feat(kis-collection): finalize sqlite migration, add fallback resilience, and update WBS documentation

tools/build_honest_performance_guard_v1.py

@@ -28,6 +28,7 @@ from pathlib import Path
 ROOT = Path(__file__).resolve().parent.parent
 
 # 입력 파일
+PREDICTION_ACCURACY = ROOT / "Temp" / "prediction_accuracy_harness_v2.json"
 REBOUND_EFF  = ROOT / "Temp" / "rebound_sell_efficiency_v1.json"
 LATE_CHASE   = ROOT / "Temp" / "late_chase_attribution_v1.json"
 PROPOSAL_HIS = ROOT / "Temp" / "proposal_evaluation_history.json"
@@ -46,6 +47,30 @@ def load_json(p: Path) -> dict | list:
     return json.loads(p.read_text(encoding="utf-8"))
 
 
+def load_prediction_accuracy() -> dict:
+    data = load_json(PREDICTION_ACCURACY)
+    return data if isinstance(data, dict) else {}
+
+
+def current_t5_status() -> tuple[float | None, str]:
+    """WBS-7.2 source-of-truth shim.
+
+    Prefer the latest prediction accuracy harness when present. Do not fall back to
+    stale hardcoded percentages when the harness explicitly says sample=0.
+    """
+    data = load_prediction_accuracy()
+    if not data:
+        return None, "ARTIFACT_MISSING"
+
+    t5_sample = int(data.get("t5_sample") or 0)
+    t5_rate = data.get("t5_op_rate")
+    if t5_sample == 0:
+        return None, "DATA_GATED"
+    if isinstance(t5_rate, (int, float)):
+        return float(t5_rate), "OK"
+    return None, "DATA_MISSING"
+
+
 def main() -> int:
     rebound  = load_json(REBOUND_EFF)
     chase    = load_json(LATE_CHASE)
@@ -90,7 +115,8 @@ def main() -> int:
         })
 
     # ── (3) T+1 / T+5 KPI 추적 ─────────────────────────────────────────
-    # operational_report 에서 일치율 추출
+    # operational_report는 보고서 텍스트용 보조 원장이고,
+    # T+5 현재값은 prediction_accuracy_harness_v2.json을 우선한다.
     t1_rate = None
     t5_rate = None
     sections = op.get("sections", []) if isinstance(op, dict) else []
@@ -109,7 +135,11 @@ def main() -> int:
 
     # 직접 알려진 값 사용 (operational_report 에서 확인된 수치)
     if t1_rate is None: t1_rate = 47.28
-    if t5_rate is None: t5_rate = 35.86
+    live_t5_rate, live_t5_status = current_t5_status()
+    if live_t5_rate is not None:
+        t5_rate = live_t5_rate
+    elif t5_rate is None:
+        t5_rate = None
 
     kpi_tracker.append({
         "metric": "T+1_match_rate_pct",
@@ -119,14 +149,24 @@ def main() -> int:
         "status": "BELOW_TARGET" if t1_rate < 55.0 else "ON_TARGET",
         "note": "동전던지기(50%) 이하 — 신호 품질 개선 필요",
     })
-    kpi_tracker.append({
-        "metric": "T+5_match_rate_pct",
-        "current": t5_rate,
-        "target_min": 55.0,
-        "gap": round(55.0 - t5_rate, 2),
-        "status": "BELOW_TARGET" if t5_rate < 55.0 else "ON_TARGET",
-        "note": "T+5 35.86% — ANTI_LATE_ENTRY_GATE_V2 임계값 보정 시 개선 목표",
-    })
+    if t5_rate is None:
+        kpi_tracker.append({
+            "metric": "T+5_match_rate_pct",
+            "current": None,
+            "target_min": 55.0,
+            "gap": None,
+            "status": "DATA_GATED",
+            "note": f"T+5 current source={live_t5_status} — sample=0 or artifact missing; do not cite stale 35.86%",
+        })
+    else:
+        kpi_tracker.append({
+            "metric": "T+5_match_rate_pct",
+            "current": t5_rate,
+            "target_min": 55.0,
+            "gap": round(55.0 - t5_rate, 2),
+            "status": "BELOW_TARGET" if t5_rate < 55.0 else "ON_TARGET",
+            "note": "T+5 current source-of-truth read from prediction_accuracy_harness_v2.json",
+        })
 
     # ── (4) OUTCOME_TRUST_GAP ───────────────────────────────────────────
     # design_score 97.12 vs 실측 T+5 35.86% 간 신뢰도 괴리
@@ -134,7 +174,8 @@ def main() -> int:
         "design_score": rb_score,
         "actual_t5_pct": t5_rate,
         "gap_note": (
-            f"설계점수 rebound_efficiency={rb_score:.2f} vs 실측 T+5 일치율 {t5_rate}% — "
+            f"설계점수 rebound_efficiency={rb_score:.2f} vs 실측 T+5 일치율 "
+            f"{('DATA_GATED' if t5_rate is None else f'{t5_rate}%')} — "
             f"설계점수가 높아도 실제 수익성 지표(T+5)는 낮을 수 있음. "
             f"두 지표를 항상 물리적으로 분리해 표시해야 한다."
         ),
@@ -153,11 +194,14 @@ def main() -> int:
     print(f"\n  [T+1/T+5 KPI 현황]")
     for k in kpi_tracker:
         status_icon = "✗" if k["status"] == "BELOW_TARGET" else "✓"
-        print(f"    {k['metric']}: {k['current']}% (목표 ≥{k['target_min']}%) {status_icon}")
+        if k["current"] is None:
+            print(f"    {k['metric']}: DATA_GATED (목표 ≥{k['target_min']}%) {status_icon}")
+        else:
+            print(f"    {k['metric']}: {k['current']}% (목표 ≥{k['target_min']}%) {status_icon}")
         print(f"    → {k['note']}")
 
     print(f"\n  [보정루프 개선 경로]")
-    print(f"    T+5 35.86% → 50%+ 목표:")
+    print(f"    T+5 {'DATA_GATED' if t5_rate is None else f'{t5_rate}%'} → 50%+ 목표:")
     print(f"    Step 1. ALEG_V2_GATE1_BLOCK_PCT(3%) → 표본 누적 후 최적값 보정")
     print(f"    Step 2. DSD_V1 가중치 → logistic regression 최적화")
     print(f"    Step 3. K2 분할비율 0.5 → 30/70/40/60/50/50 backtest 비교")
@@ -191,7 +235,7 @@ def main() -> int:
         "correction_steps": [
             f"rebound_efficiency_score={rb_score:.2f} → 보고서 표시 시 [UNVALIDATED_DESIGN_SCORE: n={rb_combo}] 주석 필수",
             f"late_chase_attribution: samples=0 → 최소 {SAMPLE_MIN}건 표본 누적 후 chase_entry_rate 검증",
-            f"T+5 {t5_rate}% → 보정루프(calibration_registry.yaml) 기반 임계값 최적화로 50%+ 목표",
+            f"T+5 {'DATA_GATED' if t5_rate is None else f'{t5_rate}%'} → 보정루프(calibration_registry.yaml) 기반 임계값 최적화로 50%+ 목표",
         ],
     }