Validation Hub

The platform measures itself.

Orion is a rule-driven decision engine for XAUUSD. This page is the public surface for how the engine is held accountable. Counters below describe the measurement system, not future performance, and never substitute for your own due diligence.

Not a broker. Not a prop firm. For educational purposes only.

Snapshot
Generated 2026-06-02 17:18 UTC. Observation window: last 30 days.
Active
Observation Window

The shadow engine is grading itself in real time.

7d
continuous
Last evaluation recorded 2026-06-02 16:05 UTC.
Definition
Operational state of the live shadow harness. Active means the harness recorded a new evaluation within the last 6 hours.
Source
Internal shadow evaluation log
Cadence
Real-time
Last updated
2026-06-02 17:18 UTC
Live
Selectivity

The engine rejects far more than it publishes.

64
Considered
4
Published
64
Rejected
Considered counts every candidate evaluated by the shadow engine in the 30-day window. Published counts decisions released publicly. Rejected counts candidates the engine declined.
Definition
Count of candidates considered, published, and rejected in the observation window. Describes the production funnel, not future performance.
Source
Internal shadow evaluation log joined with the published decision log
Cadence
Daily snapshot, 30-day window
Last updated
2026-06-02 17:18 UTC
Live
Reject Discipline

The engine declines more candidates than it publishes.

100.0
%
64 declined of 64 evaluated.
Definition
Share of evaluated candidates that the engine declined to publish. A high value reflects selectivity by design, not engine quality.
Source
Internal shadow evaluation log
Cadence
Daily snapshot, 30-day window
Last updated
2026-06-02 17:18 UTC
Live
Attribution Coverage

Every published decision is accounted for.

0.0
%
0 resolved of 4 published.
Definition
Share of published decisions in the window that have a resolved outcome recorded. Describes completeness of the measurement pipeline, not the result distribution.
Source
Decision log joined with outcome log
Cadence
Daily snapshot, 30-day window
Last updated
2026-06-02 17:18 UTC
Pending
Published vs Shadow Agreement

Method consistency between the two engine paths.

Deferred from this release. Requires a dedicated, denominator-disciplined join surface so the rate cannot be misread as a performance claim.
Definition
Share of evaluations where the published decision and the shadow evaluation agree on direction. Measures internal consistency only. Not a measure of trade outcomes.
Source
Published decision log vs shadow evaluation log
Cadence
Daily snapshot, 30-day window
Last updated
2026-06-02 17:18 UTC
Pending
Historical Coverage Period

Span of XAUUSD history replayed against the current engine version.

Awaiting a versioned, append-only public summary record. Internal historical replay coverage exists but is not yet surfaced here.
Definition
Total span of historical XAUUSD data replayed against the current engine version. Describes scope of testing, not future performance.
Source
Historical replay harness summary
Cadence
Periodic, on each historical replay run
Last updated
2026-06-02 17:18 UTC
Pending
Historical Sample Size

Candidate setups evaluated during historical replay.

Published alongside the coverage period once the versioned summary record is exposed.
Definition
Number of candidate setups evaluated during historical replay for the current engine version. A count of evaluations, not a measure of quality.
Source
Historical replay harness summary
Cadence
Periodic, on each historical replay run
Last updated
2026-06-02 17:18 UTC
Methodology Statement

The rules that govern this audit.

This statement records the operating principles that produce every number on this page. It is reviewed when the integrity rules change and is intended to remain stable across engine versions.

  1. 01
    One method, one engine

    Published decisions, the shadow harness, and the historical replay are produced by the same rule set. There is no marketing engine separate from the production engine.

  2. 02
    Reject discipline by design

    Most candidates are declined. Selectivity is a property of the method, not a marketing claim. Rejection counts are recorded alongside publication counts.

  3. 03
    Append only

    Decisions, outcomes, and shadow evaluations are written to immutable tables. Historical values shown here are derived at read time and are never rewritten.

  4. 04
    Denominators visible

    Every ratio published on this page carries its numerator, denominator, and observation window. A percentage without its sample is treated as meaningless and is not shown.

  5. 05
    Honest degradation

    If upstream data is missing, stale, or insufficient, the metric is shown as Pending or Stale. The platform does not infer, smooth, or back fill.

  6. 06
    Out of scope is published

    Win rate, profit factor, expectancy, adapter weights, calibration thresholds, and per-decision audit rows are not published on this page. The boundary is documented, not hidden.

Signed
XAU Pro Academy, Engineering and Decision Operations.
Last reviewed 2026-05-31. The statement is versioned with the integrity rules it describes.
Validation Integrity Rules

How these numbers are kept honest.

  • Append-only snapshots

    Each snapshot is computed at request time from immutable event tables. Historical values are never rewritten.

  • Reproducible from archive

    Every metric is derivable from the underlying decision and outcome records. If a value cannot be reproduced, it is not displayed.

  • Denominators visible

    Ratio metrics always carry their numerator, denominator, and observation window. A percentage without its sample is treated as meaningless.

  • Honest degradation

    If upstream data is missing or stale, the metric is shown as Pending or Stale, never inferred or smoothed.

  • Volume-resistant

    No metric improves automatically as signal volume increases without a corresponding quality constraint.

  • Forward-compatible naming

    Names, definitions, and time windows are aligned with internal reporting so they remain stable for future investor reporting.

Out of scope

What this page does not claim.

This is not a performance dashboard. The hub deliberately excludes win rate, profit factor, expectancy, average R, per-regime breakdowns, adapter weights, calibration thresholds, decision audit rows, and any number that resembles a future-performance claim. Resolved trade outcomes per signal are listed publicly on the Trade Log. The methodology is described on the Orion page.