Document S/R findings and isolate legacy gate features

README.md

@@ -430,7 +430,7 @@ Research-only flags, all off by default (the default report is byte-identical to
 |---|---|
 | `BACKTEST_HOLDOUT_SPLIT=YYYY-MM-DD` | Adds a `holdout` section: train (entries before) vs test (entries on/after), as disjoint books |
 | `BACKTEST_MIN_RR_SWEEP=1` | Sweeps the activation R:R floor against portfolio Sharpe. Combine with `BACKTEST_HOLDOUT_SPLIT` to sweep out-of-sample |
-| `BACKTEST_SR_VARIANT=<arm>` | Research-only S/R arm: `production_control`, `rr_aligned_control`, `rewrite`, `soft_zones`, `confirmed_rounds`, or `gate_v2` |
+| `BACKTEST_SR_VARIANT=<arm>` | Research-only S/R detector/gate arm; see `docs/research/sr-levels-and-exits.md` for the detector and hidden-feature matrices |
 | `BACKTEST_ENTRY_START=YYYY-MM-DD` | Restrict candidate entry dates to a validation window |
 | `BACKTEST_ENTRY_END=YYYY-MM-DD` | Restrict candidate entry dates to a training window |
 | `BACKTEST_SR_AUDIT=1` | Add momentum-slice candidate rows for paired S/R cohort comparison |

app/services/backtest_service.py

@@ -141,6 +141,9 @@ SR_RESEARCH_VARIANTS = {
     "soft_zones_legacy_primary",
     "confirmed_rounds_legacy_primary",
     "gate_v2_legacy_primary",
+    "legacy_geometry_neutral",
+    "legacy_pivots_only",
+    "legacy_traffic_grid_only",
 }
 
 
@@ -273,7 +276,19 @@ def _window_setups(
 
     sr_variant = _sr_research_variant()
     detector_variant = sr_variant.removesuffix("_legacy_primary")
-    if detector_variant in {"production_control", "rr_aligned_control"}:
+    if sr_variant == "legacy_geometry_neutral":
+        detected_levels = detect_sr_levels_legacy(
+            highs, lows, closes, volumes, neutral_strength=True
+        )
+    elif sr_variant == "legacy_pivots_only":
+        detected_levels = detect_sr_levels_legacy(
+            highs, lows, closes, volumes, include_volume_profile=False
+        )
+    elif sr_variant == "legacy_traffic_grid_only":
+        detected_levels = detect_sr_levels_legacy(
+            highs, lows, closes, volumes, include_pivots=False
+        )
+    elif detector_variant in {"production_control", "rr_aligned_control"}:
         detected_levels = detect_sr_levels_legacy(highs, lows, closes, volumes)
     else:
         detector_cap = 0 if detector_variant == "gate_v2" else MAX_LEVELS
@@ -331,7 +346,9 @@ def _window_setups(
         targets = _prune_floor_pinned_targets(targets)
         primary_min_rr = (
             1.5
-            if sr_variant == "production_control" or sr_variant.endswith("_legacy_primary")
+            if sr_variant == "production_control"
+            or sr_variant.endswith("_legacy_primary")
+            or sr_variant.startswith("legacy_")
             else float(activation.get("min_rr", 0.0))
         )
         primary = _select_primary_target(

app/services/sr_service.py

@@ -400,25 +400,36 @@ def detect_sr_levels_legacy(
     closes: list[float],
     volumes: list[int],
     tolerance: float = DEFAULT_TOLERANCE,
+    *,
+    include_volume_profile: bool = True,
+    include_pivots: bool = True,
+    neutral_strength: bool = False,
 ) -> list[dict]:
-    """Exact research control for the deployed pre-rewrite detector."""
+    """Deployed detector plus source-isolation controls for local research.
+
+    The defaults remain the exact production control. Keyword-only switches
+    allow a research arm to remove one source or neutralize strength without
+    copying or subtly changing the legacy implementation.
+    """
     if not closes:
         return []
 
     candidates: list[tuple[float, str]] = []
-    try:
+    if include_volume_profile:
-        for price in _legacy_volume_profile_nodes(highs, lows, closes, volumes):
+        try:
-            candidates.append((float(price), "volume_profile"))
+            for price in _legacy_volume_profile_nodes(highs, lows, closes, volumes):
-    except ValidationError:
+                candidates.append((float(price), "volume_profile"))
-        pass
+        except ValidationError:
-    try:
+            pass
-        pivots = compute_pivot_points(highs, lows, closes)
+    if include_pivots:
-        candidates.extend(
+        try:
-            (float(price), "pivot_point")
+            pivots = compute_pivot_points(highs, lows, closes)
-            for price in pivots.get("swing_highs", []) + pivots.get("swing_lows", [])
+            candidates.extend(
-        )
+                (float(price), "pivot_point")
-    except ValidationError:
+                for price in pivots.get("swing_highs", []) + pivots.get("swing_lows", [])
-        pass
+            )
+        except ValidationError:
+            pass
     if not candidates:
         return []
 
@@ -469,6 +480,9 @@ def detect_sr_levels_legacy(
         )
 
     _tag_levels(merged, closes[-1])
+    if neutral_strength:
+        for level in merged:
+            level["strength"] = 50
     merged.sort(key=lambda row: row["strength"], reverse=True)
     return merged
 

docs/research/sr-levels-and-exits.md

@@ -452,6 +452,52 @@ than a pristine holdout; do not sweep variants on it. No deployment follows
 automatically. A lower validation Sharpe or higher drawdown remains a no-ship
 result even when CAGR rises.
 
+## 8. Final detector-only result: no rewritten gate arm advances
+
+The last matrix froze production's effective gate (`primary min_rr=1.5`,
+activation `min_rr=2.0`) and varied only level detection/zone policy on entries
+through 2024-06-30. Corrected portfolio calendars end after the last position can
+resolve; there is no flat-cash tail.
+
+| arm | Sharpe | CAGR | MaxDD | qualified | net avg R | ex-top-5% |
+|---|---:|---:|---:|---:|---:|---:|
+| production control | **1.28** | **28.8%** | 21.4% | 676 | **0.230** | **0.066** |
+| rewrite + legacy primary | 0.96 | 21.0% | 22.2% | 1,200 | 0.037 | -0.102 |
+| soft zones + legacy primary | 1.08 | 25.1% | **17.8%** | 1,161 | 0.045 | -0.096 |
+| confirmed rounds + legacy primary | 1.14 | 22.1% | 20.2% | 570 | 0.189 | 0.045 |
+| gate v2 + legacy primary | 0.87 | 16.7% | 20.7% | 604 | 0.187 | 0.041 |
+
+No arm advances to validation. The raw rewrite retains only 249 of 676 production
+setups, removes 427 good setups, and adds 951 setups with negative expectancy.
+Round confirmation repairs the added cohort but still removes 430 production
+setups whose 30-day average (+0.680R) exceeds the additions (+0.511R). Uncapping
+recovers only 16 of those missing setups. The old detector averages 43.3 gate
+levels versus 15.0 rewritten and 24.0 rewritten-uncapped levels.
+
+**Standing no-ship decision:** keep both the deployed detector and the legacy 1.5
+primary-selection behavior in the trading path. The rewritten structure may only
+proceed as a separately computed display model. Do not merge this research branch
+into production as-is.
+
+### Hidden-feature isolation
+
+The remaining hypothesis is that the deployed detector accidentally measures
+long-memory historical price traffic rather than genuine S/R. A dedicated matrix
+holds the complete gate fixed and changes one legacy component at a time:
+
+- `legacy_geometry_neutral`: old locations, every merged strength fixed at 50;
+- `legacy_pivots_only`: unfiltered full-history pivots, no VP grid;
+- `legacy_traffic_grid_only`: old range-volume price grid, no pivots.
+
+Run on macOS:
+
+```bash
+.venv/bin/python scripts/run_sr_v2_matrix.py traffic --workers 14
+```
+
+Do not validate any traffic arm yet. First establish whether geometry, pivots, or
+the range-occupancy grid reproduces production on pre-2024 training data.
+
 **Next runs, if picked back up:**
 
 - A **per-name target model** for clear-air setups instead of a constant k×ATR. This

scripts/run_backtest_snapshot.py

@@ -53,6 +53,8 @@ def _parse_args() -> argparse.Namespace:
             "soft_zones", "confirmed_rounds", "gate_v2",
             "rewrite_legacy_primary", "soft_zones_legacy_primary",
             "confirmed_rounds_legacy_primary", "gate_v2_legacy_primary",
+            "legacy_geometry_neutral", "legacy_pivots_only",
+            "legacy_traffic_grid_only",
         ),
         default=None,
         help="Research-only S/R detector/gate arm.",

scripts/run_sr_v2_matrix.py

@@ -1,4 +1,4 @@
-"""Run the S/R v2 training matrix or one locked validation comparison.
+"""Run S/R detector, hidden-feature, or locked validation comparisons.
 
 This is a cross-platform orchestrator around ``run_backtest_snapshot.py``. It
 contains no backtest logic; every arm still runs through the production-parity
@@ -23,6 +23,12 @@ TRAINING_ARMS = (
     "gate_v2_legacy_primary",
 )
 LOCKABLE_ARMS = TRAINING_ARMS[1:]
+TRAFFIC_ARMS = (
+    "production_control",
+    "legacy_geometry_neutral",
+    "legacy_pivots_only",
+    "legacy_traffic_grid_only",
+)
 
 
 def _add_common(parser: argparse.ArgumentParser) -> None:
@@ -39,6 +45,11 @@ def _args() -> argparse.Namespace:
     commands = parser.add_subparsers(dest="command", required=True)
     train = commands.add_parser("train", help="Run all arms before 2024-07-01.")
     _add_common(train)
+    traffic = commands.add_parser(
+        "traffic",
+        help="Isolate legacy geometry, pivots, and price-traffic grid on training data.",
+    )
+    _add_common(traffic)
     validate = commands.add_parser(
         "validate",
         help="Run production control and one locked arm from 2024-07-01.",
@@ -75,17 +86,21 @@ def _run_arm(
     )
 
 
-def _train(args: argparse.Namespace) -> None:
+def _train(
-    for arm in TRAINING_ARMS:
+    args: argparse.Namespace,
+    arms: tuple[str, ...],
+    filename_prefix: str,
+) -> None:
+    for arm in arms:
         _run_arm(
             arm,
             args.snapshot,
             args.workers,
             entry_flag="--entry-end",
             entry_date="2024-06-30",
-            output=ROOT / "reports" / f"backtest-sr-v2-train-{arm}.json",
+            output=ROOT / "reports" / f"{filename_prefix}-{arm}.json",
         )
-    print("Training matrix complete. Lock one arm before running validation.")
+    print("Training matrix complete. Review results before running validation.")
 
 
 def _validate(args: argparse.Namespace) -> None:
@@ -118,7 +133,9 @@ def _validate(args: argparse.Namespace) -> None:
 def main() -> None:
     args = _args()
     if args.command == "train":
-        _train(args)
+        _train(args, TRAINING_ARMS, "backtest-sr-v2-train")
+    elif args.command == "traffic":
+        _train(args, TRAFFIC_ARMS, "backtest-sr-traffic-train")
     else:
         _validate(args)
 

tests/unit/test_detect_sr_levels.py

@@ -242,3 +242,28 @@ class TestDetectSrLevels:
         # The research control intentionally keeps the deployed detector's much
         # denser output instead of borrowing the rewrite's presentation cap.
         assert len(levels) > MAX_LEVELS
+
+    def test_legacy_geometry_neutral_changes_only_strength(self):
+        highs, lows, closes, volumes = _make_series(n=500)
+        control = detect_sr_levels_legacy(highs, lows, closes, volumes)
+        neutral = detect_sr_levels_legacy(
+            highs, lows, closes, volumes, neutral_strength=True
+        )
+        assert [level["price_level"] for level in neutral] == [
+            level["price_level"] for level in sorted(
+                control, key=lambda row: row["price_level"]
+            )
+        ]
+        assert all(level["strength"] == 50 for level in neutral)
+
+    def test_legacy_source_ablation_isolates_candidates(self):
+        highs, lows, closes, volumes = _make_series(n=500)
+        pivots = detect_sr_levels_legacy(
+            highs, lows, closes, volumes, include_volume_profile=False
+        )
+        traffic = detect_sr_levels_legacy(
+            highs, lows, closes, volumes, include_pivots=False
+        )
+        assert pivots and traffic
+        assert all(level["sources"] == ["pivot_point"] for level in pivots)
+        assert all(level["sources"] == ["volume_profile"] for level in traffic)