exp 20: add deterministic baseline (parallel=1) to isolate LightGBM non-determinism as root cause of exp18 vs R0 divergence

finish experiment 20 (exp/20-improve-the-risk-limit-reference-signal)
exp 20: add R0 same-env 5-seed TopkDropout baseline (fair base for R2/R3/R5; pre-reset exp18 baseline is not comparable across env resets)
2026-08-17 09:59:26 +00:00 · 2026-08-17 09:33:24 +00:00 · 2026-08-17 07:16:03 +00:00 · 2026-08-17 05:52:48 +00:00 · 2026-08-17 05:46:09 +00:00 · 2026-08-17 04:08:07 +00:00
47 changed files with 1448 additions and 1364 deletions
@@ -1,48 +0,0 @@
-#!/usr/bin/env python3
-"""Precompute the signal-quality gate series and save to pickle.
-
-Usage:
-    python precompute_signal_quality_gate.py <pred_path> <output_path> [topk] [lookback] [threshold]
-
-Example:
-    python precompute_signal_quality_gate.py \
-        /home/data/lake/mlruns/49/34165f27e4a34378ad54843a079a78c0/artifacts/pred.pkl \
-        /app/experiments/book/data/signal_quality_gate/sq_gate_5d_0.50.pkl \
-        10 5 0.5
-"""
-import sys
-import pickle
-from pathlib import Path
-
-# Add tac-qlib to path
-sys.path.insert(0, "/app/tac-qlib")
-
-from tac_qlib.contrib.strategy.signal_quality_gate import compute_signal_quality_gate
-
-if __name__ == "__main__":
-    if len(sys.argv) < 3:
-        print(__doc__)
-        sys.exit(1)
-
-    pred_path = sys.argv[1]
-    output_path = sys.argv[2]
-    topk = int(sys.argv[3]) if len(sys.argv) > 3 else 10
-    lookback = int(sys.argv[4]) if len(sys.argv) > 4 else 5
-    threshold = float(sys.argv[5]) if len(sys.argv) > 5 else 0.5
-
-    lake_root = "/home/data/lake"
-
-    print(f"Computing signal-quality gate: topk={topk}, lookback={lookback}, threshold={threshold}")
-    gate = compute_signal_quality_gate(
-        pred_path,
-        lake_root=lake_root,
-        topk=topk,
-        lookback=lookback,
-        threshold=threshold,
-    )
-
-    print(f"Gate: {gate.sum()}/{len(gate)} days open ({gate.mean():.1%})")
-
-    with open(output_path, "wb") as f:
-        pickle.dump(gate, f)
-    print(f"Saved to {output_path}")
@@ -1,115 +0,0 @@
-# Signal-quality gate backtest for 2021
-{%- set LAKE = TAC_LAKE_DIR %}
-{%- set UNIVERSE = "SPY,QQQ,DIA,IWM,MDY,VTI,VOO,VEA,VWO,VT,EFA,EEM,TLT,IEF,SHY,AGG,BND,LQD,HYG,JNK,EMB,GLD,SLV,USO,UNG,DBA,DBC,XLK,XLF,XLE,XLV,XLI,XLY,XLP,XLU,XLB,XLRE,ARKK,SMH,SOXX,IBB,XBI,ITA,XAR,ICLN,TAN,FDN,IGV,ESPO,REM" %}
-{%- set SP_FIELDS = "sp_ret,sp_ou_zscore,sp_ou_half_life,sp_ou_revert,sp_hmm_p_regime1,sp_hmm_state,sp_jump_ratio,sp_jump_flag,sp_jump_tail,sp_max_move,sp_rv1,sp_rv5,sp_rv22,sp_vol_ratio_5_22,sp_vol_ratio_1_22,sp_trend_slope_5,sp_trend_slope_20,sp_trend_slope_60,sp_logp,sp_hurst_exponent,sp_sig_level1_lead,sp_sig_level1_lag,sp_sig_level2_lead_lag,sp_sig_level2_lag_lead" %}
-{# Gate computed on-the-fly from signal + lake close prices #}
-
-qlib_init:
-    provider_uri: "{{ LAKE }}"
-    region: us
-    expression_cache: null
-    dataset_cache: null
-    calendar_provider:
-        class: tac_qlib.data.providers.LakeCalendarProvider
-        kwargs: { lake_root: "{{ LAKE }}", market: US }
-    instrument_provider:
-        class: tac_qlib.data.providers.LakeInstrumentProvider
-        kwargs: { lake_root: "{{ LAKE }}", market: US, markets: {} }
-    feature_provider:
-        class: tac_qlib.data.providers.LakeFeatureProvider
-        kwargs: { lake_root: "{{ LAKE }}", market: US }
-    exp_manager:
-        class: MLflowExpManager
-        module_path: qlib.workflow.expm
-        kwargs:
-            uri: "sqlite:///{{ LAKE }}/mlruns.db"
-            default_exp_name: "tac-rd-sq-gate-2021"
-
-task:
-    model:
-        class: LGBModel
-        module_path: qlib.contrib.model.gbdt
-        kwargs:
-            loss: mse
-            learning_rate: 0.05
-            num_leaves: 15
-            n_estimators: 200
-            colsample_bytree: 0.8
-            subsample: 0.8
-            subsample_freq: 1
-            reg_alpha: 0.01
-            reg_lambda: 0.01
-
-    dataset:
-        class: DatasetH
-        module_path: qlib.data.dataset
-        kwargs:
-            handler:
-                class: TACHandler
-                module_path: tac_qlib.contrib.data.handler
-                kwargs:
-                    instruments: "{{ UNIVERSE }}"
-                    start_time: 2015-01-03
-                    end_time: 2021-12-31
-                    fit_start_time: 2015-01-03
-                    fit_end_time: 2021-01-03
-                    freq: day
-                    lake_root: "{{ LAKE }}"
-                    market: US
-                    label: "Ref($close,-6)/Ref($close,-1)-1"
-                    feature_fields: "$open,$high,$low,$close,$vwap,$volume,{{ SP_FIELDS }}"
-                    infer_processors:
-                        - class: DropAllNaN
-                          kwargs: {}
-                        - class: ProcessInf
-                          kwargs: {}
-                        - class: CSRankNorm
-                          kwargs: {}
-                        - class: ZScoreNorm
-                          kwargs: {}
-                        - class: Fillna
-                          kwargs: {}
-            segments:
-                train: [2015-01-03, 2020-09-01]
-                valid: [2020-09-03, 2021-01-03]
-                test: [2021-01-04, 2021-12-31]
-
-    record:
-        - class: SignalRecord
-          module_path: qlib.workflow.record_temp
-          kwargs: {}
-        - class: SigAnaRecord
-          module_path: qlib.workflow.record_temp
-          kwargs: { ana_long_short: true, ann_scaler: 252 }
-        - class: PortAnaRecord
-          module_path: qlib.workflow.record_temp
-          kwargs:
-              config:
-                  strategy:
-                      class: SignalQualityGateStrategy
-                      module_path: tac_qlib.contrib.strategy.signal_quality_gate
-                      kwargs:
-                          signal: "<PRED>"
-                          lake_root: "{{ LAKE }}"
-                          gate_topk: 10
-                          gate_lookback: 5
-                          gate_threshold: 0.5
-                          gate_start: "2015-01-03"
-                          gate_end: "2021-12-31"
-                          topk: 10
-                          n_drop: 1
-                          only_tradable: true
-                          risk_degree: 0.95
-                  backtest:
-                      start_time: 2021-01-04
-                      end_time: 2021-12-31
-                      account: 1000000
-                      benchmark: SPY
-                      exchange_kwargs:
-                          codes: "{{ UNIVERSE }}"
-                          deal_price: $close
-                          freq: day
-                          open_cost: 0.0005
-                          close_cost: 0.0015
-                          min_cost: 5.0
-              risk_analysis_freq: 1d
@@ -1,115 +0,0 @@
-# Signal-quality gate backtest for 2024
-{%- set LAKE = TAC_LAKE_DIR %}
-{%- set UNIVERSE = "SPY,QQQ,DIA,IWM,MDY,VTI,VOO,VEA,VWO,VT,EFA,EEM,TLT,IEF,SHY,AGG,BND,LQD,HYG,JNK,EMB,GLD,SLV,USO,UNG,DBA,DBC,XLK,XLF,XLE,XLV,XLI,XLY,XLP,XLU,XLB,XLRE,ARKK,SMH,SOXX,IBB,XBI,ITA,XAR,ICLN,TAN,FDN,IGV,ESPO,REM" %}
-{%- set SP_FIELDS = "sp_ret,sp_ou_zscore,sp_ou_half_life,sp_ou_revert,sp_hmm_p_regime1,sp_hmm_state,sp_jump_ratio,sp_jump_flag,sp_jump_tail,sp_max_move,sp_rv1,sp_rv5,sp_rv22,sp_vol_ratio_5_22,sp_vol_ratio_1_22,sp_trend_slope_5,sp_trend_slope_20,sp_trend_slope_60,sp_logp,sp_hurst_exponent,sp_sig_level1_lead,sp_sig_level1_lag,sp_sig_level2_lead_lag,sp_sig_level2_lag_lead" %}
-{# Gate computed on-the-fly from signal + lake close prices #}
-
-qlib_init:
-    provider_uri: "{{ LAKE }}"
-    region: us
-    expression_cache: null
-    dataset_cache: null
-    calendar_provider:
-        class: tac_qlib.data.providers.LakeCalendarProvider
-        kwargs: { lake_root: "{{ LAKE }}", market: US }
-    instrument_provider:
-        class: tac_qlib.data.providers.LakeInstrumentProvider
-        kwargs: { lake_root: "{{ LAKE }}", market: US, markets: {} }
-    feature_provider:
-        class: tac_qlib.data.providers.LakeFeatureProvider
-        kwargs: { lake_root: "{{ LAKE }}", market: US }
-    exp_manager:
-        class: MLflowExpManager
-        module_path: qlib.workflow.expm
-        kwargs:
-            uri: "sqlite:///{{ LAKE }}/mlruns.db"
-            default_exp_name: "tac-rd-sq-gate-2024"
-
-task:
-    model:
-        class: LGBModel
-        module_path: qlib.contrib.model.gbdt
-        kwargs:
-            loss: mse
-            learning_rate: 0.05
-            num_leaves: 15
-            n_estimators: 200
-            colsample_bytree: 0.8
-            subsample: 0.8
-            subsample_freq: 1
-            reg_alpha: 0.01
-            reg_lambda: 0.01
-
-    dataset:
-        class: DatasetH
-        module_path: qlib.data.dataset
-        kwargs:
-            handler:
-                class: TACHandler
-                module_path: tac_qlib.contrib.data.handler
-                kwargs:
-                    instruments: "{{ UNIVERSE }}"
-                    start_time: 2015-01-03
-                    end_time: 2024-12-31
-                    fit_start_time: 2015-01-03
-                    fit_end_time: 2024-01-03
-                    freq: day
-                    lake_root: "{{ LAKE }}"
-                    market: US
-                    label: "Ref($close,-6)/Ref($close,-1)-1"
-                    feature_fields: "$open,$high,$low,$close,$vwap,$volume,{{ SP_FIELDS }}"
-                    infer_processors:
-                        - class: DropAllNaN
-                          kwargs: {}
-                        - class: ProcessInf
-                          kwargs: {}
-                        - class: CSRankNorm
-                          kwargs: {}
-                        - class: ZScoreNorm
-                          kwargs: {}
-                        - class: Fillna
-                          kwargs: {}
-            segments:
-                train: [2015-01-03, 2023-09-01]
-                valid: [2023-09-03, 2024-01-03]
-                test: [2024-01-02, 2024-12-31]
-
-    record:
-        - class: SignalRecord
-          module_path: qlib.workflow.record_temp
-          kwargs: {}
-        - class: SigAnaRecord
-          module_path: qlib.workflow.record_temp
-          kwargs: { ana_long_short: true, ann_scaler: 252 }
-        - class: PortAnaRecord
-          module_path: qlib.workflow.record_temp
-          kwargs:
-              config:
-                  strategy:
-                      class: SignalQualityGateStrategy
-                      module_path: tac_qlib.contrib.strategy.signal_quality_gate
-                      kwargs:
-                          signal: "<PRED>"
-                          lake_root: "{{ LAKE }}"
-                          gate_topk: 10
-                          gate_lookback: 5
-                          gate_threshold: 0.5
-                          gate_start: "2015-01-03"
-                          gate_end: "2024-12-31"
-                          topk: 10
-                          n_drop: 1
-                          only_tradable: true
-                          risk_degree: 0.95
-                  backtest:
-                      start_time: 2024-01-02
-                      end_time: 2024-12-31
-                      account: 1000000
-                      benchmark: SPY
-                      exchange_kwargs:
-                          codes: "{{ UNIVERSE }}"
-                          deal_price: $close
-                          freq: day
-                          open_cost: 0.0005
-                          close_cost: 0.0015
-                          min_cost: 5.0
-              risk_analysis_freq: 1d
@@ -1,35 +1,29 @@
 # TradeAC custom-qlib-code snapshot (auto-generated)
-# parent repo HEAD : e952feed0a66a20439f4f24ad5524233429cd0c3
+# parent repo HEAD : 125be7b96fb5975e798a0b4301eeb5809a8a181c
 #   tac-qlib/tac_qlib/contrib
 #   tac-qlib/tac_qlib/data
 # per-file hashes (git hash-object):
  1b6298c4a5652f2e863cbdc385a1014a570fcd59  tac-qlib/tac_qlib/contrib/__init__.py
-  861592c63edd6a0853a9cb174b5970435b135fc8  tac-qlib/tac_qlib/contrib/__pycache__/__init__.cpython-312.pyc
+  b8112569f9b2537c45b6535e1a505a207878d322  tac-qlib/tac_qlib/contrib/__pycache__/__init__.cpython-312.pyc
  c76a9f17f680e74eea766eff27f7624359749ed6  tac-qlib/tac_qlib/contrib/data/__init__.py
-  5c547a2ef92e075e550fe6d01508a2f1d3f536bc  tac-qlib/tac_qlib/contrib/data/__pycache__/__init__.cpython-312.pyc
-  4f656130d167e79dcaaeb7783a121f0b36852374  tac-qlib/tac_qlib/contrib/data/__pycache__/handler.cpython-312.pyc
-  0dd25ef161c6e0f15eafc84886e7e1381deb38c3  tac-qlib/tac_qlib/contrib/data/handler.py
+  8d5333ebd2b44165c50cba639ca2d4ac3fc7cfec  tac-qlib/tac_qlib/contrib/data/__pycache__/__init__.cpython-312.pyc
+  18cb37c0354184c49fa2e598396d7df0634cce0f  tac-qlib/tac_qlib/contrib/data/__pycache__/handler.cpython-312.pyc
+  871ff1e163c29261f140c3f53d42a41e6504c779  tac-qlib/tac_qlib/contrib/data/handler.py
  b151d139a0dcde87d74b21e7c4b729176ba5c39b  tac-qlib/tac_qlib/contrib/model/__init__.py
-  b1489f2fc0dee85f0a4f90b2e6ad545ed9c8967b  tac-qlib/tac_qlib/contrib/model/__pycache__/__init__.cpython-312.pyc
-  121ef237da1df1b8e21a561c3ad0db200b901339  tac-qlib/tac_qlib/contrib/model/__pycache__/rank_ensemble.cpython-312.pyc
-  74d0da348cbcc3700c96b6f4fe4391488e61efc5  tac-qlib/tac_qlib/contrib/model/__pycache__/rank_gbdt.cpython-312.pyc
-  d3f051f3a8650c42fedc7b367b966f7c74fb5789  tac-qlib/tac_qlib/contrib/model/rank_ensemble.py
-  d03e6611338918d4aac5eea4adf26f85a3763652  tac-qlib/tac_qlib/contrib/model/rank_gbdt.py
-  2c2f167b693f4366a769998e3c9d4804f29e31e0  tac-qlib/tac_qlib/contrib/strategy/__init__.py
-  f9cd9ab729e3248542ccc490af7adc2e51c71914  tac-qlib/tac_qlib/contrib/strategy/__pycache__/__init__.cpython-312.pyc
-  cd3133cfbd2556b25c106e39ae97fb128df0e326  tac-qlib/tac_qlib/contrib/strategy/__pycache__/ic_gate.cpython-312.pyc
-  6dd1c568a2961842793674390d5abffd1a0e71b8  tac-qlib/tac_qlib/contrib/strategy/__pycache__/optimal_stop.cpython-312.pyc
-  03b5e4d80da00800f1b108bee0735d3d18d856d1  tac-qlib/tac_qlib/contrib/strategy/__pycache__/regime_gate.cpython-312.pyc
-  a6a1c21ab71b62080830df45c4784b73c1531036  tac-qlib/tac_qlib/contrib/strategy/__pycache__/signal_quality_gate.cpython-312.pyc
-  755e3b139496a5e22b0328db45c8199c33029fbc  tac-qlib/tac_qlib/contrib/strategy/__pycache__/weekly_rebalance.cpython-312.pyc
-  519a1f4c05dbe0ac018ab8b779eb33d53b4dd545  tac-qlib/tac_qlib/contrib/strategy/ic_gate.py
+  ab958203f33a99d12c7d923b6efb435189231666  tac-qlib/tac_qlib/contrib/model/__pycache__/__init__.cpython-312.pyc
+  7478f6b0f6de419615c02d4d92b54529f689ef04  tac-qlib/tac_qlib/contrib/model/__pycache__/rank_ensemble.cpython-312.pyc
+  9f9014ddd9bce37490061312d51e8e6fe540fec4  tac-qlib/tac_qlib/contrib/model/__pycache__/rank_gbdt.cpython-312.pyc
+  ce77dea53f6a87c5379782709293bf8ff55b2c75  tac-qlib/tac_qlib/contrib/model/rank_ensemble.py
+  ccfe7d554989aa7f3e5a2128ae663e51b2207149  tac-qlib/tac_qlib/contrib/model/rank_gbdt.py
+  4afcf9058231111c412925f4c4b84e81d656db87  tac-qlib/tac_qlib/contrib/strategy/__init__.py
+  74e5ecbbbb20bb71fd5cd083383de4ce88476712  tac-qlib/tac_qlib/contrib/strategy/__pycache__/__init__.cpython-312.pyc
+  afaf562aeaa12cebc8529cd916153252e7e3c38a  tac-qlib/tac_qlib/contrib/strategy/__pycache__/optimal_stop.cpython-312.pyc
+  96a0a25201f0a1bb2fc2190e26228c5c0e711a79  tac-qlib/tac_qlib/contrib/strategy/hmm_risk.py
+  816de5d58ae23d996635d42331cf9fc8963d5dbe  tac-qlib/tac_qlib/contrib/strategy/momentum_gate.py
  79aaad9e39fcc740a773f4f63c512ce1086cfde0  tac-qlib/tac_qlib/contrib/strategy/optimal_stop.py
-  7bcee5f0b09cfa721440f1354f16f2dd9a112b12  tac-qlib/tac_qlib/contrib/strategy/regime_gate.py
-  16ab80b731aab6d8bb818525615d77c2d1fcb0c8  tac-qlib/tac_qlib/contrib/strategy/signal_quality_gate.py
-  fe60bacdfedd48617863be31f24b7c7daebfac5a  tac-qlib/tac_qlib/contrib/strategy/weekly_rebalance.py
  92e6e90eb0cd0a25142034560f27adb6b705b1a8  tac-qlib/tac_qlib/data/__init__.py
-  316bf4aa160cc8d15929ea648be03f4b4999667d  tac-qlib/tac_qlib/data/__pycache__/__init__.cpython-312.pyc
-  554a3f29d181b64effbf49a8161b32e7f93d8d3e  tac-qlib/tac_qlib/data/__pycache__/config.cpython-312.pyc
-  8b47f6d78ac046b6b7b2fb07bd7f3382773ffb73  tac-qlib/tac_qlib/data/__pycache__/providers.cpython-312.pyc
-  53c9007a928841fd3c3b08450f9a6520ce1ac091  tac-qlib/tac_qlib/data/config.py
-  8d0644f6f0d1efb94798ed444cc73e63b643459b  tac-qlib/tac_qlib/data/providers.py
+  0ed1ead6c1314a3f25784d453e54a15a8a04baaa  tac-qlib/tac_qlib/data/__pycache__/__init__.cpython-312.pyc
+  9609782800944c45b78bb58eaa7b51ba1b7f8f43  tac-qlib/tac_qlib/data/__pycache__/config.cpython-312.pyc
+  a85628d71d12cfe5b18b1c884c5d829c89594579  tac-qlib/tac_qlib/data/__pycache__/providers.cpython-312.pyc
+  686d36f6d101c547491ca866aa143aa542e17518  tac-qlib/tac_qlib/data/config.py
+  d9f839be30026f337754a3f015425a8efdbe8e2a  tac-qlib/tac_qlib/data/providers.py
@@ -64,13 +64,9 @@ def check_transform_proc(proc_l, fit_start_time, fit_end_time):


 def get_common_feature_fields(lake_root=None, market="US", timeframe="1d") -> List[str]:
-    """Discover feature columns present in *every* feature file of the lake.
+    """Discover ta-lib columns present in *every* features parquet file of the lake.

-    Walks the `family=ta|sp` partition layout (plus any legacy flat files).
-    TA and SP columns are disjoint by construction, so the common set is
-    computed per family (columns shared by all symbol files of that family),
-    then the per-family results are unioned. Returns sorted field names
-    (without the ``$`` prefix). Empty if no features are persisted.
+    Returns sorted field names (without the ``$`` prefix). Empty if no features are persisted.
    """
    cfg = LakeConfig(lake_root, market)
    feat_dir = cfg.features_dir(timeframe)
@@ -78,30 +74,16 @@ def get_common_feature_fields(lake_root=None, market="US", timeframe="1d") -> Li
        return []
    import pyarrow.parquet as pq

-    def _family_common(fam_dir: Path) -> set:
-        common = None
-        for p in sorted(fam_dir.glob("symbol=*.parquet")):
-            try:
-                cols = set(pq.read_schema(p).names) - set(NON_FEATURE_COLUMNS)
-            except Exception:  # pragma: no cover - skip unreadable files
-                continue
-            common = cols if common is None else (common & cols)
-            if not common:
-                break
-        return common or set()
-
-    common: set = set()
-    # family tier: features/market=*/timeframe=*/family=*/symbol=*.parquet
-    for fam in ("ta", "sp"):
-        fam_dir = feat_dir / f"family={fam}"
-        if fam_dir.is_dir():
-            common |= _family_common(fam_dir)
-    # legacy flat: features/market=*/timeframe=*/symbol=*.parquet
-    if (feat_dir / "family=ta").exists() or (feat_dir / "family=sp").exists():
-        pass  # family layout already covered
-    else:
-        common |= _family_common(feat_dir)
-    return sorted(common)
+    common = None
+    for p in sorted(feat_dir.glob("symbol=*.parquet")):
+        try:
+            cols = set(pq.read_schema(p).names) - set(NON_FEATURE_COLUMNS)
+        except Exception:  # pragma: no cover - skip unreadable files
+            continue
+        common = cols if common is None else (common & cols)
+        if not common:
+            break
+    return sorted(common) if common else []


 class DropAllNaN(processor_module.Processor):
@@ -56,6 +56,7 @@ import os
 from concurrent.futures import ThreadPoolExecutor
 from typing import List, Optional

+import numpy as np
 import pandas as pd

 from qlib.data.dataset import DatasetH
@@ -79,11 +80,15 @@ class RankICEnsembleLGBModel(RankICLGBModel):
        forwarded.
    """

-    def __init__(self, seeds: str = "42", parallel: int = 0, **kwargs):
+    def __init__(self, seeds: str = "42", parallel: int = 0, weight_mode: str = "equal", **kwargs):
        self.seeds = [int(s.strip()) for s in str(seeds).split(",") if s.strip()]
        if not self.seeds:
            raise ValueError("seeds must contain at least one integer")
        self.parallel = int(parallel)
+        if weight_mode not in ("equal", "rolling_ic"):
+            raise ValueError(f"weight_mode must be 'equal' or 'rolling_ic', got {weight_mode!r}")
+        self.weight_mode = weight_mode
+        self.rolling_ic_window = int(kwargs.pop("rolling_ic_window", 21))
        # drop seed/parallel handling from the base kwargs, keep everything else
        self._model_kwargs = dict(kwargs)
        super().__init__(**self._model_kwargs)
@@ -179,11 +184,44 @@ class RankICEnsembleLGBModel(RankICLGBModel):

    # -------------------------------------------------------------- predict
    def predict(self, dataset: DatasetH, segment="test") -> pd.Series:
-        """Average the per-seed predictions over the given segment."""
+        """Combine per-seed predictions.
+
+        ``weight_mode='equal'`` (default): simple average, as before.
+        ``weight_mode='rolling_ic'``: weight each seed by its trailing
+        per-day RankIC over the last ``rolling_ic_window`` days of the segment,
+        normalised to sum to 1 — adaptive ensemble blending that up-weights the
+        seed that is currently working (cheap alpha gain; same trained models).
+        """
        if not self._models:
            raise ValueError("model is not fitted yet!")
        preds = [m.predict(dataset, segment=segment) for m in self._models]
        if len(preds) == 1:
            return preds[0]
        frame = pd.concat(preds, axis=1)
-        return frame.mean(axis=1)
+        frame.columns = [f"seed{m.params.get('seed', i)}" for i, m in enumerate(self._models)]
+        if self.weight_mode == "equal":
+            return frame.mean(axis=1)
+
+        # rolling-IC blend: weight by per-day Spearman IC of each seed vs the
+        # cross-sectional mean prediction (proxy for the true label) on the last
+        # `rolling_ic_window` days of this segment. No lookahead: only past days
+        # of the segment are used; the final (trading) day is excluded from the
+        # window so the weights are causal.
+        mean_pred = frame.mean(axis=1)
+        dates = sorted(frame.index.get_level_values(0).unique())
+        win = [d for d in dates if d < dates[-1]][-self.rolling_ic_window :]
+        ics = {}
+        for col in frame.columns:
+            if not win:
+                ics[col] = 1.0
+                continue
+            sub = pd.DataFrame({"p": frame[col], "m": mean_pred})
+            vals = []
+            for d in win:
+                s = sub[sub.index.get_level_values(0) == d]
+                if len(s) >= 3 and s["p"].nunique() > 1 and s["m"].nunique() > 1:
+                    vals.append(s["p"].rank().corr(s["m"].rank()))
+            ics[col] = float(np.mean(vals)) if vals else 1.0
+        wsum = sum(ics.values()) or len(ics)
+        weights = {c: v / wsum for c, v in ics.items()}
+        return sum(frame[c] * weights[c] for c in frame.columns)
@@ -53,61 +53,23 @@ from qlib.workflow import R
 __all__ = ["RankICLGBModel", "rankic_feval"]


-def _group_averaged_rank(values: np.ndarray, gid: np.ndarray, offs: np.ndarray) -> np.ndarray:
-    """Averaged (tie-corrected) rank of ``values`` within each group, vectorized.
-
-    ``gid`` maps each row to its group id; ``offs`` holds the cumulative row
-    offsets so that group ``i`` occupies rows ``[offs[i], offs[i+1])``. Returns
-    the same result as ``pandas.Series.rank(method='average')`` applied per
-    group, but in one pass (``np.lexsort`` is the only non-linear step).
-    """
-    n = len(values)
-    order = np.lexsort((values, gid))
-    ord_rank = np.empty(n, dtype=np.float64)
-    ord_rank[order] = np.arange(n, dtype=np.float64) - offs[gid[order]] + 1.0
-    sg = gid[order]
-    sv = values[order]
-    newblock = np.empty(n, dtype=bool)
-    newblock[0] = True
-    newblock[1:] = (sg[1:] != sg[:-1]) | (sv[1:] != sv[:-1])
-    blockid = np.cumsum(newblock) - 1
-    block_mean = np.bincount(blockid, weights=ord_rank[order]) / np.bincount(blockid)
-    out = np.empty(n)
-    out[order] = block_mean[blockid]
-    return out
-
-
 def _per_day_spearman(preds: np.ndarray, labels: np.ndarray, group: np.ndarray) -> float:
    """Mean per-day Spearman rank correlation of preds vs labels.

    ``group`` holds the number of rows of each trading day (query group), in
    order. Days with <3 valid rows or a constant pred/label are skipped.
-
-    Vectorized: per-day Spearman == Pearson of the per-day rank transforms,
-    and the Pearson moments (``sum``, ``sum`` of products/squares) aggregate
-    over each day with ``np.bincount``. Runs ~10x faster than the per-day
-    ``pd.Series.rank()`` loop that preceded it — this feval is invoked on the
-    train and valid panels every boosting round, per seed.
    """
    if group is None or len(group) == 0:
        return 0.0
    offs = np.concatenate([[0], np.cumsum(group.astype(int))])
-    gid = np.repeat(np.arange(len(group)), group.astype(int))
-    rp = _group_averaged_rank(preds, gid, offs)
-    rl = _group_averaged_rank(labels, gid, offs)
-    n_g = group.astype(float)
-    s_p = np.bincount(gid, weights=rp)
-    s_l = np.bincount(gid, weights=rl)
-    s_pl = np.bincount(gid, weights=rp * rl)
-    s_pp = np.bincount(gid, weights=rp * rp)
-    s_ll = np.bincount(gid, weights=rl * rl)
-    cov = n_g * s_pl - s_p * s_l
-    var_p = n_g * s_pp - s_p ** 2
-    var_l = n_g * s_ll - s_l ** 2
-    denom = np.sqrt(var_p * var_l)
-    valid = (n_g >= 3) & (denom > 0)
-    corr = np.where(valid, cov / np.where(denom == 0, 1, denom), 0.0)
-    return float(corr[valid].mean()) if valid.any() else 0.0
+    vals = []
+    for i in range(len(group)):
+        s = slice(offs[i], offs[i + 1])
+        p, l = preds[s], labels[s]
+        if len(p) < 3 or np.std(p) == 0 or np.std(l) == 0:
+            continue
+        vals.append(np.corrcoef(pd.Series(p).rank(), pd.Series(l).rank())[0, 1])
+    return float(np.mean(vals)) if vals else 0.0


 def rankic_feval(preds, dataset):
@@ -1,11 +1,3 @@
-from .ic_gate import ICGateTopkDropoutStrategy  # noqa: F401
 from .optimal_stop import OptimalStopControl  # noqa: F401
-from .regime_gate import RegimeGateTopkDropoutStrategy  # noqa: F401
-from .weekly_rebalance import WeeklyRebalanceDropoutStrategy  # noqa: F401

-__all__ = [
-    "ICGateTopkDropoutStrategy",
-    "OptimalStopControl",
-    "RegimeGateTopkDropoutStrategy",
-    "WeeklyRebalanceDropoutStrategy",
-]
+__all__ = ["OptimalStopControl"]
@@ -0,0 +1,138 @@
+"""TopkDropout with HMM high-volatility + drawdown-pause risk gates.
+
+Gates NEW entries on two risk conditions (held names are never force-sold):
+
+1. **HMM high-vol pause**: when the cross-sectional mean of ``sp_hmm_p_regime1``
+   (HMM high-vol regime probability) on the signal date is >= ``hmm_pause_pct``,
+   new buys are paused. The time-series study showed HMM high-vol probability
+   pulses BEFORE sharp moves (regime-change cut) — pausing new exposure at the
+   boundary reduces drawdown from price over-reaction.
+2. **Drawdown pause**: when the account equity drawdown from its running peak
+   exceeds ``drawdown_pause_pct``, new buys are paused. This is the
+   ``drawdown_pause_pct`` risk-limit expressed inside the backtest (the pure
+   executor-side gate is documented as not expressible in a one-shot backtest).
+3. **Liquidity floor**: names whose 20-day average daily dollar volume is below
+   ``liquidity_floor_adv`` are dropped from BUY candidates (the proven mitigant
+   from exp-18: $5M floor cut drawdown 7.9%->5.4% at higher IR).
+
+Implementation: pre-filter the signal score before the base TopkDropout
+decision — non-held names get score 0 when any gate fires.
+
+Wired into a workflow yaml like:
+
+    strategy:
+        class: HmmRiskTopk
+        module_path: tac_qlib.contrib.strategy.hmm_risk
+        kwargs:
+            signal: "<PRED>"
+            topk: 10
+            n_drop: 2
+            only_tradable: true
+            risk_degree: 0.95
+            hmm_pause_pct: 0.70
+            drawdown_pause_pct: 8.0
+            liquidity_floor_adv: 5000000
+"""
+
+from __future__ import annotations
+
+import copy
+from typing import Dict
+
+import numpy as np
+import pandas as pd
+
+from qlib.backtest.decision import TradeDecisionWO
+from qlib.backtest.position import Position
+from qlib.contrib.strategy.signal_strategy import TopkDropoutStrategy
+
+__all__ = ["HmmRiskTopk"]
+
+
+class HmmRiskTopk(TopkDropoutStrategy):
+    """TopkDropoutStrategy with HMM high-vol pause + drawdown pause + liquidity floor."""
+
+    def __init__(
+        self,
+        *,
+        hmm_pause_pct: float = 0.70,
+        drawdown_pause_pct: float = 8.0,
+        liquidity_floor_adv: float = 0.0,
+        **kwargs,
+    ):
+        super().__init__(**kwargs)
+        self.hmm_pause_pct = float(hmm_pause_pct)
+        self.drawdown_pause_pct = float(drawdown_pause_pct)
+        self.liquidity_floor_adv = float(liquidity_floor_adv)
+        self._peak_equity = 0.0
+
+    # ------------------------------------------------------------- gates
+    def _hmm_high_vol(self, pred_date) -> bool:
+        """Cross-sectional mean HMM high-vol regime probability >= threshold."""
+        try:
+            from qlib.data import D
+
+            feat = D.features(D.instruments("all"), ["$sp_hmm_p_regime1"],
+                              start_time=pred_date, end_time=pred_date)
+            if feat is None or len(feat) == 0:
+                return False
+            p = feat["$sp_hmm_p_regime1"].dropna()
+            if len(p) == 0:
+                return False
+            return float(p.mean()) >= self.hmm_pause_pct
+        except Exception:
+            return False
+
+    def _drawdown_active(self, equity: float) -> bool:
+        if self.drawdown_pause_pct <= 0:
+            return False
+        self._peak_equity = max(self._peak_equity, equity)
+        if self._peak_equity <= 0:
+            return False
+        dd = (self._peak_equity - equity) / self._peak_equity * 100.0
+        return dd >= self.drawdown_pause_pct
+
+    def _illiquid(self, codes, asof) -> Dict[str, bool]:
+        if self.liquidity_floor_adv <= 0 or not codes:
+            return {}
+        from tac_qlib.risk_limits import dollar_adv
+
+        adv = dollar_adv(codes, market="US", asof=asof, lookback=20)
+        return {c: adv.get(str(c).upper(), 0.0) < self.liquidity_floor_adv for c in codes}
+
+    # ------------------------------------------------------------- decision
+    def generate_trade_decision(self, execute_result=None):
+        trade_step = self.trade_calendar.get_trade_step()
+        trade_start_time, trade_end_time = self.trade_calendar.get_step_time(trade_step)
+        pred_start_time, pred_end_time = self.trade_calendar.get_step_time(trade_step, shift=1)
+        pred_score = self.signal.get_signal(start_time=pred_start_time, end_time=pred_end_time)
+        if pred_score is None:
+            return TradeDecisionWO([], self)
+        if isinstance(pred_score, pd.DataFrame):
+            pred_score = pred_score.iloc[:, 0]
+
+        current_temp = copy.deepcopy(self.trade_position)
+        assert isinstance(current_temp, Position)
+        held = {c for c in current_temp.get_stock_list() if abs(current_temp.get_stock_amount(c)) > 1e-6}
+
+        equity = current_temp.get_cash()
+        for code in held:
+            mark = self.trade_exchange.get_deal_price(
+                stock_id=code, start_time=trade_start_time, end_time=trade_end_time, direction=1
+            )
+            if mark is not None and np.isfinite(mark):
+                equity += abs(current_temp.get_stock_amount(code)) * mark
+
+        hmm_pause = self._hmm_high_vol(str(pd.Timestamp(pred_start_time).date()))
+        dd_pause = self._drawdown_active(equity)
+        buys_paused = hmm_pause or dd_pause
+
+        pred_score = pred_score.copy()
+        if buys_paused or self.liquidity_floor_adv > 0:
+            new_codes = [c for c in pred_score.index if c not in held]
+            illiquid = self._illiquid(new_codes, str(pd.Timestamp(pred_start_time).date()))
+            for code in new_codes:
+                if buys_paused or illiquid.get(code, False):
+                    pred_score[code] = -1e9  # cannot enter today
+
+        return super().generate_trade_decision(execute_result)
@@ -1,117 +0,0 @@
-"""Realized-IC circuit breaker TopkDropout strategy.
-
-Subclass of ``qlib.contrib.strategy.signal_strategy.TopkDropoutStrategy`` that
-holds the book (issues NO orders) while the streaming realized RankIC of the
-deployed signal is below threshold — i.e. the model's cross-sectional
-predictions are no longer earning against realized forward returns. When the
-gate is open it behaves exactly like the reference TopkDropoutStrategy.
-
-The gate is evaluated per trade step on the trailing mean realized RankIC of
-the signal over the last ``ic_window`` trading days whose label is fully
-realized as of the decision date (no lookahead — a 5d fwd label ``close[t+6]/
-close[t+1]-1`` is only known at ``t+6``).
-
-Two wiring modes:
-
-* ``ic_gate``: a precomputed ``pd.Series`` indexed by datetime of booleans
-  (True = gate open / trade allowed). Computed once by the caller (e.g.
-  ``rd_backtest``) and looked up per step. Missing dates default to open.
-* realized-IC self-computation: when ``ic_min_rankic`` is given but no
-  ``ic_gate``, the strategy computes the per-date realized RankIC itself from
-  ``self.signal`` (the pred scores) and the lake 1d bars via
-  ``tac_qlib.risk_limits.realized_rankic_series``, then applies the same
-  trailing-window comparison. Works when instantiated from a workflow YAML
-  PortAnaRecord config (``lake_root`` / ``market`` must be provided).
-"""
-
-from __future__ import annotations
-
-import pandas as pd
-
-from qlib.backtest.decision import TradeDecisionWO
-from qlib.contrib.strategy.signal_strategy import TopkDropoutStrategy
-
-from tac_qlib.risk_limits import ic_circuit_breaker, realized_rankic_series
-
-__all__ = ["ICGateTopkDropoutStrategy"]
-
-
-class ICGateTopkDropoutStrategy(TopkDropoutStrategy):
-    """TopkDropout with a streaming realized-IC circuit breaker.
-
-    Parameters
-    ----------
-    topk, n_drop, method_sell, method_buy, hold_thresh, only_tradable,
-    forbid_all_trade_at_limit : same as ``TopkDropoutStrategy``.
-    ic_min_rankic : float — pause new trading while trailing realized RankIC is
-        below this threshold (0 disables the gate).
-    ic_window : int — trailing window for the realized RankIC mean (default 22).
-    ic_label_horizon : int — label horizon in trading days (default 6).
-    ic_min_obs : int — min realized labels before the gate arms (default 10).
-    ic_gate : pd.Series, optional — precomputed per-date gate (bool indexed by
-        datetime). When provided, it overrides self-computation.
-    lake_root, market : str — lake location for self-computed realized IC.
-    """
-
-    def __init__(
-        self,
-        *,
-        topk,
-        n_drop,
-        ic_min_rankic: float = 0.0,
-        ic_window: int = 22,
-        ic_label_horizon: int = 6,
-        ic_min_obs: int = 10,
-        ic_gate=None,
-        lake_root: str = "",
-        market: str = "US",
-        **kwargs,
-    ):
-        super().__init__(topk=topk, n_drop=n_drop, **kwargs)
-        self.ic_min_rankic = float(ic_min_rankic or 0.0)
-        self.ic_window = int(ic_window or 22)
-        self.ic_label_horizon = int(ic_label_horizon or 6)
-        self.ic_min_obs = int(ic_min_obs or 10)
-        self._ic_gate = ic_gate
-        self._realized_ic = None
-        self.lake_root = lake_root or ""
-        self.market = market or "US"
-
-    def _load_realized_ic(self):
-        if self._realized_ic is None:
-            pred_start_time, pred_end_time = self.trade_calendar.get_step_time(
-                self.trade_calendar.get_trade_step(), shift=-self.ic_label_horizon
-            )
-            pred = self.signal.get_signal(start_time=pred_start_time, end_time=pred_end_time)
-            if isinstance(pred, pd.DataFrame):
-                pred = pred.iloc[:, 0]
-            self._realized_ic = realized_rankic_series(
-                pred, self.lake_root, self.market, label_horizon=self.ic_label_horizon
-            )
-        return self._realized_ic
-
-    def _gate_open(self, trade_start_time) -> bool:
-        ts = pd.Timestamp(trade_start_time)
-        if self._ic_gate is not None:
-            # precomputed gate series: look up the latest known decision date <= ts
-            known = self._ic_gate[self._ic_gate.index <= ts]
-            if len(known):
-                return bool(known.iloc[-1])
-            return True
-        if self.ic_min_rankic <= 0:
-            return True
-        realized = self._load_realized_ic()
-        limits = {
-            "ic_min_rankic": self.ic_min_rankic,
-            "ic_window": self.ic_window,
-            "ic_min_obs": self.ic_min_obs,
-        }
-        tripped, _reason, _trail = ic_circuit_breaker(realized, ts, limits)
-        return not tripped
-
-    def generate_trade_decision(self, execute_result=None):
-        trade_step = self.trade_calendar.get_trade_step()
-        trade_start_time, _ = self.trade_calendar.get_step_time(trade_step)
-        if not self._gate_open(trade_start_time):
-            return TradeDecisionWO([], self)
-        return super().generate_trade_decision(execute_result)
@@ -0,0 +1,91 @@
+"""TopkDropout with a 1-day momentum entry-confirmation gate.
+
+Gates NEW entries on short-term momentum: a name that is not currently held
+may only be bought when its trailing 1-day return is above ``min_momentum``
+(Lag-1 autocorr ~ +0.45 in the time-series study => short-term momentum
+continuation). Held names are never force-sold by this gate — exits stay the
+pure TopkDropout rule.
+
+Implementation: override ``generate_trade_decision`` and zero out the signal
+score of any non-held name that fails the momentum check BEFORE calling the
+base TopkDropout decision, so it can never be selected as a buy candidate.
+This is a clean pre-filter: the rest of the strategy (top-k, n_drop, sizing,
+costs) is untouched.
+
+Wired into a workflow yaml like:
+
+    strategy:
+        class: MomentumGateTopk
+        module_path: tac_qlib.contrib.strategy.momentum_gate
+        kwargs:
+            signal: "<PRED>"
+            topk: 10
+            n_drop: 2
+            only_tradable: true
+            risk_degree: 0.95
+            min_momentum: 0.0
+"""
+
+from __future__ import annotations
+
+import copy
+
+import pandas as pd
+
+from qlib.backtest.decision import TradeDecisionWO
+from qlib.backtest.position import Position
+from qlib.contrib.strategy.signal_strategy import TopkDropoutStrategy
+
+__all__ = ["MomentumGateTopk"]
+
+
+class MomentumGateTopk(TopkDropoutStrategy):
+    """TopkDropoutStrategy gated on 1-day momentum for new entries."""
+
+    def __init__(self, *, min_momentum: float = 0.0, **kwargs):
+        super().__init__(**kwargs)
+        self.min_momentum = float(min_momentum)
+
+    def _momentum_ok(self, code, trade_start, trade_end) -> bool:
+        """True when the trailing 1-day return is above the momentum floor."""
+        try:
+            cur = self.trade_exchange.get_deal_price(
+                stock_id=code, start_time=trade_start, end_time=trade_end, direction=1
+            )
+        except Exception:
+            return False
+        if cur is None or cur != cur or cur <= 0:
+            return False
+        prev_start = trade_start - pd.Timedelta(days=5)
+        prev_end = trade_start - pd.Timedelta(seconds=1)
+        prev = self.trade_exchange.get_deal_price(
+            stock_id=code, start_time=prev_start, end_time=prev_end, direction=0
+        )
+        if prev is None or prev != prev or prev <= 0:
+            return False
+        return (cur / prev - 1.0) >= self.min_momentum
+
+    def generate_trade_decision(self, execute_result=None):
+        trade_step = self.trade_calendar.get_trade_step()
+        trade_start_time, trade_end_time = self.trade_calendar.get_step_time(trade_step)
+        pred_start_time, pred_end_time = self.trade_calendar.get_step_time(trade_step, shift=1)
+        pred_score = self.signal.get_signal(start_time=pred_start_time, end_time=pred_end_time)
+        if pred_score is None:
+            return TradeDecisionWO([], self)
+        if isinstance(pred_score, pd.DataFrame):
+            pred_score = pred_score.iloc[:, 0]
+
+        current_temp = copy.deepcopy(self.trade_position)
+        assert isinstance(current_temp, Position)
+        held = set(current_temp.get_stock_list())
+        held = {c for c in held if abs(current_temp.get_stock_amount(c)) > 1e-6}
+
+        # pre-filter: zero the score of non-held names that fail momentum
+        pred_score = pred_score.copy()
+        for code in pred_score.index:
+            if code in held:
+                continue  # never gate exits / re-balancing of held names
+            if not self._momentum_ok(code, trade_start_time, trade_end_time):
+                pred_score[code] = -1e9  # cannot enter today
+
+        return super().generate_trade_decision(execute_result)
@@ -1,215 +0,0 @@
-"""Regime-gate TopkDropout strategy.
-
-Subclass of ``qlib.contrib.strategy.signal_strategy.TopkDropoutStrategy`` that
-holds the book (issues NO orders) while a regime detector says the market is in
-an unfavorable state.  When the gate is open it behaves exactly like the
-reference TopkDropoutStrategy.
-
-Three detector types are supported (all causal — no lookahead):
-
-* ``dispersion``: cross-sectional standard deviation of 22-day rolling returns
-  across the universe.  Gate closes when CS dispersion < threshold (low
-  dispersion means the spread between winners and losers is too narrow for
-  TopkDropout to exploit).
-* ``vol``: cross-sectional mean of 22-day rolling realized volatility.  Gate
-  closes when avg vol is outside a band ``[vol_low, vol_high]`` (strategy
-  needs moderate vol — too calm or too turbulent both hurt).
-* ``hmm``: pre-computed HMM posterior for regime 1 (``sp_hmm_p_regime1``).
-  Gate closes when posterior < threshold (model is not confident the calm
-  regime is active).
-
-The gate is provided as a precomputed ``pd.Series`` of booleans indexed by
-datetime (True = trade allowed).  The companion ``compute_regime_gate``
-function builds this series from lake bars; call it once before backtesting
-and pass the result as the ``regime_gate`` parameter.
-"""
-
-from __future__ import annotations
-
-import pandas as pd
-
-from qlib.backtest.decision import TradeDecisionWO
-from qlib.contrib.strategy.signal_strategy import TopkDropoutStrategy
-
-__all__ = ["RegimeGateTopkDropoutStrategy", "compute_regime_gate"]
-
-
-class RegimeGateTopkDropoutStrategy(TopkDropoutStrategy):
-    """TopkDropout with a regime-gate circuit breaker.
-
-    Parameters
-    ----------
-    topk, n_drop, method_sell, method_buy, hold_thresh, only_tradable,
-    forbid_all_trade_at_limit : same as ``TopkDropoutStrategy``.
-    regime_gate : pd.Series — precomputed per-date gate (bool indexed by
-        datetime).  True = trade allowed, False = no orders.  Missing dates
-        default to open (trade allowed).
-    """
-
-    def __init__(self, *, regime_gate=None, **kwargs):
-        super().__init__(**kwargs)
-        self._regime_gate = regime_gate
-
-    def _gate_open(self, trade_start_time) -> bool:
-        if self._regime_gate is None:
-            return True
-        ts = pd.Timestamp(trade_start_time)
-        known = self._regime_gate[self._regime_gate.index <= ts]
-        if len(known):
-            return bool(known.iloc[-1])
-        return True  # default open if no history yet
-
-    def generate_trade_decision(self, execute_result=None):
-        trade_step = self.trade_calendar.get_trade_step()
-        trade_start_time, _ = self.trade_calendar.get_step_time(trade_step)
-        if not self._gate_open(trade_start_time):
-            return TradeDecisionWO([], self)
-        return super().generate_trade_decision(execute_result)
-
-
-# ---------------------------------------------------------------------------
-# Precomputation helper
-# ---------------------------------------------------------------------------
-
-def compute_regime_gate(
-    detector: str,
-    threshold: float = 0.0,
-    *,
-    lake_root: str = "",
-    market: str = "US",
-    start: str = "2015-01-03",
-    end: str = "2026-08-19",
-    vol_low: float = 0.0,
-    vol_high: float = 999.0,
-    hmm_field: str = "sp_hmm_p_regime1",
-) -> pd.Series:
-    """Build a per-date regime gate series from lake bars.
-
-    Parameters
-    ----------
-    detector : str — ``"dispersion"``, ``"vol"``, or ``"hmm"``.
-    threshold : float — for ``dispersion``: min CS dispersion to allow trading.
-        For ``hmm``: min HMM posterior to allow trading.
-        Ignored for ``vol`` (uses ``vol_low``/``vol_high`` band instead).
-    lake_root, market : str — lake location.
-    start, end : str — date window.
-    vol_low, vol_high : float — annualized vol band for the ``vol`` detector.
-    hmm_field : str — HMM feature column name for the ``hmm`` detector.
-
-    Returns
-    -------
-    pd.Series — bool, indexed by datetime.  True = trade allowed.
-    """
-    from tac_qlib.data.config import LakeConfig, resolve_lake_root
-
-    cfg = LakeConfig(resolve_lake_root(lake_root or None), market)
-    symbols = _universe_symbols(cfg)
-    close_df, vol_df = _load_daily_bars(symbols, cfg, start, end)
-    if close_df.empty:
-        return pd.Series(dtype=bool)
-
-    if detector == "dispersion":
-        return _dispersion_gate(close_df, threshold)
-    elif detector == "vol":
-        return _vol_gate(close_df, vol_low, vol_high)
-    elif detector == "hmm":
-        return _hmm_gate(cfg, symbols, threshold, start, end, hmm_field)
-    else:
-        raise ValueError(f"Unknown detector: {detector!r}")
-
-
-def _universe_symbols(cfg) -> list:
-    """Read symbols from the lake symbols.parquet."""
-    import pathlib
-
-    sp = cfg.lake_root / "symbols.parquet"
-    if sp.exists():
-        df = pd.read_parquet(sp)
-        col = "symbol" if "symbol" in df.columns else df.columns[0]
-        return sorted(df[col].astype(str).str.upper().tolist())
-    return []
-
-
-def _load_daily_bars(symbols, cfg, start, end):
-    """Load daily close prices for all symbols into a wide DataFrame."""
-    closes = {}
-    vols = {}
-    for sym in symbols:
-        p = cfg.bar_path("1d", sym)
-        if not p.exists():
-            continue
-        try:
-            df = pd.read_parquet(p)
-        except Exception:
-            continue
-        if not len(df):
-            continue
-        tcol = df["t"] if "t" in df.columns else df["date"]
-        ts = pd.to_datetime(tcol)
-        df = df.assign(_t=ts).set_index("_t").sort_index()
-        df = df.loc[start:end]
-        if len(df) < 22:
-            continue
-        closes[sym] = df["c"]
-        if "v" in df.columns:
-            vols[sym] = df["v"]
-    close_df = pd.DataFrame(closes)
-    vol_df = pd.DataFrame(vols) if vols else None
-    return close_df, vol_df
-
-
-def _dispersion_gate(close_df, threshold):
-    """Cross-sectional dispersion of 22-day rolling returns."""
-    if close_df.empty or close_df.shape[1] < 2:
-        return pd.Series(dtype=bool)
-    ret = close_df.pct_change(22)
-    cs_disp = ret.std(axis=1)
-    gate = cs_disp >= threshold
-    gate.iloc[:22] = True  # warmup: allow trading
-    return gate
-
-
-def _vol_gate(close_df, vol_low, vol_high):
-    """Cross-sectional mean of 22-day rolling realized vol."""
-    if close_df.empty or close_df.shape[1] < 2:
-        return pd.Series(dtype=bool)
-    import numpy as np
-    log_ret = np.log(close_df / close_df.shift(1))
-    rv22 = log_ret.rolling(22).std() * (252 ** 0.5)
-    cs_mean_vol = rv22.mean(axis=1)
-    gate = (cs_mean_vol >= vol_low) & (cs_mean_vol <= vol_high)
-    gate.iloc[:22] = True  # warmup
-    return gate
-
-
-def _hmm_gate(cfg, symbols, threshold, start, end, hmm_field):
-    """HMM regime posterior gate from persisted SP features."""
-    feat_root = cfg.lake_root / "features"
-    all_posteriors = {}
-    for sym in symbols:
-        # check both ta and sp family paths
-        for family in ("sp", "ta"):
-            p = feat_root / f"market=US" / f"timeframe=1d" / f"family={family}" / f"symbol={sym}.parquet"
-            if not p.exists():
-                continue
-            try:
-                df = pd.read_parquet(p)
-            except Exception:
-                continue
-            if hmm_field not in df.columns:
-                continue
-            tcol = df["t"] if "t" in df.columns else df["date"]
-            ts = pd.to_datetime(tcol)
-            s = pd.Series(df[hmm_field].values, index=ts, name=sym)
-            s = s.loc[start:end].dropna()
-            if len(s) > 0:
-                all_posteriors[sym] = s
-                break
-    if not all_posteriors:
-        # no HMM features found — default open
-        idx = pd.date_range(start, end, freq="B")
-        return pd.Series(True, index=idx)
-    post_df = pd.DataFrame(all_posteriors)
-    cs_mean = post_df.mean(axis=1)
-    gate = cs_mean >= threshold
-    return gate
@@ -1,301 +0,0 @@
-"""Signal-quality gate TopkDropout strategy.
-
-Subclass of ``qlib.contrib.strategy.signal_strategy.TopkDropoutStrategy`` that
-holds the book (issues NO orders) when the model's recent prediction accuracy
-is below a threshold.  When the gate is open it behaves exactly like the
-reference TopkDropoutStrategy.
-
-Unlike the regime gate (which asks "is the market calm?"), the signal-quality
-gate asks "are my predictions accurate?" — and works across ALL years.
-
-The gate is provided as a precomputed ``pd.Series`` of booleans indexed by
-datetime (True = trade allowed).  The companion ``compute_signal_quality_gate``
-function builds this series from a pred.pkl and lake bars; call it once before
-backtesting and pass the result as the ``signal_quality_gate`` parameter.
-"""
-
-from __future__ import annotations
-
-import numpy as np
-import pandas as pd
-
-from qlib.backtest.decision import TradeDecisionWO
-from qlib.contrib.strategy.signal_strategy import TopkDropoutStrategy
-
-__all__ = ["SignalQualityGateStrategy", "compute_signal_quality_gate"]
-
-
-class SignalQualityGateStrategy(TopkDropoutStrategy):
-    """TopkDropout with signal-quality gate overlay.
-
-    When ``lake_root`` is provided the gate is computed on-the-fly from the
-    signal (``<PRED>``) and close prices — no precomputed gate file needed.
-    This ensures the gate matches the model that is actually generating the
-    predictions (critical when the model is retrained each year).
-
-    Parameters
-    ----------
-    topk, n_drop, method_sell, method_buy, hold_thresh, only_tradable,
-    forbid_all_trade_at_limit : same as ``TopkDropoutStrategy``.
-    signal_quality_gate : pd.Series — precomputed per-date gate (bool).
-    signal_quality_gate_path : str — path to pickled gate Series.
-    lake_root : str — lake root for on-the-fly gate computation (preferred).
-    gate_topk, gate_lookback, gate_threshold : int/float — gate params.
-    gate_start, gate_end : str — date window for loading close prices.
-    """
-
-    def __init__(self, *, signal_quality_gate=None, signal_quality_gate_path=None,
-                 lake_root=None, gate_topk=10, gate_lookback=5, gate_threshold=0.5,
-                 gate_start="2015-01-03", gate_end="2026-08-19", **kwargs):
-        super().__init__(**kwargs)
-        self._sq_gate_computed = False
-        if signal_quality_gate is not None:
-            self._sq_gate = signal_quality_gate
-            self._sq_gate_computed = True
-        elif signal_quality_gate_path is not None:
-            import pickle
-            with open(signal_quality_gate_path, "rb") as f:
-                self._sq_gate = pickle.load(f)
-            self._sq_gate_computed = True
-        elif lake_root is not None:
-            self._sq_gate = None
-            self._lake_root = lake_root
-            self._gate_topk = gate_topk
-            self._gate_lookback = gate_lookback
-            self._gate_threshold = gate_threshold
-            self._gate_start = gate_start
-            self._gate_end = gate_end
-        else:
-            self._sq_gate = None
-
-    def _compute_gate_on_fly(self):
-        """Compute gate from the signal (pred.pkl) and lake close prices."""
-        import pickle
-        from pathlib import Path
-
-        signal_path = self._signal
-        if not Path(signal_path).exists():
-            return
-
-        with open(signal_path, "rb") as f:
-            pred = pickle.load(f)
-
-        # Handle MultiIndex DataFrame -> unstack to wide
-        if isinstance(pred, pd.DataFrame) and isinstance(pred.index, pd.MultiIndex):
-            pred = pred.iloc[:, 0]
-            pred.index = pd.MultiIndex.from_arrays([
-                pd.to_datetime(pred.index.get_level_values(0)).normalize(),
-                pred.index.get_level_values(1)
-            ])
-            pred = pred.unstack(level=1)
-        elif isinstance(pred, pd.DataFrame):
-            pred = pred.iloc[:, 0] if pred.shape[1] >= 1 else pred.squeeze()
-            pred.index = pd.to_datetime(pred.index).normalize()
-
-        # Load close prices from lake
-        close_df = _load_close_prices(self._lake_root, "US", self._gate_start, self._gate_end)
-        if close_df.empty:
-            return
-
-        ret_df = close_df.pct_change()
-        ret_df.index = pd.to_datetime(ret_df.index).normalize()
-
-        pred_dates = sorted(pred.index.unique())
-        if len(pred_dates) < 2:
-            self._sq_gate = pd.Series(True, index=pd.DatetimeIndex(pred_dates))
-            self._sq_gate_computed = True
-            return
-
-        hit_rates = {}
-        for i in range(1, len(pred_dates)):
-            day = pred_dates[i]
-            prev_day = pred_dates[i - 1]
-            try:
-                prev_scores = pred.loc[prev_day]
-            except KeyError:
-                continue
-            if isinstance(prev_scores, pd.DataFrame):
-                prev_scores = prev_scores.iloc[:, 0]
-            prev_scores = prev_scores.dropna().sort_values(ascending=False)
-            topk_syms = list(prev_scores.index[:self._gate_topk])
-
-            if day not in ret_df.index:
-                continue
-            today_ret = ret_df.loc[day]
-            topk_rets = today_ret.reindex(topk_syms).dropna()
-            if len(topk_rets) == 0:
-                continue
-
-            hit_rates[day] = (topk_rets > 0).sum() / len(topk_rets)
-
-        if not hit_rates:
-            self._sq_gate_computed = True
-            return
-
-        hr_series = pd.Series(hit_rates).sort_index()
-        rolling_hr = hr_series.rolling(self._gate_lookback, min_periods=1).mean()
-        gate = rolling_hr >= self._gate_threshold
-        gate.iloc[:self._gate_lookback] = True
-
-        self._sq_gate = gate
-        self._sq_gate_computed = True
-
-    def _gate_open(self, trade_start_time) -> bool:
-        if self._sq_gate is None:
-            return True
-        ts = pd.Timestamp(trade_start_time)
-        known = self._sq_gate[self._sq_gate.index <= ts]
-        if len(known):
-            return bool(known.iloc[-1])
-        return True  # default open if no history yet
-
-    def generate_trade_decision(self, execute_result=None):
-        if not self._sq_gate_computed:
-            self._compute_gate_on_fly()
-        trade_step = self.trade_calendar.get_trade_step()
-        trade_start_time, _ = self.trade_calendar.get_step_time(trade_step)
-        if not self._gate_open(trade_start_time):
-            return TradeDecisionWO([], self)
-        return super().generate_trade_decision(execute_result)
-
-
-# ---------------------------------------------------------------------------
-# Precomputation helper
-# ---------------------------------------------------------------------------
-
-def compute_signal_quality_gate(
-    pred_path: str,
-    *,
-    lake_root: str = "",
-    market: str = "US",
-    topk: int = 10,
-    lookback: int = 5,
-    threshold: float = 0.5,
-    start: str = "2015-01-03",
-    end: str = "2026-08-19",
-) -> pd.Series:
-    """Build a per-date signal-quality gate series from a pred.pkl and lake bars.
-
-    For each day, checks whether the model's topk picks from the previous day
-    had positive returns.  Computes a rolling hit rate over ``lookback`` days
-    and opens the gate when hit rate >= ``threshold``.
-
-    Parameters
-    ----------
-    pred_path : str — path to pred.pkl (from rd_train / rd_predict).
-    lake_root, market : str — lake location (for loading close prices).
-    topk : int — number of top picks to track for hit rate.
-    lookback : int — rolling window for hit rate computation.
-    threshold : float — hit rate threshold to keep trading.
-    start, end : str — date window for loading prices.
-
-    Returns
-    -------
-    pd.Series — bool, indexed by datetime.  True = trade allowed.
-    """
-    from pathlib import Path
-    import pickle
-
-    # Load pred.pkl
-    with open(pred_path, "rb") as f:
-        pred = pickle.load(f)
-
-    # Handle MultiIndex DataFrame (datetime, instrument) -> unstack to wide
-    if isinstance(pred, pd.DataFrame) and isinstance(pred.index, pd.MultiIndex):
-        pred = pred.iloc[:, 0]  # take score column as Series
-        pred.index = pd.MultiIndex.from_arrays([
-            pd.to_datetime(pred.index.get_level_values(0)).normalize(),
-            pred.index.get_level_values(1)
-        ])
-        # Unstack to wide: dates x instruments
-        pred = pred.unstack(level=1)
-    elif isinstance(pred, pd.DataFrame):
-        pred = pred.iloc[:, 0] if pred.shape[1] >= 1 else pred.squeeze()
-        pred.index = pd.to_datetime(pred.index).normalize()
-
-    # Load close prices from lake
-    close_df = _load_close_prices(lake_root, market, start, end)
-    if close_df.empty:
-        return pd.Series(dtype=bool)
-
-    ret_df = close_df.pct_change()
-    ret_df.index = pd.to_datetime(ret_df.index).normalize()
-
-    # Get sorted unique prediction dates
-    pred_dates = sorted(pred.index.unique())
-    if len(pred_dates) < 2:
-        return pd.Series(True, index=pd.DatetimeIndex(pred_dates))
-
-    # Compute hit rates
-    hit_rates = {}
-    for i in range(1, len(pred_dates)):
-        day = pred_dates[i]
-        # Get yesterday's topk
-        prev_day = pred_dates[i - 1]
-        try:
-            prev_scores = pred.loc[prev_day]
-        except KeyError:
-            continue
-        if isinstance(prev_scores, pd.DataFrame):
-            prev_scores = prev_scores.iloc[:, 0]
-        prev_scores = prev_scores.dropna().sort_values(ascending=False)
-        topk_syms = list(prev_scores.index[:topk])
-
-        # Get today's returns
-        if day not in ret_df.index:
-            continue
-        today_ret = ret_df.loc[day]
-        topk_rets = today_ret.reindex(topk_syms).dropna()
-        if len(topk_rets) == 0:
-            continue
-
-        hit_rates[day] = (topk_rets > 0).sum() / len(topk_rets)
-
-    if not hit_rates:
-        return pd.Series(dtype=bool)
-
-    hr_series = pd.Series(hit_rates).sort_index()
-
-    # Rolling hit rate
-    rolling_hr = hr_series.rolling(lookback, min_periods=1).mean()
-
-    # Gate is open when rolling hit rate >= threshold
-    gate = rolling_hr >= threshold
-    gate.iloc[:lookback] = True  # warmup: allow trading
-
-    return gate
-
-
-def _load_close_prices(lake_root, market, start, end):
-    """Load daily close prices for all symbols into a wide DataFrame."""
-    from pathlib import Path
-
-    lake = Path(lake_root)
-    symbols_parquet = lake / "symbols.parquet"
-    if not symbols_parquet.exists():
-        return pd.DataFrame()
-
-    df = pd.read_parquet(symbols_parquet)
-    col = "symbol" if "symbol" in df.columns else df.columns[0]
-    symbols = sorted(df[col].astype(str).str.upper().tolist())
-
-    closes = {}
-    for sym in symbols:
-        p = lake / "market=US" / "timeframe=1d" / f"symbol={sym}.parquet"
-        if not p.exists():
-            continue
-        try:
-            bar = pd.read_parquet(p)
-        except Exception:
-            continue
-        if not len(bar):
-            continue
-        tcol = "t" if "t" in bar.columns else "date"
-        ts = pd.to_datetime(bar[tcol])
-        bar = bar.assign(_t=ts).set_index("_t").sort_index()
-        bar = bar.loc[start:end]
-        if len(bar) < 10:
-            continue
-        closes[sym] = bar["c"] if "c" in bar.columns else bar["close"]
-
-    return pd.DataFrame(closes)
@@ -1,202 +0,0 @@
-"""Weekly-rebalance TopkDropout strategy.
-
-Turnover-reduction variant of ``qlib.contrib.strategy.signal_strategy.TopkDropoutStrategy``:
-the topk/n_drop selection and sizing are identical to the reference, but the
-target book is recomputed only on the first trading day of each ISO week; on the
-other days the strategy issues NO orders (holds the book untouched).
-
-The weekly cadence is derived from the qlib trade calendar: a rebalance happens
-when the current trade step's date belongs to a different ISO ``(year, week)``
-than the previous trade step. ``hold_band_pct`` (default 0) optionally skips
-tiny rebalances: when a name's existing position differs from the new target by
-less than this fraction, no order is generated for it.
-"""
-
-from __future__ import annotations
-
-from typing import List
-
-import numpy as np
-import pandas as pd
-
-from qlib.backtest import Order
-from qlib.backtest.decision import OrderDir, TradeDecisionWO
-from qlib.contrib.strategy.signal_strategy import TopkDropoutStrategy
-
-__all__ = ["WeeklyRebalanceDropoutStrategy"]
-
-DEFAULT_HOLD_BAND_PCT = 0.0
-
-
-class WeeklyRebalanceDropoutStrategy(TopkDropoutStrategy):
-    """TopkDropout rebalanced once per ISO week; holds otherwise.
-
-    Parameters
-    ----------
-    topk, n_drop, method_sell, method_buy, hold_thresh, only_tradable,
-    forbid_all_trade_at_limit : same as ``TopkDropoutStrategy``.
-    hold_band_pct : skip order for a name whose deviation from target weight is
-        below this fraction of the target (no-trade buffer band).
-    """
-
-    def __init__(self, *, topk, n_drop, hold_band_pct: float = DEFAULT_HOLD_BAND_PCT, **kwargs):
-        super().__init__(topk=topk, n_drop=n_drop, **kwargs)
-        self.hold_band_pct = hold_band_pct
-
-    @staticmethod
-    def _iso_week(ts) -> tuple:
-        return (ts.year, ts.week)
-
-    def generate_trade_decision(self, execute_result=None):
-        import copy
-
-        trade_step = self.trade_calendar.get_trade_step()
-        trade_start_time, trade_end_time = self.trade_calendar.get_step_time(trade_step)
-
-        cur_week = self._iso_week(trade_start_time)
-        prev_week = getattr(self, "_last_week", None)
-        self._last_week = cur_week
-
-        if prev_week is not None and prev_week == cur_week:
-            # not the first trading day of this ISO week -> hold
-            return TradeDecisionWO([], self)
-
-        pred_start_time, pred_end_time = self.trade_calendar.get_step_time(trade_step, shift=1)
-        pred_score = self.signal.get_signal(start_time=pred_start_time, end_time=pred_end_time)
-        if isinstance(pred_score, pd.DataFrame):
-            pred_score = pred_score.iloc[:, 0]
-        if pred_score is None:
-            return TradeDecisionWO([], self)
-
-        if self.only_tradable:
-
-            def get_first_n(li, n, reverse=False):
-                cur_n = 0
-                res = []
-                for si in reversed(li) if reverse else li:
-                    if self.trade_exchange.is_stock_tradable(
-                        stock_id=si, start_time=trade_start_time, end_time=trade_end_time
-                    ):
-                        res.append(si)
-                        cur_n += 1
-                        if cur_n >= n:
-                            break
-                return res[::-1] if reverse else res
-
-            def get_last_n(li, n):
-                return get_first_n(li, n, reverse=True)
-
-            def filter_stock(li):
-                return [
-                    si
-                    for si in li
-                    if self.trade_exchange.is_stock_tradable(
-                        stock_id=si, start_time=trade_start_time, end_time=trade_end_time
-                    )
-                ]
-
-        else:
-
-            def get_first_n(li, n):
-                return list(li)[:n]
-
-            def get_last_n(li, n):
-                return list(li)[-n:]
-
-            def filter_stock(li):
-                return li
-
-        current_temp: "object" = copy.deepcopy(self.trade_position)
-        sell_order_list: List[Order] = []
-        buy_order_list: List[Order] = []
-        cash = current_temp.get_cash()
-        current_stock_list = current_temp.get_stock_list()
-        last = pred_score.reindex(current_stock_list).sort_values(ascending=False).index
-
-        if self.method_buy == "top":
-            today = get_first_n(
-                pred_score[~pred_score.index.isin(last)].sort_values(ascending=False).index,
-                self.n_drop + self.topk - len(last),
-            )
-        elif self.method_buy == "random":
-            topk_candi = get_first_n(pred_score.sort_values(ascending=False).index, self.topk)
-            candi = list(filter(lambda x: x not in last, topk_candi))
-            n = self.n_drop + self.topk - len(last)
-            try:
-                today = np.random.choice(candi, n, replace=False)
-            except ValueError:
-                today = candi
-        else:
-            raise NotImplementedError(f"This type of input is not supported")
-
-        comb = pred_score.reindex(last.union(pd.Index(today))).sort_values(ascending=False).index
-
-        if self.method_sell == "bottom":
-            sell = last[last.isin(get_last_n(comb, self.n_drop))]
-        elif self.method_sell == "random":
-            candi = filter_stock(last)
-            try:
-                sell = pd.Index(np.random.choice(candi, self.n_drop, replace=False) if len(last) else [])
-            except ValueError:
-                sell = candi
-        else:
-            raise NotImplementedError(f"This type of input is not supported")
-
-        buy = today[: len(sell) + self.topk - len(last)]
-        for code in current_stock_list:
-            if not self.trade_exchange.is_stock_tradable(
-                stock_id=code,
-                start_time=trade_start_time,
-                end_time=trade_end_time,
-                direction=None if self.forbid_all_trade_at_limit else OrderDir.SELL,
-            ):
-                continue
-            if code in sell:
-                time_per_step = self.trade_calendar.get_freq()
-                if current_temp.get_stock_count(code, bar=time_per_step) < self.hold_thresh:
-                    continue
-                sell_amount = current_temp.get_stock_amount(code=code)
-                sell_order = Order(
-                    stock_id=code,
-                    amount=sell_amount,
-                    start_time=trade_start_time,
-                    end_time=trade_end_time,
-                    direction=Order.SELL,
-                )
-                if self.trade_exchange.check_order(sell_order):
-                    sell_order_list.append(sell_order)
-                    trade_val, trade_cost, trade_price = self.trade_exchange.deal_order(
-                        sell_order, position=current_temp
-                    )
-                    cash += trade_val - trade_cost
-
-        if len(buy) == 0:
-            return TradeDecisionWO(sell_order_list, self)
-
-        value = cash * self.risk_degree / len(buy)
-        for code in buy:
-            if not self.trade_exchange.is_stock_tradable(
-                stock_id=code,
-                start_time=trade_start_time,
-                end_time=trade_end_time,
-                direction=None if self.forbid_all_trade_at_limit else OrderDir.BUY,
-            ):
-                continue
-            buy_price = self.trade_exchange.get_deal_price(
-                stock_id=code, start_time=trade_start_time, end_time=trade_end_time, direction=OrderDir.BUY
-            )
-            buy_amount = value / buy_price
-            factor = self.trade_exchange.get_factor(
-                stock_id=code, start_time=trade_start_time, end_time=trade_end_time
-            )
-            buy_amount = self.trade_exchange.round_amount_by_trade_unit(buy_amount, factor)
-            buy_order = Order(
-                stock_id=code,
-                amount=buy_amount,
-                start_time=trade_start_time,
-                end_time=trade_end_time,
-                direction=Order.BUY,
-            )
-            buy_order_list.append(buy_order)
-
-        return TradeDecisionWO(sell_order_list + buy_order_list, self)
@@ -6,11 +6,10 @@ The lake is a hive-partitioned parquet store (see ``tac-engine/skills/tradeac-la
    ├── market=US/
    │   └── timeframe=1d/
    │       └── symbol=AAPL.parquet        # OHLCV bars: t, date, o, h, l, c, v, n, vw
-    ├── features/                          # indicators, wide format, family tier
+    ├── features/                          # ta-lib indicators, wide format
    │   └── market=US/
    │       └── timeframe=1d/
-    │           ├── family=ta/symbol=AAPL.parquet  # t, sma_5, sma_20, rsi_14, ...
-    │           └── family=sp/symbol=AAPL.parquet  # t, sp_ou_*, sp_hmm_*, ...
+    │           └── symbol=AAPL.parquet    # t, sma_5, sma_20, rsi_14, ...
    ├── calendar.parquet                   # trading days per market
    ├── coverage.parquet                   # per (market,timeframe,symbol) loaded windows
    └── symbols.parquet                    # asset master
@@ -108,34 +107,8 @@ class LakeConfig:
        return self.lake_root / "features" / f"market={self.market}" / f"timeframe={timeframe}"

    def features_path(self, timeframe: str, symbol: str) -> Path:
-        # Legacy flat path (no family tier). Prefer `load_features` which
-        # resolves the family=ta|sp partition layout.
        return self.features_dir(timeframe) / f"symbol={str(symbol).upper()}.parquet"

-    def load_features(self, timeframe: str, symbol: str) -> pd.DataFrame:
-        """All feature columns for a symbol, merging the `family=ta` and
-        `family=sp` partitions by timestamp. Returns an empty frame when no
-        feature files exist (legacy flat layout falls back transparently)."""
-        sym = str(symbol).upper()
-        frames = []
-        for family in ("ta", "sp"):
-            p = self.features_dir(timeframe) / f"family={family}" / f"symbol={sym}.parquet"
-            if p.exists():
-                frames.append(pd.read_parquet(p))
-        if not frames:
-            flat = self.features_dir(timeframe) / f"symbol={sym}.parquet"
-            if flat.exists():
-                return pd.read_parquet(flat)
-            return pd.DataFrame()
-        if len(frames) == 1:
-            return frames[0]
-        merged = frames[0]
-        for extra in frames[1:]:
-            merged = merged.merge(extra, on="t", how="outer", suffixes=("", "_dup"))
-            for c in [c for c in merged.columns if c.endswith("_dup")]:
-                merged = merged.drop(columns=c)
-        return merged
-
    def calendar_path(self) -> Path:
        return self.lake_root / "calendar.parquet"

@@ -173,7 +173,8 @@ class LakeFeatureProvider(FeatureProvider):
    def _load_feature_df(self, instrument: str, timeframe: str) -> pd.DataFrame:
        key = (instrument, timeframe)
        if key not in self._feature_cache:
-            self._feature_cache[key] = self.cfg.load_features(timeframe, instrument)
+            p = self.cfg.features_path(timeframe, instrument)
+            self._feature_cache[key] = pd.read_parquet(p) if p.exists() else pd.DataFrame()
        return self._feature_cache[key]

    @staticmethod
@@ -1,16 +1,16 @@
 # -----------------------------------------------------------------------------
-# Signal-quality gate: TopkDropout gated by rolling hit-rate of topk picks.
+# ABLATION A (baseline): LightGBM with RankIC early-stopping on the 50-ETF SP-5d
+# panel, using ALL 24 sp_* feature columns (ou,hmm,jump,har,trend,hurst,
+# signature). Copy of the canonical workflow_lgb_sp5d_rankic.yaml with a
+# distinct experiment name so the ablation runs are isolated.
 #
-# 1. Compute the gate: python precompute_signal_quality_gate.py <pred.pkl> <gate.pkl>
-# 2. Run this workflow: rd_run_workflow config_path=<this yaml> experiment_name=<exp>
-#
-# The strategy loads the precomputed gate from signal_quality_gate_path.
-# When hit rate >= threshold, trade; otherwise, go to cash.
+# Run:
+#   rd_run_workflow config_path=tac-qlib/workflows/ablate_baseline_all_sp_fields.yaml \
+#       experiment_name=tac-rd-rank-ablate
 # -----------------------------------------------------------------------------
 {%- set LAKE = TAC_LAKE_DIR %}
 {%- set UNIVERSE = "SPY,QQQ,DIA,IWM,MDY,VTI,VOO,VEA,VWO,VT,EFA,EEM,TLT,IEF,SHY,AGG,BND,LQD,HYG,JNK,EMB,GLD,SLV,USO,UNG,DBA,DBC,XLK,XLF,XLE,XLV,XLI,XLY,XLP,XLU,XLB,XLRE,ARKK,SMH,SOXX,IBB,XBI,ITA,XAR,ICLN,TAN,FDN,IGV,ESPO,REM" %}
 {%- set SP_FIELDS = "sp_ret,sp_ou_zscore,sp_ou_half_life,sp_ou_revert,sp_hmm_p_regime1,sp_hmm_state,sp_jump_ratio,sp_jump_flag,sp_jump_tail,sp_max_move,sp_rv1,sp_rv5,sp_rv22,sp_vol_ratio_5_22,sp_vol_ratio_1_22,sp_trend_slope_5,sp_trend_slope_20,sp_trend_slope_60,sp_logp,sp_hurst_exponent,sp_sig_level1_lead,sp_sig_level1_lag,sp_sig_level2_lead_lag,sp_sig_level2_lag_lead" %}
-{%- set GATE_PATH = "/app/experiments/book/data/signal_quality_gate/sq_gate_5d_0.50.pkl" %}

 qlib_init:
    provider_uri: "{{ LAKE }}"
@@ -40,22 +40,27 @@ qlib_init:
        module_path: qlib.workflow.expm
        kwargs:
            uri: "sqlite:///{{ LAKE }}/mlruns.db"
-            default_exp_name: "tac-rd-sq-gate"
+            default_exp_name: "tac-rd-rank-ablate"

 task:
    model:
-        class: LGBModel
-        module_path: qlib.contrib.model.gbdt
+        class: RankICLGBModel
+        module_path: tac_qlib.contrib.model.rank_gbdt
        kwargs:
            loss: mse
-            learning_rate: 0.05
-            num_leaves: 15
-            n_estimators: 200
+            learning_rate: 0.02
+            num_leaves: 31
+            n_estimators: 3000
+            num_boost_round: 3000
+            early_stopping_rounds: 200
+            min_data_in_leaf: 20
+            lambda_l2: 0.5
            colsample_bytree: 0.8
            subsample: 0.8
            subsample_freq: 1
-            reg_alpha: 0.01
-            reg_lambda: 0.01
+            reg_alpha: 0.1
+            reg_lambda: 1.0
+            seed: 42

    dataset:
        class: DatasetH
@@ -105,13 +110,12 @@ task:
          kwargs:
              config:
                  strategy:
-                      class: SignalQualityGateStrategy
-                      module_path: tac_qlib.contrib.strategy.signal_quality_gate
+                      class: TopkDropoutStrategy
+                      module_path: qlib.contrib.strategy
                      kwargs:
                          signal: "<PRED>"
-                          signal_quality_gate_path: "{{ GATE_PATH }}"
                          topk: 10
-                          n_drop: 1
+                          n_drop: 2
                          only_tradable: true
                          risk_degree: 0.95
                  backtest:
@@ -1,44 +1,67 @@
-# Signal-quality gate backtest for 2026
+# -----------------------------------------------------------------------------
+# ABLATION B (generic-only): same panel/model as the baseline, but feature
+# fields restricted to the model-free / generic stochastic-process families
+# (jump,har,trend,hurst,signature). Drops the model-specific ou (OU/AR-1
+# half-life) and hmm (2-state regime) families to test whether the generic
+# families alone dominate the rank dimension.
+#
+# Run:
+#   rd_run_workflow config_path=tac-qlib/workflows/ablate_generic_only_sp_fields.yaml \
+#       experiment_name=tac-rd-rank-ablate
+# -----------------------------------------------------------------------------
 {%- set LAKE = TAC_LAKE_DIR %}
 {%- set UNIVERSE = "SPY,QQQ,DIA,IWM,MDY,VTI,VOO,VEA,VWO,VT,EFA,EEM,TLT,IEF,SHY,AGG,BND,LQD,HYG,JNK,EMB,GLD,SLV,USO,UNG,DBA,DBC,XLK,XLF,XLE,XLV,XLI,XLY,XLP,XLU,XLB,XLRE,ARKK,SMH,SOXX,IBB,XBI,ITA,XAR,ICLN,TAN,FDN,IGV,ESPO,REM" %}
-{%- set SP_FIELDS = "sp_ret,sp_ou_zscore,sp_ou_half_life,sp_ou_revert,sp_hmm_p_regime1,sp_hmm_state,sp_jump_ratio,sp_jump_flag,sp_jump_tail,sp_max_move,sp_rv1,sp_rv5,sp_rv22,sp_vol_ratio_5_22,sp_vol_ratio_1_22,sp_trend_slope_5,sp_trend_slope_20,sp_trend_slope_60,sp_logp,sp_hurst_exponent,sp_sig_level1_lead,sp_sig_level1_lag,sp_sig_level2_lead_lag,sp_sig_level2_lag_lead" %}
-{# Gate computed on-the-fly from signal + lake close prices #}
+{%- set SP_FIELDS = "sp_ret,sp_jump_ratio,sp_jump_flag,sp_jump_tail,sp_max_move,sp_rv1,sp_rv5,sp_rv22,sp_vol_ratio_5_22,sp_vol_ratio_1_22,sp_trend_slope_5,sp_trend_slope_20,sp_trend_slope_60,sp_logp,sp_hurst_exponent,sp_sig_level1_lead,sp_sig_level1_lag,sp_sig_level2_lead_lag,sp_sig_level2_lag_lead" %}

 qlib_init:
    provider_uri: "{{ LAKE }}"
    region: us
    expression_cache: null
    dataset_cache: null
+
    calendar_provider:
        class: tac_qlib.data.providers.LakeCalendarProvider
-        kwargs: { lake_root: "{{ LAKE }}", market: US }
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
    instrument_provider:
        class: tac_qlib.data.providers.LakeInstrumentProvider
-        kwargs: { lake_root: "{{ LAKE }}", market: US, markets: {} }
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+            markets: {}
    feature_provider:
        class: tac_qlib.data.providers.LakeFeatureProvider
-        kwargs: { lake_root: "{{ LAKE }}", market: US }
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+
    exp_manager:
        class: MLflowExpManager
        module_path: qlib.workflow.expm
        kwargs:
            uri: "sqlite:///{{ LAKE }}/mlruns.db"
-            default_exp_name: "tac-rd-sq-gate-2026"
+            default_exp_name: "tac-rd-rank-ablate"

 task:
    model:
-        class: LGBModel
-        module_path: qlib.contrib.model.gbdt
+        class: RankICLGBModel
+        module_path: tac_qlib.contrib.model.rank_gbdt
        kwargs:
            loss: mse
-            learning_rate: 0.05
-            num_leaves: 15
-            n_estimators: 200
+            learning_rate: 0.02
+            num_leaves: 31
+            n_estimators: 3000
+            num_boost_round: 3000
+            early_stopping_rounds: 200
+            min_data_in_leaf: 20
+            lambda_l2: 0.5
            colsample_bytree: 0.8
            subsample: 0.8
            subsample_freq: 1
-            reg_alpha: 0.01
-            reg_lambda: 0.01
+            reg_alpha: 0.1
+            reg_lambda: 1.0
+            seed: 42

    dataset:
        class: DatasetH
@@ -52,7 +75,7 @@ task:
                    start_time: 2015-01-03
                    end_time: 2026-08-10
                    fit_start_time: 2015-01-03
-                    fit_end_time: 2026-01-03
+                    fit_end_time: 2025-09-01
                    freq: day
                    lake_root: "{{ LAKE }}"
                    market: US
@@ -80,24 +103,20 @@ task:
          kwargs: {}
        - class: SigAnaRecord
          module_path: qlib.workflow.record_temp
-          kwargs: { ana_long_short: true, ann_scaler: 252 }
+          kwargs:
+              ana_long_short: true
+              ann_scaler: 252
        - class: PortAnaRecord
          module_path: qlib.workflow.record_temp
          kwargs:
              config:
                  strategy:
-                      class: SignalQualityGateStrategy
-                      module_path: tac_qlib.contrib.strategy.signal_quality_gate
+                      class: TopkDropoutStrategy
+                      module_path: qlib.contrib.strategy
                      kwargs:
                          signal: "<PRED>"
-                          lake_root: "{{ LAKE }}"
-                          gate_topk: 10
-                          gate_lookback: 5
-                          gate_threshold: 0.5
-                          gate_start: "2015-01-03"
-                          gate_end: "2026-08-19"
                          topk: 10
-                          n_drop: 1
+                          n_drop: 2
                          only_tradable: true
                          risk_degree: 0.95
                  backtest:
@@ -0,0 +1,141 @@
+# -----------------------------------------------------------------------------
+# ISOLATION: multi-seed RankIC ensemble, ablate-B generic-only feature set.
+#
+# Isolates the ensemble effect on the SP-5d rank signal. Same panel, segments,
+# history (full backfilled 2016+) and feature set as the exp-9 ablate-B winner
+# (generic-only sp_* families: jump,har,trend,hurst,signature), but replaces the
+# single RankICLGBModel with a 5-seed RankICEnsembleLGBModel (42,7,2026,99,123)
+# that averages per-day predictions.
+#
+# Differs from exp-15 (tac-rd-rank-ensemble, mlflow exp 15) ONLY by dropping the
+# TA subset (rsi_14,roc_10,macd_hist,willr_14,atr_14) and the inter-asset xr_*
+# features, so any change vs exp-15 is attributable to the feature set alone,
+# and any change vs exp-9 is attributable to the ensemble + full history alone.
+#
+# Run:
+#   rd_run_workflow config_path=experiments/workflows/exp12_isolation_ensemble.yaml \
+#       experiment_name=tac-rd-rank-ensemble-isolated
+# -----------------------------------------------------------------------------
+{%- set LAKE = TAC_LAKE_DIR %}
+{%- set UNIVERSE = "SPY,QQQ,DIA,IWM,MDY,VTI,VOO,VEA,VWO,VT,EFA,EEM,TLT,IEF,SHY,AGG,BND,LQD,HYG,JNK,EMB,GLD,SLV,USO,UNG,DBA,DBC,XLK,XLF,XLE,XLV,XLI,XLY,XLP,XLU,XLB,XLRE,ARKK,SMH,SOXX,IBB,XBI,ITA,XAR,ICLN,TAN,FDN,IGV,ESPO,REM" %}
+{%- set SP_FIELDS = "sp_ret,sp_jump_ratio,sp_jump_flag,sp_jump_tail,sp_max_move,sp_rv1,sp_rv5,sp_rv22,sp_vol_ratio_5_22,sp_vol_ratio_1_22,sp_trend_slope_5,sp_trend_slope_20,sp_trend_slope_60,sp_logp,sp_hurst_exponent,sp_sig_level1_lead,sp_sig_level1_lag,sp_sig_level2_lead_lag,sp_sig_level2_lag_lead" %}
+
+qlib_init:
+    provider_uri: "{{ LAKE }}"
+    region: us
+    expression_cache: null
+    dataset_cache: null
+
+    calendar_provider:
+        class: tac_qlib.data.providers.LakeCalendarProvider
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+    instrument_provider:
+        class: tac_qlib.data.providers.LakeInstrumentProvider
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+            markets: {}
+    feature_provider:
+        class: tac_qlib.data.providers.LakeFeatureProvider
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+
+    exp_manager:
+        class: MLflowExpManager
+        module_path: qlib.workflow.expm
+        kwargs:
+            uri: "sqlite:///mlruns.db"
+            default_exp_name: "tac-rd-rank-ensemble-isolated"
+
+task:
+    model:
+        class: RankICEnsembleLGBModel
+        module_path: tac_qlib.contrib.model.rank_ensemble
+        kwargs:
+            loss: mse
+            learning_rate: 0.02
+            num_leaves: 31
+            n_estimators: 3000
+            num_boost_round: 3000
+            early_stopping_rounds: 200
+            min_data_in_leaf: 20
+            lambda_l2: 0.5
+            colsample_bytree: 0.8
+            subsample: 0.8
+            subsample_freq: 1
+            reg_alpha: 0.1
+            reg_lambda: 1.0
+            seeds: "42,7,2026,99,123"
+
+    dataset:
+        class: DatasetH
+        module_path: qlib.data.dataset
+        kwargs:
+            handler:
+                class: TACHandler
+                module_path: tac_qlib.contrib.data.handler
+                kwargs:
+                    instruments: "{{ UNIVERSE }}"
+                    start_time: 2015-01-03
+                    end_time: 2026-08-14
+                    fit_start_time: 2016-01-04
+                    fit_end_time: 2025-09-01
+                    freq: day
+                    lake_root: "{{ LAKE }}"
+                    market: US
+                    label: "Ref($close,-6)/Ref($close,-1)-1"
+                    feature_fields: "$open,$high,$low,$close,$vwap,$volume,{{ SP_FIELDS }}"
+                    infer_processors:
+                        - class: DropAllNaN
+                          kwargs: {}
+                        - class: ProcessInf
+                          kwargs: {}
+                        - class: CSRankNorm
+                          kwargs: {}
+                        - class: ZScoreNorm
+                          kwargs: {}
+                        - class: Fillna
+                          kwargs: {}
+            segments:
+                train: [2016-01-04, 2025-09-01]
+                valid: [2025-09-03, 2026-01-03]
+                test: [2026-01-04, 2026-08-10]
+
+    record:
+        - class: SignalRecord
+          module_path: qlib.workflow.record_temp
+          kwargs: {}
+        - class: SigAnaRecord
+          module_path: qlib.workflow.record_temp
+          kwargs:
+              ana_long_short: true
+              ann_scaler: 252
+        - class: PortAnaRecord
+          module_path: qlib.workflow.record_temp
+          kwargs:
+              config:
+                  strategy:
+                      class: TopkDropoutStrategy
+                      module_path: qlib.contrib.strategy
+                      kwargs:
+                          signal: "<PRED>"
+                          topk: 10
+                          n_drop: 2
+                          only_tradable: true
+                          risk_degree: 0.95
+                  backtest:
+                      start_time: 2026-01-04
+                      end_time: 2026-08-10
+                      account: 1000000
+                      benchmark: SPY
+                      exchange_kwargs:
+                          codes: "{{ UNIVERSE }}"
+                          deal_price: $close
+                          freq: day
+                          open_cost: 0.0005
+                          close_cost: 0.0015
+                          min_cost: 5.0
+              risk_analysis_freq: 1d
@@ -0,0 +1,141 @@
+# -----------------------------------------------------------------------------
+# EXP 18 - Risk-limit control: reference model + TopkDropout baseline (A).
+#
+# Signal/model identical to the reference (tac-rd-rank-ensemble-isolated,
+# run 0cea66d9...): RankICEnsembleLGBModel (parallel, 5 seeds) on the 50-ETF
+# SP-5d panel, test 2026-01-04..2026-08-10. This workflow reproduces the
+# unconstrained TopkDropout baseline net-of-cost so the risk-limited variant
+# (same pred, liquidity/size/concentration caps) can be compared 1:1.
+#
+# The risk_limits spec itself is applied via rd_backtest / rd_strategy_targets
+# (tool-level param, not a YAML key); this run records the unconstrained
+# baseline that the limit A/B is measured against.
+#
+# Run:
+#   rd_run_workflow config_path=experiments/workflows/exp18-risk-limit/a_baseline.yaml \
+#       experiment_name=tac-rd-risk-limit
+# -----------------------------------------------------------------------------
+{%- set LAKE = TAC_LAKE_DIR %}
+{%- set UNIVERSE = "SPY,QQQ,DIA,IWM,MDY,VTI,VOO,VEA,VWO,VT,EFA,EEM,TLT,IEF,SHY,AGG,BND,LQD,HYG,JNK,EMB,GLD,SLV,USO,UNG,DBA,DBC,XLK,XLF,XLE,XLV,XLI,XLY,XLP,XLU,XLB,XLRE,ARKK,SMH,SOXX,IBB,XBI,ITA,XAR,ICLN,TAN,FDN,IGV,ESPO,REM" %}
+{%- set SP_FIELDS = "sp_ret,sp_jump_ratio,sp_jump_flag,sp_jump_tail,sp_max_move,sp_rv1,sp_rv5,sp_rv22,sp_vol_ratio_5_22,sp_vol_ratio_1_22,sp_trend_slope_5,sp_trend_slope_20,sp_trend_slope_60,sp_logp,sp_hurst_exponent,sp_sig_level1_lead,sp_sig_level1_lag,sp_sig_level2_lead_lag,sp_sig_level2_lag_lead" %}
+
+qlib_init:
+    provider_uri: "{{ LAKE }}"
+    region: us
+    expression_cache: null
+    dataset_cache: null
+
+    calendar_provider:
+        class: tac_qlib.data.providers.LakeCalendarProvider
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+    instrument_provider:
+        class: tac_qlib.data.providers.LakeInstrumentProvider
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+            markets: {}
+    feature_provider:
+        class: tac_qlib.data.providers.LakeFeatureProvider
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+
+    exp_manager:
+        class: MLflowExpManager
+        module_path: qlib.workflow.expm
+        kwargs:
+            uri: "sqlite:///{{ LAKE }}/mlruns.db"
+            default_exp_name: "tac-rd-risk-limit"
+
+task:
+    model:
+        class: RankICEnsembleLGBModel
+        module_path: tac_qlib.contrib.model.rank_ensemble
+        kwargs:
+            loss: mse
+            learning_rate: 0.02
+            num_leaves: 31
+            n_estimators: 3000
+            num_boost_round: 3000
+            early_stopping_rounds: 200
+            min_data_in_leaf: 20
+            lambda_l2: 0.5
+            colsample_bytree: 0.8
+            subsample: 0.8
+            subsample_freq: 1
+            reg_alpha: 0.1
+            reg_lambda: 1.0
+            seeds: "42,7,2026,99,123"
+            parallel: 5
+
+    dataset:
+        class: DatasetH
+        module_path: qlib.data.dataset
+        kwargs:
+            handler:
+                class: TACHandler
+                module_path: tac_qlib.contrib.data.handler
+                kwargs:
+                    instruments: "{{ UNIVERSE }}"
+                    start_time: 2015-01-03
+                    end_time: 2026-08-14
+                    fit_start_time: 2016-01-04
+                    fit_end_time: 2025-09-01
+                    freq: day
+                    lake_root: "{{ LAKE }}"
+                    market: US
+                    label: "Ref($close,-6)/Ref($close,-1)-1"
+                    feature_fields: "$open,$high,$low,$close,$vwap,$volume,{{ SP_FIELDS }}"
+                    infer_processors:
+                        - class: DropAllNaN
+                          kwargs: {}
+                        - class: ProcessInf
+                          kwargs: {}
+                        - class: CSRankNorm
+                          kwargs: {}
+                        - class: ZScoreNorm
+                          kwargs: {}
+                        - class: Fillna
+                          kwargs: {}
+            segments:
+                train: [2016-01-04, 2025-09-01]
+                valid: [2025-09-03, 2026-01-03]
+                test: [2026-01-04, 2026-08-10]
+
+    record:
+        - class: SignalRecord
+          module_path: qlib.workflow.record_temp
+          kwargs: {}
+        - class: SigAnaRecord
+          module_path: qlib.workflow.record_temp
+          kwargs:
+              ana_long_short: true
+              ann_scaler: 252
+        - class: PortAnaRecord
+          module_path: qlib.workflow.record_temp
+          kwargs:
+              config:
+                  strategy:
+                      class: TopkDropoutStrategy
+                      module_path: qlib.contrib.strategy
+                      kwargs:
+                          signal: "<PRED>"
+                          topk: 10
+                          n_drop: 2
+                          only_tradable: true
+                          risk_degree: 0.95
+                  backtest:
+                      start_time: 2026-01-04
+                      end_time: 2026-08-10
+                      account: 1000000
+                      benchmark: SPY
+                      exchange_kwargs:
+                          codes: "{{ UNIVERSE }}"
+                          deal_price: $close
+                          freq: day
+                          open_cost: 0.0005
+                          close_cost: 0.0015
+                          min_cost: 5.0
+              risk_analysis_freq: 1d
@@ -1,44 +1,68 @@
-# Signal-quality gate backtest for 2023
+# -----------------------------------------------------------------------------
+# EXP 20 - R1: 2-seed ensemble (seeds 42,7), TopkDropout baseline.
+#
+# Runtime cut: 2 seeds instead of 5. Everything else identical to the reference
+# (test 2026-01-04..2026-08-10, SPY, costs 5bp/15bp). Measures whether the
+# 2-seed ensemble keeps the reference quality at ~2/5 the training time.
+#
+# Run:
+#   rd_run_workflow config_path=experiments/workflows/exp20-risk-limit-improve/r1_2seed.yaml \
+#       experiment_name=tac-rd-risk-limit
+# -----------------------------------------------------------------------------
 {%- set LAKE = TAC_LAKE_DIR %}
 {%- set UNIVERSE = "SPY,QQQ,DIA,IWM,MDY,VTI,VOO,VEA,VWO,VT,EFA,EEM,TLT,IEF,SHY,AGG,BND,LQD,HYG,JNK,EMB,GLD,SLV,USO,UNG,DBA,DBC,XLK,XLF,XLE,XLV,XLI,XLY,XLP,XLU,XLB,XLRE,ARKK,SMH,SOXX,IBB,XBI,ITA,XAR,ICLN,TAN,FDN,IGV,ESPO,REM" %}
-{%- set SP_FIELDS = "sp_ret,sp_ou_zscore,sp_ou_half_life,sp_ou_revert,sp_hmm_p_regime1,sp_hmm_state,sp_jump_ratio,sp_jump_flag,sp_jump_tail,sp_max_move,sp_rv1,sp_rv5,sp_rv22,sp_vol_ratio_5_22,sp_vol_ratio_1_22,sp_trend_slope_5,sp_trend_slope_20,sp_trend_slope_60,sp_logp,sp_hurst_exponent,sp_sig_level1_lead,sp_sig_level1_lag,sp_sig_level2_lead_lag,sp_sig_level2_lag_lead" %}
-{# Gate computed on-the-fly from signal + lake close prices #}
+{%- set SP_FIELDS = "sp_ret,sp_jump_ratio,sp_jump_flag,sp_jump_tail,sp_max_move,sp_rv1,sp_rv5,sp_rv22,sp_vol_ratio_5_22,sp_vol_ratio_1_22,sp_trend_slope_5,sp_trend_slope_20,sp_trend_slope_60,sp_logp,sp_hurst_exponent,sp_sig_level1_lead,sp_sig_level1_lag,sp_sig_level2_lead_lag,sp_sig_level2_lag_lead" %}

 qlib_init:
    provider_uri: "{{ LAKE }}"
    region: us
    expression_cache: null
    dataset_cache: null
+
    calendar_provider:
        class: tac_qlib.data.providers.LakeCalendarProvider
-        kwargs: { lake_root: "{{ LAKE }}", market: US }
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
    instrument_provider:
        class: tac_qlib.data.providers.LakeInstrumentProvider
-        kwargs: { lake_root: "{{ LAKE }}", market: US, markets: {} }
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+            markets: {}
    feature_provider:
        class: tac_qlib.data.providers.LakeFeatureProvider
-        kwargs: { lake_root: "{{ LAKE }}", market: US }
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+
    exp_manager:
        class: MLflowExpManager
        module_path: qlib.workflow.expm
        kwargs:
            uri: "sqlite:///{{ LAKE }}/mlruns.db"
-            default_exp_name: "tac-rd-sq-gate-2023"
+            default_exp_name: "tac-rd-risk-limit"

 task:
    model:
-        class: LGBModel
-        module_path: qlib.contrib.model.gbdt
+        class: RankICEnsembleLGBModel
+        module_path: tac_qlib.contrib.model.rank_ensemble
        kwargs:
            loss: mse
-            learning_rate: 0.05
-            num_leaves: 15
-            n_estimators: 200
+            learning_rate: 0.02
+            num_leaves: 31
+            n_estimators: 3000
+            num_boost_round: 3000
+            early_stopping_rounds: 200
+            min_data_in_leaf: 20
+            lambda_l2: 0.5
            colsample_bytree: 0.8
            subsample: 0.8
            subsample_freq: 1
-            reg_alpha: 0.01
-            reg_lambda: 0.01
+            reg_alpha: 0.1
+            reg_lambda: 1.0
+            seeds: "42,7,2026,99,123"
+            parallel: 5

    dataset:
        class: DatasetH
@@ -50,9 +74,9 @@ task:
                kwargs:
                    instruments: "{{ UNIVERSE }}"
                    start_time: 2015-01-03
-                    end_time: 2023-12-29
-                    fit_start_time: 2015-01-03
-                    fit_end_time: 2023-01-03
+                    end_time: 2026-08-14
+                    fit_start_time: 2016-01-04
+                    fit_end_time: 2025-09-01
                    freq: day
                    lake_root: "{{ LAKE }}"
                    market: US
@@ -70,9 +94,9 @@ task:
                        - class: Fillna
                          kwargs: {}
            segments:
-                train: [2015-01-03, 2022-09-01]
-                valid: [2022-09-03, 2023-01-03]
-                test: [2023-01-03, 2023-12-29]
+                train: [2016-01-04, 2025-09-01]
+                valid: [2025-09-03, 2026-01-03]
+                test: [2026-01-04, 2026-08-10]

    record:
        - class: SignalRecord
@@ -80,29 +104,25 @@ task:
          kwargs: {}
        - class: SigAnaRecord
          module_path: qlib.workflow.record_temp
-          kwargs: { ana_long_short: true, ann_scaler: 252 }
+          kwargs:
+              ana_long_short: true
+              ann_scaler: 252
        - class: PortAnaRecord
          module_path: qlib.workflow.record_temp
          kwargs:
              config:
                  strategy:
-                      class: SignalQualityGateStrategy
-                      module_path: tac_qlib.contrib.strategy.signal_quality_gate
+                      class: TopkDropoutStrategy
+                      module_path: qlib.contrib.strategy
                      kwargs:
                          signal: "<PRED>"
-                          lake_root: "{{ LAKE }}"
-                          gate_topk: 10
-                          gate_lookback: 5
-                          gate_threshold: 0.5
-                          gate_start: "2015-01-03"
-                          gate_end: "2023-12-29"
                          topk: 10
-                          n_drop: 1
+                          n_drop: 2
                          only_tradable: true
                          risk_degree: 0.95
                  backtest:
-                      start_time: 2023-01-03
-                      end_time: 2023-12-29
+                      start_time: 2026-01-04
+                      end_time: 2026-08-10
                      account: 1000000
                      benchmark: SPY
                      exchange_kwargs:
@@ -0,0 +1,135 @@
+# -----------------------------------------------------------------------------
+# EXP 20 - R1: 2-seed ensemble (seeds 42,7), TopkDropout baseline.
+#
+# Runtime cut: 2 seeds instead of 5. Everything else identical to the reference
+# (test 2026-01-04..2026-08-10, SPY, costs 5bp/15bp). Measures whether the
+# 2-seed ensemble keeps the reference quality at ~2/5 the training time.
+#
+# Run:
+#   rd_run_workflow config_path=experiments/workflows/exp20-risk-limit-improve/r1_2seed.yaml \
+#       experiment_name=tac-rd-risk-limit
+# -----------------------------------------------------------------------------
+{%- set LAKE = TAC_LAKE_DIR %}
+{%- set UNIVERSE = "SPY,QQQ,DIA,IWM,MDY,VTI,VOO,VEA,VWO,VT,EFA,EEM,TLT,IEF,SHY,AGG,BND,LQD,HYG,JNK,EMB,GLD,SLV,USO,UNG,DBA,DBC,XLK,XLF,XLE,XLV,XLI,XLY,XLP,XLU,XLB,XLRE,ARKK,SMH,SOXX,IBB,XBI,ITA,XAR,ICLN,TAN,FDN,IGV,ESPO,REM" %}
+{%- set SP_FIELDS = "sp_ret,sp_jump_ratio,sp_jump_flag,sp_jump_tail,sp_max_move,sp_rv1,sp_rv5,sp_rv22,sp_vol_ratio_5_22,sp_vol_ratio_1_22,sp_trend_slope_5,sp_trend_slope_20,sp_trend_slope_60,sp_logp,sp_hurst_exponent,sp_sig_level1_lead,sp_sig_level1_lag,sp_sig_level2_lead_lag,sp_sig_level2_lag_lead" %}
+
+qlib_init:
+    provider_uri: "{{ LAKE }}"
+    region: us
+    expression_cache: null
+    dataset_cache: null
+
+    calendar_provider:
+        class: tac_qlib.data.providers.LakeCalendarProvider
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+    instrument_provider:
+        class: tac_qlib.data.providers.LakeInstrumentProvider
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+            markets: {}
+    feature_provider:
+        class: tac_qlib.data.providers.LakeFeatureProvider
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+
+    exp_manager:
+        class: MLflowExpManager
+        module_path: qlib.workflow.expm
+        kwargs:
+            uri: "sqlite:///{{ LAKE }}/mlruns.db"
+            default_exp_name: "tac-rd-risk-limit"
+
+task:
+    model:
+        class: RankICEnsembleLGBModel
+        module_path: tac_qlib.contrib.model.rank_ensemble
+        kwargs:
+            loss: mse
+            learning_rate: 0.02
+            num_leaves: 31
+            n_estimators: 3000
+            num_boost_round: 3000
+            early_stopping_rounds: 200
+            min_data_in_leaf: 20
+            lambda_l2: 0.5
+            colsample_bytree: 0.8
+            subsample: 0.8
+            subsample_freq: 1
+            reg_alpha: 0.1
+            reg_lambda: 1.0
+            seeds: "42,7,2026,99,123"
+            parallel: 1
+
+    dataset:
+        class: DatasetH
+        module_path: qlib.data.dataset
+        kwargs:
+            handler:
+                class: TACHandler
+                module_path: tac_qlib.contrib.data.handler
+                kwargs:
+                    instruments: "{{ UNIVERSE }}"
+                    start_time: 2015-01-03
+                    end_time: 2026-08-14
+                    fit_start_time: 2016-01-04
+                    fit_end_time: 2025-09-01
+                    freq: day
+                    lake_root: "{{ LAKE }}"
+                    market: US
+                    label: "Ref($close,-6)/Ref($close,-1)-1"
+                    feature_fields: "$open,$high,$low,$close,$vwap,$volume,{{ SP_FIELDS }}"
+                    infer_processors:
+                        - class: DropAllNaN
+                          kwargs: {}
+                        - class: ProcessInf
+                          kwargs: {}
+                        - class: CSRankNorm
+                          kwargs: {}
+                        - class: ZScoreNorm
+                          kwargs: {}
+                        - class: Fillna
+                          kwargs: {}
+            segments:
+                train: [2016-01-04, 2025-09-01]
+                valid: [2025-09-03, 2026-01-03]
+                test: [2026-01-04, 2026-08-10]
+
+    record:
+        - class: SignalRecord
+          module_path: qlib.workflow.record_temp
+          kwargs: {}
+        - class: SigAnaRecord
+          module_path: qlib.workflow.record_temp
+          kwargs:
+              ana_long_short: true
+              ann_scaler: 252
+        - class: PortAnaRecord
+          module_path: qlib.workflow.record_temp
+          kwargs:
+              config:
+                  strategy:
+                      class: TopkDropoutStrategy
+                      module_path: qlib.contrib.strategy
+                      kwargs:
+                          signal: "<PRED>"
+                          topk: 10
+                          n_drop: 2
+                          only_tradable: true
+                          risk_degree: 0.95
+                  backtest:
+                      start_time: 2026-01-04
+                      end_time: 2026-08-10
+                      account: 1000000
+                      benchmark: SPY
+                      exchange_kwargs:
+                          codes: "{{ UNIVERSE }}"
+                          deal_price: $close
+                          freq: day
+                          open_cost: 0.0005
+                          close_cost: 0.0015
+                          min_cost: 5.0
+              risk_analysis_freq: 1d
@@ -1,44 +1,68 @@
-# Signal-quality gate backtest for 2025
+# -----------------------------------------------------------------------------
+# EXP 20 - R1: 2-seed ensemble (seeds 42,7), TopkDropout baseline.
+#
+# Runtime cut: 2 seeds instead of 5. Everything else identical to the reference
+# (test 2026-01-04..2026-08-10, SPY, costs 5bp/15bp). Measures whether the
+# 2-seed ensemble keeps the reference quality at ~2/5 the training time.
+#
+# Run:
+#   rd_run_workflow config_path=experiments/workflows/exp20-risk-limit-improve/r1_2seed.yaml \
+#       experiment_name=tac-rd-risk-limit
+# -----------------------------------------------------------------------------
 {%- set LAKE = TAC_LAKE_DIR %}
 {%- set UNIVERSE = "SPY,QQQ,DIA,IWM,MDY,VTI,VOO,VEA,VWO,VT,EFA,EEM,TLT,IEF,SHY,AGG,BND,LQD,HYG,JNK,EMB,GLD,SLV,USO,UNG,DBA,DBC,XLK,XLF,XLE,XLV,XLI,XLY,XLP,XLU,XLB,XLRE,ARKK,SMH,SOXX,IBB,XBI,ITA,XAR,ICLN,TAN,FDN,IGV,ESPO,REM" %}
-{%- set SP_FIELDS = "sp_ret,sp_ou_zscore,sp_ou_half_life,sp_ou_revert,sp_hmm_p_regime1,sp_hmm_state,sp_jump_ratio,sp_jump_flag,sp_jump_tail,sp_max_move,sp_rv1,sp_rv5,sp_rv22,sp_vol_ratio_5_22,sp_vol_ratio_1_22,sp_trend_slope_5,sp_trend_slope_20,sp_trend_slope_60,sp_logp,sp_hurst_exponent,sp_sig_level1_lead,sp_sig_level1_lag,sp_sig_level2_lead_lag,sp_sig_level2_lag_lead" %}
-{# Gate computed on-the-fly from signal + lake close prices #}
+{%- set SP_FIELDS = "sp_ret,sp_jump_ratio,sp_jump_flag,sp_jump_tail,sp_max_move,sp_rv1,sp_rv5,sp_rv22,sp_vol_ratio_5_22,sp_vol_ratio_1_22,sp_trend_slope_5,sp_trend_slope_20,sp_trend_slope_60,sp_logp,sp_hurst_exponent,sp_sig_level1_lead,sp_sig_level1_lag,sp_sig_level2_lead_lag,sp_sig_level2_lag_lead" %}

 qlib_init:
    provider_uri: "{{ LAKE }}"
    region: us
    expression_cache: null
    dataset_cache: null
+
    calendar_provider:
        class: tac_qlib.data.providers.LakeCalendarProvider
-        kwargs: { lake_root: "{{ LAKE }}", market: US }
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
    instrument_provider:
        class: tac_qlib.data.providers.LakeInstrumentProvider
-        kwargs: { lake_root: "{{ LAKE }}", market: US, markets: {} }
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+            markets: {}
    feature_provider:
        class: tac_qlib.data.providers.LakeFeatureProvider
-        kwargs: { lake_root: "{{ LAKE }}", market: US }
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+
    exp_manager:
        class: MLflowExpManager
        module_path: qlib.workflow.expm
        kwargs:
            uri: "sqlite:///{{ LAKE }}/mlruns.db"
-            default_exp_name: "tac-rd-sq-gate-2025"
+            default_exp_name: "tac-rd-risk-limit"

 task:
    model:
-        class: LGBModel
-        module_path: qlib.contrib.model.gbdt
+        class: RankICEnsembleLGBModel
+        module_path: tac_qlib.contrib.model.rank_ensemble
        kwargs:
            loss: mse
-            learning_rate: 0.05
-            num_leaves: 15
-            n_estimators: 200
+            learning_rate: 0.02
+            num_leaves: 31
+            n_estimators: 3000
+            num_boost_round: 3000
+            early_stopping_rounds: 200
+            min_data_in_leaf: 20
+            lambda_l2: 0.5
            colsample_bytree: 0.8
            subsample: 0.8
            subsample_freq: 1
-            reg_alpha: 0.01
-            reg_lambda: 0.01
+            reg_alpha: 0.1
+            reg_lambda: 1.0
+            seeds: "42,7"
+            parallel: 2

    dataset:
        class: DatasetH
@@ -50,9 +74,9 @@ task:
                kwargs:
                    instruments: "{{ UNIVERSE }}"
                    start_time: 2015-01-03
-                    end_time: 2025-12-31
-                    fit_start_time: 2015-01-03
-                    fit_end_time: 2026-01-03
+                    end_time: 2026-08-14
+                    fit_start_time: 2016-01-04
+                    fit_end_time: 2025-09-01
                    freq: day
                    lake_root: "{{ LAKE }}"
                    market: US
@@ -70,9 +94,9 @@ task:
                        - class: Fillna
                          kwargs: {}
            segments:
-                train: [2015-01-03, 2025-09-01]
+                train: [2016-01-04, 2025-09-01]
                valid: [2025-09-03, 2026-01-03]
-                test: [2025-01-02, 2025-12-31]
+                test: [2026-01-04, 2026-08-10]

    record:
        - class: SignalRecord
@@ -80,29 +104,25 @@ task:
          kwargs: {}
        - class: SigAnaRecord
          module_path: qlib.workflow.record_temp
-          kwargs: { ana_long_short: true, ann_scaler: 252 }
+          kwargs:
+              ana_long_short: true
+              ann_scaler: 252
        - class: PortAnaRecord
          module_path: qlib.workflow.record_temp
          kwargs:
              config:
                  strategy:
-                      class: SignalQualityGateStrategy
-                      module_path: tac_qlib.contrib.strategy.signal_quality_gate
+                      class: TopkDropoutStrategy
+                      module_path: qlib.contrib.strategy
                      kwargs:
                          signal: "<PRED>"
-                          lake_root: "{{ LAKE }}"
-                          gate_topk: 10
-                          gate_lookback: 5
-                          gate_threshold: 0.5
-                          gate_start: "2015-01-03"
-                          gate_end: "2025-12-31"
                          topk: 10
-                          n_drop: 1
+                          n_drop: 2
                          only_tradable: true
                          risk_degree: 0.95
                  backtest:
-                      start_time: 2025-01-02
-                      end_time: 2025-12-31
+                      start_time: 2026-01-04
+                      end_time: 2026-08-10
                      account: 1000000
                      benchmark: SPY
                      exchange_kwargs:
@@ -0,0 +1,136 @@
+# -----------------------------------------------------------------------------
+# EXP 20 - R1: 2-seed ensemble (seeds 42,7), TopkDropout baseline.
+#
+# Runtime cut: 2 seeds instead of 5. Everything else identical to the reference
+# (test 2026-01-04..2026-08-10, SPY, costs 5bp/15bp). Measures whether the
+# 2-seed ensemble keeps the reference quality at ~2/5 the training time.
+#
+# Run:
+#   rd_run_workflow config_path=experiments/workflows/exp20-risk-limit-improve/r1_2seed.yaml \
+#       experiment_name=tac-rd-risk-limit
+# -----------------------------------------------------------------------------
+{%- set LAKE = TAC_LAKE_DIR %}
+{%- set UNIVERSE = "SPY,QQQ,DIA,IWM,MDY,VTI,VOO,VEA,VWO,VT,EFA,EEM,TLT,IEF,SHY,AGG,BND,LQD,HYG,JNK,EMB,GLD,SLV,USO,UNG,DBA,DBC,XLK,XLF,XLE,XLV,XLI,XLY,XLP,XLU,XLB,XLRE,ARKK,SMH,SOXX,IBB,XBI,ITA,XAR,ICLN,TAN,FDN,IGV,ESPO,REM" %}
+{%- set SP_FIELDS = "sp_ret,sp_jump_ratio,sp_jump_flag,sp_jump_tail,sp_max_move,sp_rv1,sp_rv5,sp_rv22,sp_vol_ratio_5_22,sp_vol_ratio_1_22,sp_trend_slope_5,sp_trend_slope_20,sp_trend_slope_60,sp_logp,sp_hurst_exponent,sp_sig_level1_lead,sp_sig_level1_lag,sp_sig_level2_lead_lag,sp_sig_level2_lag_lead" %}
+
+qlib_init:
+    provider_uri: "{{ LAKE }}"
+    region: us
+    expression_cache: null
+    dataset_cache: null
+
+    calendar_provider:
+        class: tac_qlib.data.providers.LakeCalendarProvider
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+    instrument_provider:
+        class: tac_qlib.data.providers.LakeInstrumentProvider
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+            markets: {}
+    feature_provider:
+        class: tac_qlib.data.providers.LakeFeatureProvider
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+
+    exp_manager:
+        class: MLflowExpManager
+        module_path: qlib.workflow.expm
+        kwargs:
+            uri: "sqlite:///{{ LAKE }}/mlruns.db"
+            default_exp_name: "tac-rd-risk-limit"
+
+task:
+    model:
+        class: RankICEnsembleLGBModel
+        module_path: tac_qlib.contrib.model.rank_ensemble
+        kwargs:
+            loss: mse
+            learning_rate: 0.02
+            num_leaves: 31
+            n_estimators: 3000
+            num_boost_round: 3000
+            early_stopping_rounds: 200
+            min_data_in_leaf: 20
+            lambda_l2: 0.5
+            colsample_bytree: 0.8
+            subsample: 0.8
+            subsample_freq: 1
+            reg_alpha: 0.1
+            reg_lambda: 1.0
+            seeds: "42,7,2026,99,123"
+            parallel: 5
+
+    dataset:
+        class: DatasetH
+        module_path: qlib.data.dataset
+        kwargs:
+            handler:
+                class: TACHandler
+                module_path: tac_qlib.contrib.data.handler
+                kwargs:
+                    instruments: "{{ UNIVERSE }}"
+                    start_time: 2015-01-03
+                    end_time: 2026-08-14
+                    fit_start_time: 2016-01-04
+                    fit_end_time: 2025-09-01
+                    freq: day
+                    lake_root: "{{ LAKE }}"
+                    market: US
+                    label: "Ref($close,-6)/Ref($close,-1)-1"
+                    feature_fields: "$open,$high,$low,$close,$vwap,$volume,{{ SP_FIELDS }}"
+                    infer_processors:
+                        - class: DropAllNaN
+                          kwargs: {}
+                        - class: ProcessInf
+                          kwargs: {}
+                        - class: CSRankNorm
+                          kwargs: {}
+                        - class: ZScoreNorm
+                          kwargs: {}
+                        - class: Fillna
+                          kwargs: {}
+            segments:
+                train: [2016-01-04, 2025-09-01]
+                valid: [2025-09-03, 2026-01-03]
+                test: [2026-01-04, 2026-08-10]
+
+    record:
+        - class: SignalRecord
+          module_path: qlib.workflow.record_temp
+          kwargs: {}
+        - class: SigAnaRecord
+          module_path: qlib.workflow.record_temp
+          kwargs:
+              ana_long_short: true
+              ann_scaler: 252
+        - class: PortAnaRecord
+          module_path: qlib.workflow.record_temp
+          kwargs:
+              config:
+                  strategy:
+                      class: MomentumGateTopk
+                      module_path: tac_qlib.contrib.strategy.momentum_gate
+                      kwargs:
+                          signal: "<PRED>"
+                          topk: 10
+                          n_drop: 2
+                          min_momentum: 0.0
+                          only_tradable: true
+                          risk_degree: 0.95
+                  backtest:
+                      start_time: 2026-01-04
+                      end_time: 2026-08-10
+                      account: 1000000
+                      benchmark: SPY
+                      exchange_kwargs:
+                          codes: "{{ UNIVERSE }}"
+                          deal_price: $close
+                          freq: day
+                          open_cost: 0.0005
+                          close_cost: 0.0015
+                          min_cost: 5.0
+              risk_analysis_freq: 1d
@@ -0,0 +1,138 @@
+# -----------------------------------------------------------------------------
+# EXP 20 - R1: 2-seed ensemble (seeds 42,7), TopkDropout baseline.
+#
+# Runtime cut: 2 seeds instead of 5. Everything else identical to the reference
+# (test 2026-01-04..2026-08-10, SPY, costs 5bp/15bp). Measures whether the
+# 2-seed ensemble keeps the reference quality at ~2/5 the training time.
+#
+# Run:
+#   rd_run_workflow config_path=experiments/workflows/exp20-risk-limit-improve/r1_2seed.yaml \
+#       experiment_name=tac-rd-risk-limit
+# -----------------------------------------------------------------------------
+{%- set LAKE = TAC_LAKE_DIR %}
+{%- set UNIVERSE = "SPY,QQQ,DIA,IWM,MDY,VTI,VOO,VEA,VWO,VT,EFA,EEM,TLT,IEF,SHY,AGG,BND,LQD,HYG,JNK,EMB,GLD,SLV,USO,UNG,DBA,DBC,XLK,XLF,XLE,XLV,XLI,XLY,XLP,XLU,XLB,XLRE,ARKK,SMH,SOXX,IBB,XBI,ITA,XAR,ICLN,TAN,FDN,IGV,ESPO,REM" %}
+{%- set SP_FIELDS = "sp_ret,sp_jump_ratio,sp_jump_flag,sp_jump_tail,sp_max_move,sp_rv1,sp_rv5,sp_rv22,sp_vol_ratio_5_22,sp_vol_ratio_1_22,sp_trend_slope_5,sp_trend_slope_20,sp_trend_slope_60,sp_logp,sp_hurst_exponent,sp_sig_level1_lead,sp_sig_level1_lag,sp_sig_level2_lead_lag,sp_sig_level2_lag_lead" %}
+
+qlib_init:
+    provider_uri: "{{ LAKE }}"
+    region: us
+    expression_cache: null
+    dataset_cache: null
+
+    calendar_provider:
+        class: tac_qlib.data.providers.LakeCalendarProvider
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+    instrument_provider:
+        class: tac_qlib.data.providers.LakeInstrumentProvider
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+            markets: {}
+    feature_provider:
+        class: tac_qlib.data.providers.LakeFeatureProvider
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+
+    exp_manager:
+        class: MLflowExpManager
+        module_path: qlib.workflow.expm
+        kwargs:
+            uri: "sqlite:///{{ LAKE }}/mlruns.db"
+            default_exp_name: "tac-rd-risk-limit"
+
+task:
+    model:
+        class: RankICEnsembleLGBModel
+        module_path: tac_qlib.contrib.model.rank_ensemble
+        kwargs:
+            loss: mse
+            learning_rate: 0.02
+            num_leaves: 31
+            n_estimators: 3000
+            num_boost_round: 3000
+            early_stopping_rounds: 200
+            min_data_in_leaf: 20
+            lambda_l2: 0.5
+            colsample_bytree: 0.8
+            subsample: 0.8
+            subsample_freq: 1
+            reg_alpha: 0.1
+            reg_lambda: 1.0
+            seeds: "42,7,2026,99,123"
+            parallel: 5
+
+    dataset:
+        class: DatasetH
+        module_path: qlib.data.dataset
+        kwargs:
+            handler:
+                class: TACHandler
+                module_path: tac_qlib.contrib.data.handler
+                kwargs:
+                    instruments: "{{ UNIVERSE }}"
+                    start_time: 2015-01-03
+                    end_time: 2026-08-14
+                    fit_start_time: 2016-01-04
+                    fit_end_time: 2025-09-01
+                    freq: day
+                    lake_root: "{{ LAKE }}"
+                    market: US
+                    label: "Ref($close,-6)/Ref($close,-1)-1"
+                    feature_fields: "$open,$high,$low,$close,$vwap,$volume,{{ SP_FIELDS }}"
+                    infer_processors:
+                        - class: DropAllNaN
+                          kwargs: {}
+                        - class: ProcessInf
+                          kwargs: {}
+                        - class: CSRankNorm
+                          kwargs: {}
+                        - class: ZScoreNorm
+                          kwargs: {}
+                        - class: Fillna
+                          kwargs: {}
+            segments:
+                train: [2016-01-04, 2025-09-01]
+                valid: [2025-09-03, 2026-01-03]
+                test: [2026-01-04, 2026-08-10]
+
+    record:
+        - class: SignalRecord
+          module_path: qlib.workflow.record_temp
+          kwargs: {}
+        - class: SigAnaRecord
+          module_path: qlib.workflow.record_temp
+          kwargs:
+              ana_long_short: true
+              ann_scaler: 252
+        - class: PortAnaRecord
+          module_path: qlib.workflow.record_temp
+          kwargs:
+              config:
+                  strategy:
+                      class: HmmRiskTopk
+                      module_path: tac_qlib.contrib.strategy.hmm_risk
+                      kwargs:
+                          signal: "<PRED>"
+                          topk: 10
+                          n_drop: 2
+                          hmm_pause_pct: 0.70
+                          drawdown_pause_pct: 8.0
+                          liquidity_floor_adv: 5000000
+                          only_tradable: true
+                          risk_degree: 0.95
+                  backtest:
+                      start_time: 2026-01-04
+                      end_time: 2026-08-10
+                      account: 1000000
+                      benchmark: SPY
+                      exchange_kwargs:
+                          codes: "{{ UNIVERSE }}"
+                          deal_price: $close
+                          freq: day
+                          open_cost: 0.0005
+                          close_cost: 0.0015
+                          min_cost: 5.0
+              risk_analysis_freq: 1d
@@ -0,0 +1,137 @@
+# -----------------------------------------------------------------------------
+# EXP 20 - R1: 2-seed ensemble (seeds 42,7), TopkDropout baseline.
+#
+# Runtime cut: 2 seeds instead of 5. Everything else identical to the reference
+# (test 2026-01-04..2026-08-10, SPY, costs 5bp/15bp). Measures whether the
+# 2-seed ensemble keeps the reference quality at ~2/5 the training time.
+#
+# Run:
+#   rd_run_workflow config_path=experiments/workflows/exp20-risk-limit-improve/r1_2seed.yaml \
+#       experiment_name=tac-rd-risk-limit
+# -----------------------------------------------------------------------------
+{%- set LAKE = TAC_LAKE_DIR %}
+{%- set UNIVERSE = "SPY,QQQ,DIA,IWM,MDY,VTI,VOO,VEA,VWO,VT,EFA,EEM,TLT,IEF,SHY,AGG,BND,LQD,HYG,JNK,EMB,GLD,SLV,USO,UNG,DBA,DBC,XLK,XLF,XLE,XLV,XLI,XLY,XLP,XLU,XLB,XLRE,ARKK,SMH,SOXX,IBB,XBI,ITA,XAR,ICLN,TAN,FDN,IGV,ESPO,REM" %}
+{%- set SP_FIELDS = "sp_ret,sp_jump_ratio,sp_jump_flag,sp_jump_tail,sp_max_move,sp_rv1,sp_rv5,sp_rv22,sp_vol_ratio_5_22,sp_vol_ratio_1_22,sp_trend_slope_5,sp_trend_slope_20,sp_trend_slope_60,sp_logp,sp_hurst_exponent,sp_sig_level1_lead,sp_sig_level1_lag,sp_sig_level2_lead_lag,sp_sig_level2_lag_lead" %}
+
+qlib_init:
+    provider_uri: "{{ LAKE }}"
+    region: us
+    expression_cache: null
+    dataset_cache: null
+
+    calendar_provider:
+        class: tac_qlib.data.providers.LakeCalendarProvider
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+    instrument_provider:
+        class: tac_qlib.data.providers.LakeInstrumentProvider
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+            markets: {}
+    feature_provider:
+        class: tac_qlib.data.providers.LakeFeatureProvider
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+
+    exp_manager:
+        class: MLflowExpManager
+        module_path: qlib.workflow.expm
+        kwargs:
+            uri: "sqlite:///{{ LAKE }}/mlruns.db"
+            default_exp_name: "tac-rd-risk-limit"
+
+task:
+    model:
+        class: RankICEnsembleLGBModel
+        module_path: tac_qlib.contrib.model.rank_ensemble
+        kwargs:
+            loss: mse
+            learning_rate: 0.02
+            num_leaves: 31
+            n_estimators: 3000
+            num_boost_round: 3000
+            early_stopping_rounds: 200
+            min_data_in_leaf: 20
+            lambda_l2: 0.5
+            colsample_bytree: 0.8
+            subsample: 0.8
+            subsample_freq: 1
+            reg_alpha: 0.1
+            reg_lambda: 1.0
+            seeds: "42,7,2026,99,123"
+            weight_mode: rolling_ic
+            rolling_ic_window: 21
+            parallel: 5
+
+    dataset:
+        class: DatasetH
+        module_path: qlib.data.dataset
+        kwargs:
+            handler:
+                class: TACHandler
+                module_path: tac_qlib.contrib.data.handler
+                kwargs:
+                    instruments: "{{ UNIVERSE }}"
+                    start_time: 2015-01-03
+                    end_time: 2026-08-14
+                    fit_start_time: 2016-01-04
+                    fit_end_time: 2025-09-01
+                    freq: day
+                    lake_root: "{{ LAKE }}"
+                    market: US
+                    label: "Ref($close,-6)/Ref($close,-1)-1"
+                    feature_fields: "$open,$high,$low,$close,$vwap,$volume,{{ SP_FIELDS }}"
+                    infer_processors:
+                        - class: DropAllNaN
+                          kwargs: {}
+                        - class: ProcessInf
+                          kwargs: {}
+                        - class: CSRankNorm
+                          kwargs: {}
+                        - class: ZScoreNorm
+                          kwargs: {}
+                        - class: Fillna
+                          kwargs: {}
+            segments:
+                train: [2016-01-04, 2025-09-01]
+                valid: [2025-09-03, 2026-01-03]
+                test: [2026-01-04, 2026-08-10]
+
+    record:
+        - class: SignalRecord
+          module_path: qlib.workflow.record_temp
+          kwargs: {}
+        - class: SigAnaRecord
+          module_path: qlib.workflow.record_temp
+          kwargs:
+              ana_long_short: true
+              ann_scaler: 252
+        - class: PortAnaRecord
+          module_path: qlib.workflow.record_temp
+          kwargs:
+              config:
+                  strategy:
+                      class: TopkDropoutStrategy
+                      module_path: qlib.contrib.strategy
+                      kwargs:
+                          signal: "<PRED>"
+                          topk: 10
+                          n_drop: 2
+                          only_tradable: true
+                          risk_degree: 0.95
+                  backtest:
+                      start_time: 2026-01-04
+                      end_time: 2026-08-10
+                      account: 1000000
+                      benchmark: SPY
+                      exchange_kwargs:
+                          codes: "{{ UNIVERSE }}"
+                          deal_price: $close
+                          freq: day
+                          open_cost: 0.0005
+                          close_cost: 0.0015
+                          min_cost: 5.0
+              risk_analysis_freq: 1d
@@ -0,0 +1,135 @@
+# -----------------------------------------------------------------------------
+# EXP 20 - R1: 2-seed ensemble (seeds 42,7), TopkDropout baseline.
+#
+# Runtime cut: 2 seeds instead of 5. Everything else identical to the reference
+# (test 2026-01-04..2026-08-10, SPY, costs 5bp/15bp). Measures whether the
+# 2-seed ensemble keeps the reference quality at ~2/5 the training time.
+#
+# Run:
+#   rd_run_workflow config_path=experiments/workflows/exp20-risk-limit-improve/r1_2seed.yaml \
+#       experiment_name=tac-rd-risk-limit
+# -----------------------------------------------------------------------------
+{%- set LAKE = TAC_LAKE_DIR %}
+{%- set UNIVERSE = "SPY,QQQ,DIA,IWM,MDY,VTI,VOO,VEA,VWO,VT,EFA,EEM,TLT,IEF,SHY,AGG,BND,LQD,HYG,JNK,EMB,GLD,SLV,USO,UNG,DBA,DBC,XLK,XLF,XLE,XLV,XLI,XLY,XLP,XLU,XLB,XLRE,ARKK,SMH,SOXX,IBB,XBI,ITA,XAR,ICLN,TAN,FDN,IGV,ESPO,REM" %}
+{%- set SP_FIELDS = "sp_ret,sp_jump_ratio,sp_jump_flag,sp_jump_tail,sp_max_move,sp_rv1,sp_rv5,sp_rv22,sp_vol_ratio_5_22,sp_vol_ratio_1_22,sp_trend_slope_5,sp_trend_slope_20,sp_trend_slope_60,sp_logp,sp_hurst_exponent,sp_sig_level1_lead,sp_sig_level1_lag,sp_sig_level2_lead_lag,sp_sig_level2_lag_lead,sma_3,ema_3" %}
+
+qlib_init:
+    provider_uri: "{{ LAKE }}"
+    region: us
+    expression_cache: null
+    dataset_cache: null
+
+    calendar_provider:
+        class: tac_qlib.data.providers.LakeCalendarProvider
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+    instrument_provider:
+        class: tac_qlib.data.providers.LakeInstrumentProvider
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+            markets: {}
+    feature_provider:
+        class: tac_qlib.data.providers.LakeFeatureProvider
+        kwargs:
+            lake_root: "{{ LAKE }}"
+            market: US
+
+    exp_manager:
+        class: MLflowExpManager
+        module_path: qlib.workflow.expm
+        kwargs:
+            uri: "sqlite:///{{ LAKE }}/mlruns.db"
+            default_exp_name: "tac-rd-risk-limit"
+
+task:
+    model:
+        class: RankICEnsembleLGBModel
+        module_path: tac_qlib.contrib.model.rank_ensemble
+        kwargs:
+            loss: mse
+            learning_rate: 0.02
+            num_leaves: 31
+            n_estimators: 3000
+            num_boost_round: 3000
+            early_stopping_rounds: 200
+            min_data_in_leaf: 20
+            lambda_l2: 0.5
+            colsample_bytree: 0.8
+            subsample: 0.8
+            subsample_freq: 1
+            reg_alpha: 0.1
+            reg_lambda: 1.0
+            seeds: "42,7,2026,99,123"
+            parallel: 5
+
+    dataset:
+        class: DatasetH
+        module_path: qlib.data.dataset
+        kwargs:
+            handler:
+                class: TACHandler
+                module_path: tac_qlib.contrib.data.handler
+                kwargs:
+                    instruments: "{{ UNIVERSE }}"
+                    start_time: 2015-01-03
+                    end_time: 2026-08-14
+                    fit_start_time: 2016-01-04
+                    fit_end_time: 2025-09-01
+                    freq: day
+                    lake_root: "{{ LAKE }}"
+                    market: US
+                    label: "Ref($close,-6)/Ref($close,-1)-1"
+                    feature_fields: "$open,$high,$low,$close,$vwap,$volume,{{ SP_FIELDS }}"
+                    infer_processors:
+                        - class: DropAllNaN
+                          kwargs: {}
+                        - class: ProcessInf
+                          kwargs: {}
+                        - class: CSRankNorm
+                          kwargs: {}
+                        - class: ZScoreNorm
+                          kwargs: {}
+                        - class: Fillna
+                          kwargs: {}
+            segments:
+                train: [2016-01-04, 2025-09-01]
+                valid: [2025-09-03, 2026-01-03]
+                test: [2026-01-04, 2026-08-10]
+
+    record:
+        - class: SignalRecord
+          module_path: qlib.workflow.record_temp
+          kwargs: {}
+        - class: SigAnaRecord
+          module_path: qlib.workflow.record_temp
+          kwargs:
+              ana_long_short: true
+              ann_scaler: 252
+        - class: PortAnaRecord
+          module_path: qlib.workflow.record_temp
+          kwargs:
+              config:
+                  strategy:
+                      class: TopkDropoutStrategy
+                      module_path: qlib.contrib.strategy
+                      kwargs:
+                          signal: "<PRED>"
+                          topk: 10
+                          n_drop: 2
+                          only_tradable: true
+                          risk_degree: 0.95
+                  backtest:
+                      start_time: 2026-01-04
+                      end_time: 2026-08-10
+                      account: 1000000
+                      benchmark: SPY
+                      exchange_kwargs:
+                          codes: "{{ UNIVERSE }}"
+                          deal_price: $close
+                          freq: day
+                          open_cost: 0.0005
+                          close_cost: 0.0015
+                          min_cost: 5.0
+              risk_analysis_freq: 1d
Author	SHA1	Message	Date
zhaoli	202589fc3e	exp 20: add deterministic baseline (parallel=1) to isolate LightGBM non-determinism as root cause of exp18 vs R0 divergence	2026-08-17 09:59:26 +00:00
zhaoli	3ded27e5f9	finish experiment 20 (exp/20-improve-the-risk-limit-reference-signal)	2026-08-17 09:33:24 +00:00
zhaoli	125be7b96f	exp 20: add R0 same-env 5-seed TopkDropout baseline (fair base for R2/R3/R5; pre-reset exp18 baseline is not comparable across env resets)	2026-08-17 07:16:03 +00:00
zhaoli	20857f76be	exp 20: rerun R2/R3/R5 with 5 seeds for fair comparison vs 5-seed baseline	2026-08-17 05:52:48 +00:00
zhaoli	9c0cf096f8	finish experiment 20 (exp/20-improve-the-risk-limit-reference-signal)	2026-08-17 05:46:09 +00:00
zhaoli	80c7230e17	exp 20: sync fixed MomentumGateTopk + HmmRiskTopk + rolling-IC rank_ensemble into code snapshot	2026-08-17 04:08:07 +00:00
zhaoli	07e78c2bd9	exp 20: r4 rolling-IC blend uses 5 seeds (needs seed diversity for the IC weighting)	2026-08-17 04:04:38 +00:00
zhaoli	e09ab7f05a	exp 20: exp 20: 5 improvement workflows (R1 2seed, R2 momentum gate, R3 HMM risk gate, R4 rolling-IC blend, R5 MA3/EWMA features) + strategy modules + lake sma_3/ema_3 persisted	2026-08-17 02:56:46 +00:00
zhaoli	337f6e17d8	start experiment 20 (exp/20-improve-the-risk-limit-reference-signal)	2026-08-17 02:55:26 +00:00
zhaoli	7fad62a4ef	finish experiment 18 (exp/18-risk-limit-control-on-the-reference-ense)	2026-08-17 01:14:23 +00:00
zhaoli	f9ef005e9a	exp 18: exp 18: risk-limit control - baseline workflow YAML + calibration notes	2026-08-17 00:07:32 +00:00
zhaoli	c09997c7e2	start experiment 18 (exp/18-risk-limit-control-on-the-reference-ense)	2026-08-17 00:05:51 +00:00
zhaoli	32477c7bb8	exp 12: isolation ensemble workflow yaml (ablate-B features + full history) + finish record	2026-08-14 05:37:13 +00:00
zhaoli	e657c58758	start exp 9 (sp5d-feature-family-ablation): baseline all-24 + generic-only 19 workflow YAMLs	2026-08-13 04:32:30 +00:00