SESSION HANDOVER — SNIPERGOLD_ML P3-S.2: FORMAL SPECIFICATION + CODE-CONFORMANCE AUDIT

Date       : 2026-08-22 (session close)
Status     : P3-S.0 / P3-S.1 CLOSED — research direction CHANGED to specification conformance
Forge      : https://forge.mql5.io/chiki2bum2/SniperGold_ML

1. RESEARCH DIRECTION CHANGE STATUS

HUMAN VERIFICATION (Human A / Human B / Adjudication) = explicitly CANCELLED.
HUMAN ANNOTATION = NOT REQUIRED for the current research path.
human_A_f7.csv / human_B_f7.csv / human_adjudicated_f7.csv = DO NOT continue/request.

The human annotation packages (ml/p3/smc_semantic/human_package/, machine_package/) remain preserved as historical/research artifacts — not an execution path. Do not delete without instructions.

2. REASON FOR THE CHANGE

Code A versus code B is not ground truth, and human visual annotation is too subjective/expensive to be the primary verification method. The more appropriate approach: validate whether the MQL5 implementation conforms to the formal specification of the SMC concept defined in advance. The code/reference implementation is only a secondary diagnostic tool, not the definition of truth.

3. NEW VALIDATION METHOD (P3-S.2 onwards)

SMC Concept
  -> Formal Specification
  -> Code-Conformance Audit
  -> Synthetic Specification Tests
  -> Historical Regression / Differential Checks
  -> Semantic Verdict

4. VERIFIED STATUS (checkpoint)

Forge HEAD (before this handover) : 8c1ef6c
Working tree                      : CLEAN
P3-S.0 commit                     : b519a34 (f7 event lifecycle fix)
P3-S.1 commit                     : 8c1ef6c (machine freeze + human package)
Machine v2 hash                   : 6d7c0b41aeb6d8ab64dcda9d0723f8e82abce771ab86fc9834ddeb874dd58583
Case set hash                     : da3a0d0154b7d4b2b853b088e1948fd780d7677fbb00929eef24fc4e0132733a
Human package                     : preserved (human_package/ 60 context files, template, protocol)

5. P3-S.0 STATUS — CLOSED

P3-S.0-BUG-001 DetectLiquidityGrabs f7 persistent-state/event-lifecycle bug
  = CONFIRMED + FIXED
Lifecycle  : NO_SWEEP -> SWEEP_ONSET -> valid <= InpSeqWindow(40) -> EXPIRED -> NO_SWEEP
Regression : 15/15 PASS (test_f7_lifecycle.py, R1-R6 + historical)
Golden     : 60 cases rebuilt (sampling unchanged)
Machine    : f7_v2 = FROZEN MACHINE REFERENCE

6. P3-S.1 STATUS — CLOSED

P3-S.1 Machine v2 integrity = PASS (integrity_audit_f7_v2.json)
Machine v2                 = FROZEN MACHINE IMPLEMENTATION OUTPUT
Human verification         = CANCELLED

7. SEMANTIC STATUS (not final)

f7 semantic definition   = NOT YET APPROVED
f10/f11 semantics        = NOT YET APPROVED
Liquidity Sweep verdict  = NOT YET ESTABLISHED

8. CURRENT RESEARCH QUESTION

Does DetectLiquidityGrabs() really implement the Liquidity Sweep concept according to the agreed formal specification?

Not: "does the machine output match human judgment?"

9. NEXT SESSION — P3-S.2

First target : Liquidity Sweep
Main function: DetectLiquidityGrabs()
Principle    : SPECIFICATION FIRST, CODE AUDIT SECOND

FORBIDDEN in P3-S.2:

modify production code
modify f7 / f10/f11 / OB / FVG / CHoCH / timeframe / ML
train model / optimize
human annotation
using AUC/PF to determine semantics

10. PROVENANCE

P2_SOURCE_SHA        : 8d330343af688e2de2b2a1d12ce19a582714709a
P2.6_MODEL_SHA       : 06df8452a112290ecb9bae1a3dbff6df492e084f8b872885d808bb0303ec6a70
FEATURE_CONTRACT_SHA : C44CC6F2B740C32D06F776BD7C3E669DC5A8A6DE0484230544EBFFCF517D38DD
DATASET_SHA          : e85a08616191fb7ce4bbe79c1754228ddbcf54562034a322747501c7aabe124e
Checkpoints          : b41059d (P3.2.2) -> 1e80833 (handover) -> b877fb6 (golden LS)
                       -> b519a34 (P3-S.0 fix) -> 8c1ef6c (P3-S.1 freeze)
symbol/tf            : XAUUSD / M15 (2017-01-01 .. 2026-08-20)

11. KEY ARTIFACTS

docs/P3_S_F7_EVENT_LIFECYCLE_FORENSIC.md        (P3-S.0)
docs/P3_S1_HUMAN_MACHINE_F7_VALIDATION.md       (P3-S.1 — human gate CANCELLED, package preserved)
docs/P3_SMC_SEMANTIC_GOLDEN_DATASET.md          (golden dataset phase)
ml/p3/smc_semantic/test_f7_lifecycle.py         (R1-R6, 15/15 PASS)
ml/p3/smc_semantic/integrity_audit.py           (+ output/integrity_audit_f7_v2.json)
ml/p3/smc_semantic/comparison.py                (extended; human path no longer used)
ml/p3/smc_semantic/output/machine_annotations_f7_v2.csv  (FROZEN, sha 6d7c0b41)
ml/p3/smc_semantic/output/machine_annotations_f7_v1_LEGACY.csv (preserved)
ml/p3/smc_semantic/human_package/ + machine_package/       (preserved, not an execution path)

4.6 KiB Raw Permalink Blame History