Evidence and Governance
Purpose
CENTAUR separates implementation progress from research evidence. A code change can be valid while an experimental conclusion remains invalid or pending.
Evidence Principles
- Preserve raw evidence.
- Do not modify an active experimental artifact to improve presentation.
- Identify the exact repository baseline used for an analysis.
- Record corrective changes separately from observations they invalidate.
- Do not treat configuration files as authorization.
- Do not promote an experiment before its gate conditions are satisfied.
EXP-003A Lesson
The central governance lesson from EXP-003A is that a zero-result scanner observation is not meaningful until scanner data orientation has been verified.
The ArraySetAsSeries defect demonstrated how an implementation error can masquerade as a research finding.
Current Evidence State
At bf19250:
- EXP-003A forensic finding: PASS.
- Corrected C0: still running.
- C0 CSV: protected from modification.
- Gate A: pending evaluation.
- EXP-003B: not authorized.
Baseline Identification
Current Wiki baseline: bf19250
Relevant corrective commit: 01b556e
Change Control
The local repository currently reports an uncommitted modification to MQL5/Experts/CentaurQuant.mqproj. It is not part of the documented baseline.
No Premature Promotion
The Wiki must not describe EXP-003B as active, successful, or authorized until the governing checkpoint permits it.
Likewise, architectural completeness must not be presented as proof of a profitable trading edge.