Branch feat/odb-ext/sign_markov_inout_strength_asymmetry_flux โ terrylica/mql5 PR #146 (base = iter-03 branch; loop never merges). First-run green everywhere: parity 5/5 ยท robustness 4/4 ยท rust:check 0 ยท check:all-gates 0 ยท pre-push โ ยท zero fix cycles.
| Substrate | Tier/TOL | Parity batch / stream (max-abs-dev, SPLIT) | non-vacuity | pinned witness | C&H |
|---|---|---|---|---|---|
| signs (RG8 gap-nulled, never-reset ring) | counts EXACT ยท value ฮต 1e-9/1e-12 (LOCKED) | 0.0 / 0.0 โ attacker-measured oracleโกgolden cross-diff: max_abs 0.0, bit-unequal 0 ร11,200 (transcribed IEEE op order) | 11,200 compared; zero-flux class = 5,420 bars pinned (48%, a value not a miss); H1 ร11,200 | EURUSD@5+0 โ 0.005076142131979683 ยท XAUUSD@25+0 โ exact 0.0 (3 of 5 witnesses are real zero-flux windows) | A 8/8 ยท D 10/10 ยท 0 conceded |
H2: 98-ROWNORM fork teeth in THREE harnesses (agnostic control 0.0056 ยท bootstrap probe 60/60 max 1.135 ยท Rust suite 60/60 in (1.13,1.14)). H4 range [0.0, 0.02010050251256283] bit-exact. Oracle sha 33a4217fโฆ (replay byte-identical); golden sha ae915175โฆ.
Treats the up/flat/down bar sequence as a 3-state cycle and measures the net probability current circulating around it. A preferred rotational direction (e.g., upโflatโdownโup) is a directional-regime signature invisible to simple averages of up and down counts.
sign_markov_inout_strength_asymmetry_flux.rs (fixed 3ร3 arrays โ H1 by construction) + rolling state + mod.rsBarOrthogonalExtV1 field ยท BarBuilder read-before-push wiring ยท writer.rs BarRow (last field)schema.sql DDL AFTER sequiturโฆ ยท feature_metadata.yaml ยท regenerated docs/feature-catalog.html.sha256 sidecars (dual-dir)Findings: transcribed-op-order โ literal 0.0 deviation at the ฮต tier (LOCKED gate untouched); zero-flux is a first-class value (5,420 bars). Next: iter 05 (#80 gini_correlation_asymmetry_signedreturn_spread โ the first two-input ret_spread substrate) stacks on this branch.