QuantStrategyLab · Pigbibi · Jul 5, 2026 · Jul 5, 2026
diff --git a/README.md b/README.md
@@ -28,7 +28,7 @@ These profiles use platform-provided `market_history` and do not require a separ
 | Profile | Name | Input | Benchmark | Current role |
 | --- | --- | --- | --- | --- |
 | `cn_dividend_quality_snapshot` | CN Dividend Quality Snapshot | `feature_snapshot` + manifest | `510300` | Dividend + quality composite selector for defensive allocation. |
-| `cn_equity_combo` | CN Equity Combo | `market_history` + `feature_snapshot` | `510300` | Research-only composite track combining ETF, stock, and dividend sleeves. |
+| `cn_equity_combo` | CN Equity Combo | `market_history` + `feature_snapshot` | `510300` | Research-only orchestrator combining ETF, growth sleeves, and dividend sleeves. |
 
 ### Live candidate
 
@@ -42,6 +42,13 @@ These profiles use platform-provided `market_history` and do not require a separ
 | --- | --- | --- | --- | --- |
 | `cn_chinext_growth_momentum_quality_snapshot` | CN ChiNext Growth Momentum Quality Snapshot | `feature_snapshot` + manifest | `399006` | Research-only ChiNext growth-momentum-quality selector. |
 
+### Research-only growth sleeves
+
+| Profile | Name | Input | Benchmark | Current role |
+| --- | --- | --- | --- | --- |
+| `cn_chinext_growth_momentum_quality` | CN ChiNext Growth Momentum Quality | `market_history` | `510300` | Research-only ChiNext growth sleeve; board-specific growth + momentum + quality. |
+| `cn_star_growth_momentum_quality` | CN Star Growth Momentum Quality | `market_history` | `510300` | Research-only STAR growth sleeve; stricter liquidity and concentration controls. |
+
 ### Planned external snapshot scaffolds
 
 | Profile | Name | Notes |
@@ -101,6 +108,9 @@ The intended A-share runtime platform is:
 
 - [`docs/platform_integration.md`](docs/platform_integration.md)
 - [`docs/research/cn_index_etf_tactical_rotation.md`](docs/research/cn_index_etf_tactical_rotation.md)
+- [`docs/research/cn_growth_sleeve_redesign_20260705.md`](docs/research/cn_growth_sleeve_redesign_20260705.md)
+- [`docs/research/cn_growth_sleeve_structure_20260705.md`](docs/research/cn_growth_sleeve_structure_20260705.md)
+- [`docs/research/cn_chinext_growth_momentum_quality_snapshot_20260705.md`](docs/research/cn_chinext_growth_momentum_quality_snapshot_20260705.md)
 
 ## Safety and contribution notes
 

diff --git a/README.zh-CN.md b/README.zh-CN.md
@@ -34,6 +34,13 @@
 | --- | --- | --- | --- | --- |
 | `cn_chinext_growth_momentum_quality_snapshot` | CN ChiNext Growth Momentum Quality Snapshot | `feature_snapshot` + manifest | `399006` | **研究轨**：创业板成长动量质量选择。 |
 
+### 研究侧成长 sleeve
+
+| Profile | 名称 | 输入 | 基准 | 当前角色 |
+| --- | --- | --- | --- | --- |
+| `cn_chinext_growth_momentum_quality` | CN ChiNext Growth Momentum Quality | `market_history` | `510300` | **研究轨**：创业板成长 sleeve，板块级成长 + 动量 + 质量。 |
+| `cn_star_growth_momentum_quality` | CN Star Growth Momentum Quality | `market_history` | `510300` | **研究轨**：科创板成长 sleeve，流动性和集中度更严格。 |
+
 ### 计划中的 snapshot-backed 策略
 
 | Profile | 名称 | 输入 | 基准 | 当前角色 |
@@ -94,6 +101,9 @@ python scripts/smoke_cn_index_etf_tactical_rotation_dry_run.py --json
 
 - [`docs/platform_integration.md`](docs/platform_integration.md)
 - [`docs/research/cn_index_etf_tactical_rotation.md`](docs/research/cn_index_etf_tactical_rotation.md)
+- [`docs/research/cn_growth_sleeve_redesign_20260705.md`](docs/research/cn_growth_sleeve_redesign_20260705.md)
+- [`docs/research/cn_growth_sleeve_structure_20260705.md`](docs/research/cn_growth_sleeve_structure_20260705.md)
+- [`docs/research/cn_chinext_growth_momentum_quality_snapshot_20260705.md`](docs/research/cn_chinext_growth_momentum_quality_snapshot_20260705.md)
 
 ## 安全和贡献说明
 

diff --git a/docs/platform_integration.md b/docs/platform_integration.md
@@ -18,8 +18,10 @@
 | --- | --- | --- | --- |
 | `cn_index_etf_tactical_rotation` | `market_history` | `qmt` | `weight` |
 | `cn_dividend_quality_snapshot` | `feature_snapshot` + manifest | `qmt` | `weight` |
+| `cn_chinext_growth_momentum_quality` | `market_history` | `qmt` | `weight` |
 | `cn_chinext_growth_momentum_quality_snapshot` | `feature_snapshot` + manifest | `qmt` | `weight` |
 | `cn_chinext_tactical_rotation` | `market_history` | `qmt` | `weight` |
+| `cn_star_growth_momentum_quality` | `market_history` | `qmt` | `weight` |
 | `cn_equity_combo` | `market_history` + `feature_snapshot` | `qmt` | `weight` |
 
 ## Environment variables (planned for QmtPlatform)

diff --git a/docs/research/cn_growth_sleeve_redesign_20260705.md b/docs/research/cn_growth_sleeve_redesign_20260705.md
@@ -0,0 +1,124 @@
+# CN Growth Sleeve Redesign Plan (2026-07-05)
+
+> This is an engineering/research plan, not investment advice.
+
+## Why this needs a redesign
+
+The earlier read treated ChiNext and STAR as weaker or secondary because the
+existing strategy shapes were too narrow. That was the wrong conclusion.
+
+Recent market evidence shows both growth boards can be strong regime leaders:
+
+- ChiNext has printed major highs and has materially outperformed in the current
+  growth-led phase.
+- STAR market / STAR50 has also led in 2026 on a number of public data sources.
+
+So the issue is not "these boards are weak." The issue is:
+
+1. the current strategy shapes are too small or too tactical,
+2. the board-specific regime is not modeled explicitly enough,
+3. the combo layer is using the wrong abstraction.
+
+## Design conclusion
+
+ChiNext and STAR should be treated as **independent growth sleeves**, not as
+subordinate research leftovers.
+
+That means:
+
+- the board itself is a first-class universe,
+- the sleeve gets its own regime gate,
+- the sleeve gets its own growth / quality / liquidity logic,
+- the sleeve can later become a combo component, but not as a dumb appendage.
+
+## Recommended target architecture
+
+### 1) ChiNext growth sleeve
+
+Proposed new shape:
+
+- universe: ChiNext board
+- style: growth + momentum + quality
+- cadence: monthly
+- gating: benchmark trend filter + liquidity + overheat / crowding penalty
+
+Key signals:
+
+- revenue growth
+- profit growth
+- ROE / profitability stability
+- 12M-1M momentum
+- 60/120-day trend
+- liquidity and execution cost
+- board breadth / regime filter
+
+This should evolve from `cn_chinext_tactical_rotation` into a more explicit
+`cn_chinext_growth_momentum_quality` line.
+
+### 2) STAR growth sleeve
+
+Proposed new shape:
+
+- universe: STAR50 or STAR composite / broader STAR universe
+- style: growth + quality + liquidity with heavier risk controls
+- cadence: monthly
+- gating: stricter liquidity and concentration constraints than ChiNext
+
+STAR should not start as a narrow champion-picking experiment. It should begin
+as a board-level enhancement sleeve so the design matches how the market is
+actually behaving.
+
+### 3) CN combo as allocator, not alpha claimant
+
+`cn_equity_combo` should be rebuilt as a composition/orchestration layer:
+
+- industry ETF core
+- ChiNext growth sleeve
+- STAR growth sleeve
+- defensive dividend/quality sleeve
+
+The combo should decide weights and risk budgets. It should not pretend to be a
+single alpha source.
+
+## What should change in policy language
+
+- `research_backtest_only` should no longer imply "low priority" for ChiNext or
+  STAR.
+- `live_candidate` should only be assigned after the board-specific sleeve beats
+  the current main line on the same horizon.
+- `cn_chinext_tactical_rotation` and `cn_chinext_growth_momentum_quality_snapshot`
+  should be treated as **growth-sleeve candidates**, not discardable leftovers.
+
+## What not to do
+
+- Do not promote a board sleeve just because a short window is hot.
+- Do not merge board growth sleeves into the industry ETF main line.
+- Do not keep multiple weak variants alive just because each one has a slightly
+  different filter.
+- Do not force live enablement before the board-specific gate is complete.
+
+## Proposed implementation sequence
+
+1. Define a board-specific growth sleeve contract for ChiNext.
+2. Define a board-specific growth sleeve contract for STAR.
+3. Reframe `cn_equity_combo` as a true allocator/orchestrator.
+4. Re-run the same-cycle comparisons against the current industry ETF main line.
+5. Promote only if the sleeve is still competitive after liquidity and risk
+   penalties.
+
+## Practical status recommendation
+
+- `cn_industry_etf_rotation`: keep as the core A-share main line.
+- `cn_chinext_*`: redesign into a first-class growth sleeve.
+- `cn_star_*`: add as a separate growth sleeve once the STAR contract is
+  defined.
+- `cn_equity_combo`: rebuild later as composition logic.
+
+## Reference anchors
+
+These are the main external reference families used for the redesign:
+
+- STAR 50 / STAR 100 / STAR 200 official index construction rules from SSE.
+- ChiNext as a board-level growth universe with its own index structure.
+- MSCI factor index families for growth, quality, momentum, value, and low-volatility framing.
+- EDHEC / AQR trend-following research for regime and gating logic.
diff --git a/docs/research/cn_growth_sleeve_structure_20260705.md b/docs/research/cn_growth_sleeve_structure_20260705.md
@@ -0,0 +1,184 @@
+# CN Growth Sleeve Structure Draft (2026-07-05)
+
+> This is a design draft for the CN growth stack. It is not a trading
+> recommendation.
+
+## Goal
+
+Turn the CN growth side from "a few tactical or research leftovers" into a
+clean structure with three first-class roles:
+
+1. **ChiNext growth sleeve**
+2. **STAR growth sleeve**
+3. **CN combo orchestrator**
+
+This is a design correction, not just a parameter tweak.
+
+## Why the old shape was wrong
+
+The earlier shape treated ChiNext / STAR too much like side research tracks.
+That is too narrow for the current market regime.
+
+Public market evidence shows that both growth boards can lead strongly in a
+growth-led phase. The better interpretation is:
+
+- the boards themselves are valid growth universes,
+- the strategy shape needs board-specific gates,
+- the combo layer should allocate sleeves, not pretend to be the only alpha.
+
+## Proposed layer 1: ChiNext growth sleeve
+
+### Proposed profile name
+
+- `cn_chinext_growth_momentum_quality`
+
+### Role
+
+Board-specific growth sleeve for the ChiNext market.
+
+### Shape
+
+- universe: ChiNext board
+- cadence: monthly
+- style: growth + momentum + quality
+- explicit regime gate: benchmark trend + breadth + liquidity + crowding
+
+### Suggested signals
+
+- revenue growth
+- profit growth
+- ROE / profitability stability
+- 12M-1M momentum
+- 60/120-day trend
+- liquidity and execution cost
+- overheat / crowding penalty
+
+### Suggested behavior
+
+- Can be promoted from the existing research snapshot work.
+- Should not be forced to share the same gate model as the industry ETF main line.
+- Can later be used as a sleeve inside the combo orchestrator.
+
+## Proposed layer 2: STAR growth sleeve
+
+### Proposed profile name
+
+- `cn_star_growth_momentum_quality`
+
+### Role
+
+Board-specific growth sleeve for the STAR market / STAR50-like universe.
+
+### Shape
+
+- universe: STAR50, or a broader STAR composite universe such as STAR100 /
+  STAR200 / special innovation board universes if later evidence supports it
+- cadence: monthly
+- style: growth + quality + liquidity
+- stricter concentration and liquidity controls than ChiNext
+
+### Suggested signals
+
+- revenue / earnings growth
+- R&D intensity or innovation proxy
+- profitability stability
+- 12M-1M momentum
+- relative trend vs STAR benchmark
+- liquidity / turnover / spread penalty
+
+### Suggested behavior
+
+- Start as a board-level enhancement sleeve, not a narrow stock picker.
+- Use a stricter gate than ChiNext because STAR is more concentration- and
+  liquidity-sensitive.
+- Keep the first version simple and reviewable.
+
+## Proposed layer 3: CN combo orchestrator
+
+### Current problem
+
+`cn_equity_combo` currently reads like a blended research strategy. That is too
+ambiguous.
+
+### New role
+
+Make it a true orchestrator / allocator:
+
+- allocates across sleeves
+- manages regime budgets
+- decides when to emphasize offense vs defense
+- does **not** claim to be a single alpha source
+
+### Suggested sleeves
+
+- core industry ETF sleeve
+- ChiNext growth sleeve
+- STAR growth sleeve
+- defensive dividend / quality sleeve
+
+### Suggested runtime shape
+
+```json
+{
+  "profile": "cn_equity_combo",
+  "sleeves": [
+    {"profile": "cn_industry_etf_rotation", "weight": 0.40},
+    {"profile": "cn_chinext_growth_momentum_quality", "weight": 0.25},
+    {"profile": "cn_star_growth_momentum_quality", "weight": 0.15},
+    {"profile": "cn_dividend_quality_snapshot", "weight": 0.20}
+  ]
+}
+```
+
+The actual weights should stay configurable, but the separation of responsibilities
+should not change.
+
+## Gate model
+
+### Keep / tune
+
+- keep the industry ETF main line as the core A-share runtime strategy
+- keep the aggressive ETF variant as the controlled enhancement lane
+
+### Redesign
+
+- ChiNext tactical and ChiNext growth snapshot should be redesigned into the
+  ChiNext growth sleeve
+- STAR should get its own sleeve instead of being folded into the ChiNext logic
+- combo should be rebuilt as an orchestrator
+
+### Do not do
+
+- do not collapse board growth into the industry ETF main line
+- do not keep many small variants alive just because they all have slightly
+  different filters
+- do not promote before the board-specific sleeve beats the current live main
+  line on the same horizon
+
+## Evidence sequence
+
+1. Build the board-specific growth sleeve contract.
+2. Run same-cycle comparisons against the current industry ETF main line.
+3. Add liquidity / spread / crowding penalties.
+4. Confirm the sleeve still wins after penalties.
+5. Only then decide whether it becomes `live_candidate` or remains research.
+
+## External reference shape
+
+This structure matches the way official board/index sources describe the
+universes:
+
+- ChiNext is a growth-oriented Shenzhen board universe with strong innovation
+  characteristics.
+- STAR Market / STAR50 is the Shanghai innovation board universe, with a
+  naturally higher concentration and liquidity sensitivity.
+
+That is why both should be modeled as dedicated sleeves instead of generic
+tactical leftovers.
+
+## Reference anchors
+
+- SSE official STAR 50 / STAR 100 / STAR 200 index construction rules
+- SSE official index and fund disclosures for STAR-linked products
+- MSCI factor index families for growth / quality / momentum / low volatility
+- EDHEC and AQR trend-following research for regime gating patterns