English | 中文

🌱 cst-quant · Cheap · Stable · Trending

A-share deep-value quantitative strategy, inspired by Graham-style "cigar butt" investing. Three transparent factors — cheap valuation, low volatility, trend momentum — equally weighted, 50 stocks, quarterly rebalancing, always fully invested. No market timing, no macro calls, no black boxes.

Key Result — 2014-01 ~ 2026-06 (12.5 yrs): +776.27% cumulative return vs benchmark +121.24%, annualized +19.08%, Sharpe 0.67, max drawdown −37.18%.

---

Live Strategy

Factor	Weight	Description
F2 EP	1/3	Earnings yield = net profit / market cap
F5 LowVol	1/3	−40d daily return std dev
F6 MOM-40d	1/3	61-21 momentum (40-day trend window)

Metric	Value
Period	2014-01-01 ~ 2026-06-28 (12.5 yrs)
Cumulative Return	+776.27% (benchmark +121.24%)
Max Drawdown	−37.18%
Sharpe / Alpha / Beta	0.67 / 0.13 / 0.77

Cross-sample validation (AllA / CSI300 / CSI500) — 2026-06-29 — OVERALL: PASS. Code: results/P5-F2F5F6-40d-final-strategy.py

Composite score (equal-weight Z-score):

Score = ⅓·Z(EP) + ⅓·Z(−Vol) + ⅓·Z(MOM)  →  top 50

---

Project Structure

cst-quant/
├── README.md
├── README-zh.md
├── requirements.txt
├── LICENSE
├── archive/                           # Historical archive (theory doc, legacy code)
├── docs/
│   ├── CURRENT-STRATEGY.md            # Final strategy overview
│   ├── manual-investment-guide.md     # Manual investing guide
│   ├── prompts/                       # Research workflow prompts (00-04)
│   └── task-state/                    # Task state & debug notes
├── research/
│   ├── _index.md                      # Research dashboard
│   ├── scripts/                       # Research scripts (JoinQuant paste-to-run)
│   ├── reports/                       # Analysis reports
│   ├── decisions/                     # Decision records
│   └── specs/                         # Research specs
└── results/                           # Final deliverables
    ├── P5-F2F5F6-40d-final-strategy.py    # Production strategy (VOL=40, 777%/S=0.67)
    ├── manual-investment-guide.html        # Investing manual (HTML)
    ├── manual-investment-guide.md          # Investing manual (Markdown)
    └── P5-F6-MOM-2026Q2-v1/               # Cross-sample validation

---

Quick Start

On JoinQuant (run backtests)

Strategy environment (production, real costs):

# Paste results/P5-F2F5F6-40d-final-strategy.py → Run

Research environment (validation, cross-sample):

# Paste research/scripts/P5-F6-MOM-2026Q2-v1-segmented-standalone.py → Run

All scripts are self-contained — no local imports needed.

Locally (code review / IDE)

git clone https://github.com/IdealAuror/Cheap-Stable-Trending-quant.git
cd Cheap-Stable-Trending-quant
pip install -r requirements.txt

---

Experiment Summary

Phase	Factor / Topic	Verdict
P1-F1	EV<0 (net cash)	❌ Size proxy, abandoned
P1-F2	EP (earnings yield)	✅ Core alpha
P1-F3	Dividend yield	🟡 IC passed, post-break failure
P1-F4	ROE (quality)	❌ Negative alpha, abandoned
P1-F5	LowVol	⚠️ Auxiliary factor only
P2	Multi-factor synthesis	🟡 F2 sole alpha, F5 risk adjuster
P3	Stress testing	🟡 3/4 robustness checks passed
P4	Live calibration V1/V2	❌ M2/M5 failed Gate 5
P5	F6-MOM momentum	✅ Cross-sample PASS

---

Docs

Document	Audience
results/manual-investment-guide.html	Investors — HTML manual (recommended)
docs/CURRENT-STRATEGY.md	Everyone — strategy overview
research/_index.md	Researchers — factor dashboard
results/P5-F2F5F6-40d-final-strategy.py	Developers — production code

License

MIT · Research only · Not investment advice · Past performance ≠ future results