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

code/MANIFEST.txt

@@ -1,5 +1,5 @@
 # TradeAC custom-qlib-code snapshot (auto-generated)
-# parent repo HEAD : f9d1fe66f6e4f8ace0d3d774e23f1c79def9bae0
+# parent repo HEAD : 125be7b96fb5975e798a0b4301eeb5809a8a181c
 #   tac-qlib/tac_qlib/contrib
 #   tac-qlib/tac_qlib/data
 # per-file hashes (git hash-object):
@@ -11,13 +11,15 @@
   871ff1e163c29261f140c3f53d42a41e6504c779  tac-qlib/tac_qlib/contrib/data/handler.py
   b151d139a0dcde87d74b21e7c4b729176ba5c39b  tac-qlib/tac_qlib/contrib/model/__init__.py
   ab958203f33a99d12c7d923b6efb435189231666  tac-qlib/tac_qlib/contrib/model/__pycache__/__init__.cpython-312.pyc
-  9dc36de7e343073b7d511349ee5aede086c38f94  tac-qlib/tac_qlib/contrib/model/__pycache__/rank_ensemble.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
-  d3f051f3a8650c42fedc7b367b966f7c74fb5789  tac-qlib/tac_qlib/contrib/model/rank_ensemble.py
+  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
   92e6e90eb0cd0a25142034560f27adb6b705b1a8  tac-qlib/tac_qlib/data/__init__.py
   0ed1ead6c1314a3f25784d453e54a15a8a04baaa  tac-qlib/tac_qlib/data/__pycache__/__init__.cpython-312.pyc

code/tac-qlib/tac_qlib/contrib/model/rank_ensemble.py

@@ -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)

code/tac-qlib/tac_qlib/contrib/strategy/hmm_risk.py

@@ -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)

code/tac-qlib/tac_qlib/contrib/strategy/momentum_gate.py

@@ -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)

workflows/README-moments.md

@@ -1,20 +0,0 @@
-# exp/10 sp5d-moment-features
-
-Variant C: generic-only 19 + 16 new moment/volatility families (skew, kurt,
-DSV+ratios, max_up/down, rv_ac1, rv_cv_22, sig lag-5). 35 sp_* fields, ou/hmm excluded.
-
-Run a3f7d1d40c3d4b839314fcf5b40f9b08 (tac-rd-moments / exp 12) — FINISHED.
-
-## Result: NEGATIVE (regression vs generic-only baseline)
-
-| Metric | generic-only 19 (run 7b1e797) | +moments 35 (run a3f7d1d) |
-|---|---|---|
-| Rank IC   | 0.0635 | 0.0466 |
-| Rank ICIR | 0.276  | 0.183  |
-| L-S Sharpe | 2.55  | 1.44  |
-| net excess (cost) | +3.1% IR 0.28 | -16.2% IR -1.57 |
-| MDD | -7.3% | -11.1% |
-
-Same failure mode as ou/hmm in exp 9: adding cross-sectional moment features
-to the 50-name panel degrades the rank signal. Generic-only 19 remains the
-best configuration. No further moment-family variants planned.

workflows/exp12_isolation_ensemble.yaml

@@ -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

workflows/exp18-risk-limit/a_baseline.yaml

@@ -1,22 +1,23 @@
 # -----------------------------------------------------------------------------
-# VARIANT C (generic + moments): keeps the winning generic-only 19-field set
+# EXP 18 - Risk-limit control: reference model + TopkDropout baseline (A).
-# (jump,har,trend,hurst,signature) and adds the NEW generic moment families the
+#
-# engine now exposes:
+# Signal/model identical to the reference (tac-rd-rank-ensemble-isolated,
-#   - realized skewness / kurtosis      (sp_rskew_5, sp_rskew_22, sp_rkurt_5, sp_rkurt_22)
+# run 0cea66d9...): RankICEnsembleLGBModel (parallel, 5 seeds) on the 50-ETF
-#   - downside semi-variance + ratios   (sp_dsv_1/5/22, sp_dsv_ratio_1/5/22)
+# SP-5d panel, test 2026-01-04..2026-08-10. This workflow reproduces the
-#   - signed max moves                  (sp_max_up, sp_max_down)
+# unconstrained TopkDropout baseline net-of-cost so the risk-limited variant
-#   - RV autocorr / vol-of-vol          (sp_rv_ac1, sp_rv_cv_22)
+# (same pred, liquidity/size/concentration caps) can be compared 1:1.
-#   - longer-lag signature terms        (sp_sig_level2_*_5)
+#
-# Drops the model-specific ou/hmm families (they scored high in importance but
+# The risk_limits spec itself is applied via rd_backtest / rd_strategy_targets
-# hurt the rank dimension in the all-24 run). Same panel/model as baseline.
+# (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/ablate_generic_moments.yaml \
+#   rd_run_workflow config_path=experiments/workflows/exp18-risk-limit/a_baseline.yaml \
-#       experiment_name=tac-rd-moments
+#       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_max_up,sp_max_down,sp_rv1,sp_rv5,sp_rv22,sp_vol_ratio_5_22,sp_vol_ratio_1_22,sp_rv_ac1,sp_rv_cv_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,sp_sig_level2_lead_lag_5,sp_sig_level2_lag_lead_5,sp_rskew_5,sp_rskew_22,sp_rkurt_5,sp_rkurt_22,sp_dsv_1,sp_dsv_5,sp_dsv_22,sp_dsv_ratio_1,sp_dsv_ratio_5,sp_dsv_ratio_22" %}
+{%- 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 }}"
@@ -46,12 +47,12 @@ qlib_init:
         module_path: qlib.workflow.expm
         kwargs:
             uri: "sqlite:///{{ LAKE }}/mlruns.db"
-            default_exp_name: "tac-rd-moments"
+            default_exp_name: "tac-rd-risk-limit"
 
 task:
     model:
-        class: RankICLGBModel
+        class: RankICEnsembleLGBModel
-        module_path: tac_qlib.contrib.model.rank_gbdt
+        module_path: tac_qlib.contrib.model.rank_ensemble
         kwargs:
             loss: mse
             learning_rate: 0.02
@@ -66,7 +67,8 @@ task:
             subsample_freq: 1
             reg_alpha: 0.1
             reg_lambda: 1.0
-            seed: 42
+            seeds: "42,7,2026,99,123"
+            parallel: 5
 
     dataset:
         class: DatasetH
@@ -78,8 +80,8 @@ task:
                 kwargs:
                     instruments: "{{ UNIVERSE }}"
                     start_time: 2015-01-03
-                    end_time: 2026-08-10
+                    end_time: 2026-08-14
-                    fit_start_time: 2015-01-03
+                    fit_start_time: 2016-01-04
                     fit_end_time: 2025-09-01
                     freq: day
                     lake_root: "{{ LAKE }}"
@@ -98,7 +100,7 @@ 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: [2026-01-04, 2026-08-10]
 

workflows/exp20-risk-limit-improve/r0_5seed_baseline.yaml

@@ -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: 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

workflows/exp20-risk-limit-improve/r0_det_baseline.yaml

@@ -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

workflows/exp20-risk-limit-improve/r1_2seed.yaml

@@ -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"
+            parallel: 2
+
+    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

workflows/exp20-risk-limit-improve/r2_momentum.yaml

@@ -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

workflows/exp20-risk-limit-improve/r3_hmm_risk.yaml

@@ -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

workflows/exp20-risk-limit-improve/r4_rolling_ic.yaml

@@ -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

workflows/exp20-risk-limit-improve/r5_ma3_ewma.yaml

@@ -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