Gate 01
Offline ledger mining
Find hypotheses without spending MT5 runtime.
Trade-level provenance, family de-duplication, hour and non-hour regime checks.
Research validation
A research system that is allowed to say no.
BoltIQ treats no-go reports, buried hypotheses, and failed robustness checks as part of the product, not as embarrassing artifacts to hide.
Gate sequence
More runtime is earned, not assumed.
Each stage answers one question. If the answer is weak, the family is paused or buried before broader MT5 runs consume time.
Gate 02
Smoke test
Verify mechanics before monthly validation.
Telemetry token, order comment lineage, dashboard reconstruction, compile health.
Gate 03
Controlled MT5 window
Run only the smallest useful date window.
June-only, JJA-only, or guard-vs-control tests before broader expansion.
Gate 04
Promotion gate
Accept, hold, or bury the family.
Profit factor, drawdown, data quality, coverage, stress proxy, Monte Carlo, and no-go report.
No-go discipline
Rejected branches remain visible.
A serious trading operation needs memory. Re-testing the same fragile family is avoided unless new independent evidence exists.
frequency_probe / native_auto reuse unless independently revalidated
hour-union overfitting without non-hour evidence
families that fail June control before July/August expansion
candidates dependent on one oversized report or narrow window