forked from chiki2bum2/SniperGold_ML
79 lines
2.5 KiB
Markdown
79 lines
2.5 KiB
Markdown
# P3-S22.3 — POPULATION REDERIVATION
| |||
| |||
```text
| |||
Date : 2026-08-26
| |||
Session : P3-S22.3 — RESEARCH M30 CONSTRUCTION REPAIR &
| |||
POPULATION REDERIVATION
| |||
Status : COMPLETE (research-only; corrected population re-derived)
| |||
Policy : docs/CODE_VERIFICATION_AND_SILENT_BUG_POLICY_v1.md
| |||
```
| |||
| |||
This document reports the exact change the corrected M30 UTC construction
| |||
makes to the historical Candidate Setup population and the frozen feature /
| |||
label rows.
| |||
| |||
## Input
| |||
| |||
- Feed: frozen XAUUSD M15, 2017-01-03 .. 2026-07-21 UTC scope (197,032 feed /
| |||
194,985 in-scope bars).
| |||
- Frozen chain: vectorized CHoCH/OB/FVG + F3SetupEngine (canonical == runtime,
| |||
P3-S17R2 FULL PARITY). Only the M30 gate differs (committed index-pair vs
| |||
corrected UTC); H4, M15, sweep, CHoCH, zones, F3 registry unchanged.
| |||
| |||
## Frozen baseline (reproduced exactly)
| |||
| |||
```text
| |||
created : 694
| |||
in-scope : 686
| |||
per-bar mismatch : 0
| |||
leads 594 + follow-ons 92 = 686
| |||
WIN 190 / LOSS 469 / UNRESOLVED 22 / AMBIGUOUS 5
| |||
```
| |||
| |||
## Corrected population (only the M30 gate substituted)
| |||
| |||
```text
| |||
created : 717
| |||
in-scope : 707
| |||
leads 610 + follow-ons 97 = 707
| |||
WIN 201 / LOSS 472 / UNRESOLVED 28 / AMBIGUOUS 6
| |||
```
| |||
| |||
## Candidate identity comparison (stable chain key)
| |||
| |||
| identity | count |
| |||
|---|---|
| |||
| shared | 484 |
| |||
| old-only | 202 |
| |||
| corrected-only | 223 |
| |||
| corrected in-scope total | 707 |
| |||
| |||
## M30 gate summary
| |||
| |||
| metric | value |
| |||
|---|---|
| |||
| nonzero (old / corrected) | 180,664 / 180,774 |
| |||
| gate diff full | 38,399 (19.49 %) |
| |||
| gate diff scope | 37,651 (19.31 %) |
| |||
| |||
Mechanism (spot-checked): shared 484 identities have IDENTICAL creation-bar
| |||
gate (0 % changed) and only ~10 % of window bars differ; the 202 old-only were
| |||
formed under a DIFFERENT creation-bar gate (~62 %) so their chain no longer
| |||
completes; the 223 corrected-only are formed under different gates instead.
| |||
| |||
## Label / feature impact (frozen contracts)
| |||
| |||
- Every shared identity has an IDENTICAL label (484/484).
| |||
- Label class counts change (WIN 201 / LOSS 472 / UNRESOLVED 28 / AMBIGUOUS 6).
| |||
- Feature rows: 202 removed (old-only), 484 retained (shared), 223 added
| |||
(corrected-only). No feature-contract change.
| |||
| |||
## Status
| |||
| |||
The corrected M30 construction MATERIALLY changes the Candidate Setup
| |||
population (686 -> 707) and the feature/label rows it populates. Production
| |||
MQL5 is unaffected (runtime reads broker UTC M30). The frozen old population
| |||
remains the historical reference; the corrected population is a CANDIDATE
| |||
corrected research dataset pending a separate validation phase.
| |||
| |||
*End of P3-S22.3 population rederivation.*
|