# Results — first pre-registered runs, 2026-08-22

Both designs frozen before outcomes (see the two prereg files). Runners kept beside this file; both rerunnable.

## pn-001 ORB by VXN tercile — **REFUTED as a system**
NQ+ES 1m 2010–2026, ~10,500 trades (quad Fridays excluded, VXN cuts frozen on 2010-15).
- No (symbol × OR window × tercile × exit) cell survives friction; best gross t ≈ +1.3–1.4; most cells NEGATIVE net of NQ 0.5pt / ES 0.25pt per side. Win rates 45–54%, expectancy ±0.1R.
- H1 (expectancy increasing in VXN) inconsistent: ES mildly supportive (best t=+1.69, 15m/1555), NQ contradicts on 30m. Not usable.
- Verdict → doctrine v10 rule 16. Matches the house prior: OR carries information as a *trigger inside a read*, not as a recipe.

## pn-070/071 GEX regime — **H1 PROMISING, accruing; H2 unreadable**
Archive spans only 2026-07-26 → (n≈20 days/symbol; ~11–15 classified after the ±8% dead zone).
- Breakouts on **−GEX** days: 8/8 winners across NQ+ES (NQ mean +151pt, ES +29pt); **+GEX** days chopped (NQ −46pt mean). NQ Welch t=+2.50 at n=3/8 — direction matches the noise claim and house regime doctrine, sample far too small.
- Wall-fade test (H2): n=1–3 per bucket; fades mostly lost; no read.
- Verdict → doctrine v10 rule 18 (WATCH). **Rerun `study_gex_regime.py` when the archive reaches ~60 classified days (~late Oct 2026); it needs no changes to rerun.**

## Disposition of the rest of the tree
- Collision #2 (fade vs break families) → doctrine v10 rule 17 (WATCH).
- Street-name → house-concept translator shipped to the Discord options-school channel (community/education use only).
- Everything else in the test queue stays queued (pn-030 VWAP reclaim next; B/C items after).
- Nothing from this tree enters the Desk Read synthesis prompt directly; the only path is pre-reg → doctrine → distiller, per house rule.