iteration 06 · 2026-07-14 · RUN-TERMINAL

DONE — PR #148 OPEN (stacked on #147) · §11 STOP CONDITION MET card 78substrate: ret_dur (FIRST duration)locked-margin 5.09e-05

#78 hoeffding_phi_squared_midreturn_duration

Branch feat/odb-ext/hoeffding_phi_squared_midreturn_duration → terrylica/mql5 PR #148 (base = iter-05 branch; loop never merges). First-run green: parity 5/5 · robustness 3/3 · rust:check 0 · check:all-gates 0 · pre-push ✓ · zero fix cycles. With this PR the run is COMPLETE: bootstrap + all 5 today-features PR-OPEN, zero deferrals.

H5 results row (ledger detail)

Substrate	Tier/TOL	Parity batch / stream (max-abs-dev, SPLIT)	non-vacuity	pinned witness	C&H
ret_dur (return leg RG8 gap-masked; duration leg unmasked — mask WITNESSED; dur derived in-state, zero extra plumbing)	ε 1e-9 rel / 1e-12 atol (LOCKED)	1.825e-15 / 0.0 — 11,200/11,200 within gate; max locked-budget margin 5.09e-05 (the iter-05 ruler applied prospectively)	11,200 compared; 4/4 cells; H3 golden 12,000; H1 ×11,200	EURUSD@5+0 → 0.033517958192861005 (RG8 gap window, bit-exact in the C&H spot-check)	A 8/8 · D 10/10 · 0 conceded

H2: 78-AVGRANK fork 73/80 breaches with 7 silents each individually verified tie-free (max-abs in the 1.502e-03 band) — integer-µs duration ties are real and load-bearing. Oracle sha 104b9c02… (replay byte-identical); golden sha 7b8a801b…. DISTINCT from card-63 Hoeffding D (disambiguated in kernel, oracle, and schema COMMENT).

Decisions + findings (recorded LOUD)

The campaign's one local-hook block: the BarBuilder impl file crossed the local advisory plugin's 1000-line ceiling at this iteration's +13-line wiring; suppressed with an honest first-line FILE-SIZE-OK marker (repo precedent). Repo gates unaffected (exit 0). Operator follow-up: split the bar-close feature block (Wave-243 pattern, task #521 family) after the chain merges.
Silents explained, not tolerated: every AVGRANK silent match is asserted !window_has_tie — average ≡ ordinal on tie-free windows by construction.
Clean-room: copBasic (GPL) never read; crypto odb-py #540 twin neither consulted nor executed (RG12).

In plain terms

Scores dependence of ANY functional shape between how far a bar's price moved and how long the bar took to form, on a 0 (independent) to 1 (fully dependent) scale. Captures move-size/time coupling that straight-line correlation would miss entirely.

What landed (lockstep, one feature commit — 19 files) + run closure

kernel hoeffding_phi_squared_midreturn_duration.rs (stable-ordinal permutation copula, faithful double-cumsum) + TWO-RING rolling state (dur in-state) + mod.rs
BarOrthogonalExtV1 field · BarBuilder 3-arg wiring · writer.rs BarRow (last field)
schema.sql DDL AFTER gini… · feature_metadata.yaml · regenerated docs/feature-catalog.html
oracle + oracle/golden jsonl + .sha256 sidecars (dual-dir) · parity (5) + robustness (3) suites · C&H record

Run closed at S7: chain #143 → #144 → #145 → #146 → #147 → #148 all OPEN, zero deferrals, zero conceded C&H gaps across all six iterations. Iters 07–11 stay NEXT IN QUEUE (operator-gated; agnostic verdicts to be minted before the second run).