to happen before the pattern is even eligible to fire) was expressed in a way that only made sense for one kind of market. We corrected the definition so it adapts to each trader’s own trading instead of assuming one market’s units, re-ran the full audit process against the corrected version, and it came back clean. The pattern’s underlying behavioral finding, that widening a stop on a losing trade is a distinct, catchable habit, was never in question and did not change; this was a fix to how the pattern is defined, not a reversal of what it detects.
- What we won’t detail
- The exact form of the corrected threshold, and the specific math behind “how much widening” counts, stay internal. That’s the recipe, not the standard.
Read the full discipline this entry refers to on the methodology page.