โ€บNavigation

Iteration 6 โ€” Q5: are the two flagged anomalies disqualifying? RESOLVED โ€” both held, neither disqualifies

Read-only evaluation ยท no ClickHouse query needed โ€” synthesised from committed verdict artifacts (read-only --readonly=2 access present) ยท matrix-evaluation-for-orthogonality loop ยท PR #543

Question (backlog Q5)

Do the two flagged anomalies โ€” the wold_R crypto inversion (conditional value โ‰ˆ 0.278, < 0.5; replicates) and the forex PBO 0.20 โ€” disqualify any standing grounded result? Required trust basis: replication + PBO/CSCV.

Method (read-only, zero new compute)

Synthesised from committed verdicts: FULL-GRID-SCALE-VERDICT.md (wold_R cross-substrate behaviour), E2-BATTERY-VERDICT.md + E4-FINAL-VERDICT.md (forex PBO), S0-PRE-REGISTRATION.md + STATUS-GROUNDING-AND-THRESHOLDS.md (dispositions). No new measurement โ†’ leakage_guard.py N/A.

Findings โ€” every claim with its trust basis

AnomalyWhat it isStanding dispositionTrust basis
wold_R crypto inversionOn crypto the metric inverts โ€” conditional value โ‰ˆ 0.278 (< 0.5), and it replicates across runs. On forex it validated 19/19 (FULL-GRID), so the inversion is crypto-specific (a likely sign/orientation issue, mechanism not yet confirmed).Already REPORTED-NOT-GATED (S0 disposition). It is not part of any gate, so the inversion changes no standing result. Do not use wold_R as a crypto gate until the inversion mechanism is understood.Replication (0.278 reproduces); pre-registered S0 disposition "reported-not-gated".
forex PBO 0.20Probability of Backtest Overfitting drifted 0.00 โ†’ 0.20 (E2/E4 forex). The certified selection-bias bound is PBO โ‰ค 0.20, so 0.20 sits at the boundary (a watch, not a breach).WATCH, and now PARKED. Forex is out of scope (crypto-only, operator directive). The forex PBO is an archived watch item โ€” re-confirm only if/when forex is reintroduced.CSCV / PBO (deflated-z family); E2-BATTERY + E4 committed.

Verdict

RESOLVED โ€” NO, neither anomaly disqualifies a standing crypto result.

The crypto FDR certificate (the stamped four; PBO 0.00 crypto, deflation z 22.15) is untouched by either anomaly.

Anti-inference guard (R7)

This iteration does not "fix" wold_R (no re-derivation of its crypto sign) and does not re-pass the forex PBO. It documents their hold disposition from committed evidence. wold_R remains off the crypto gate; the forex PBO re-confirmation is deferred. No metric is promoted to a gate.

Read-only & resource-cap proof (this firing)

GuardEvidence
ClickHouse READ-ONLY (R1)Read-only access present (--readonly=2); no query required โ€” committed-artifact synthesis (0 queries, 0 writes).
Resource cap (R2)No compute envelope launched. Box well under the load1 โ‰ค 24 watchdog.
Leakage discipline (R7)No new measurement โ†’ leakage_guard.py N/A. wold_R 0.278 and forex PBO 0.20 are committed, replicated/CSCV-based numbers.
Write surfaces (R3)Only this dashboard folder + PR #543 body.

What's next

SSoT: FULL-GRID-SCALE-VERDICT.md, E2-BATTERY-VERDICT.md, E4-FINAL-VERDICT.md, S0-PRE-REGISTRATION.md, STATUS-GROUNDING-AND-THRESHOLDS.md ยท audit twin 2026-05-26-forward-orthogonality-prediction/ ยท read-only evaluator ยท do NOT merge PR #543