价值投资导向的 26 天深度上市公司研究方法论 A 26-Day Systematic Equity Research Methodology for Value Investors
Deep Stock Research 是一套价值投资导向的系统化上市公司研究方法论，通过 26 天（Day01–Day26） 的结构化深度研究，对任意 A 股 / 港股 / 美股公司做一次"穿透到骨头里"的全景体检。不荐股，只帮你看清数据背后的真实信号。
- 买股票就是买公司——以生意视角而非交易视角看待每一笔投资
- 事实底是估值地基——所有估值必须建立在可核查的事实之上（neil-fcff 工序）
- 护城河是唯一安全边际——没有护城河的"便宜"是陷阱
- 一美元法则——每留存 1 元利润，是否至少创造 1 元市值？这是管理层的终极试金石
- 数字会说真话——但前提是你愿意直面它们
- 管理层是最大变量——再好的生意配差管理层也是价值毁灭
- 宁缺毋滥——看不懂的公司不投，等得起
基础研究 Day01–Day20（必读，建立基础认知）
| 层级 | Day | 主题 |
|---|---|---|
| 起点层 | Day01–08 | 商业模式、历史、业务拆解、竞争格局 |
| 数字层 | Day09–11 | 财务分析、分红、管理层 |
| 资本配置层 | Day12–14 | 护城河深度、五大去向、一美元法则 |
| 判断层 | Day15–17 | neil-fcff 事实底、配倍数、估值风险 |
| 收束层 | Day18–20 | 投资结论、心智模型校验、跟踪决策协议 |
高级工具箱 Day21–Day26（可选，按需调用）
| Day | 主题 | 理论源头 | 工具箱说明 |
|---|---|---|---|
| Day21 | 一美元法则 | Buffett 1983 致股东信 | 累计资本回报率测试（严格公式） |
| Day22 | 红皇后效应 | Van Valen 1973 + Ridley 1999 | 制造业竞争动态学 |
| Day23 | 反脆弱 | Taleb 2012《Antifragile》 | 压力测试 + 脆弱性扫描 |
| Day24 | 隐含期权 | Black-Scholes 1973 + CRR 1979 | 实物期权估值 |
| Day25 | Munger 反向思考 | Munger 1994 USC 演讲 + 2005《Poor Charlie's Almanack》 | 失败学 + Stop Doing List |
| Day26 | Hold or Fold 决策树 | Kelly 1956《A New Interpretation of Information Rate》 | 凯利公式仓位管理 |
将本仓库 clone 到你的 agent workspace 的 skills/ 目录下：
cd ~/.qclaw/workspace-xxx/skills/
git clone https://github.com/zhujianneil/deep-stock-research.git或作为 OpenClaw skill 安装：
skillhub install deep-stock-research- 「研究 XX」「深度研究 XX」「XX 公司分析」
- 「帮我分析 XX」「XX 的护城河」「XX 的估值」
- 「XX 的 owner earnings / 自由现金流」
- 「XX 的管理层 / 资本配置 / 一美元法则」
| Skill | 作用 | 是否必需 |
|---|---|---|
| neodata-financial-search | 实时金融数据查询（行情/财报/资金流向/研报评级） | 必需 |
| neil-fcff | owner earnings 提取器（OCF 瀑布 → 维护性 capex 拆分 → 红旗扫描） | 强烈推荐（v3.0 Day15 配套） |
| moat-multiple-translator | 给真现金配护城河倍数 | 强烈推荐（v3.0 Day16 配套） |
- 每篇 Day 文件 ≥ 5,000 中文字（不含 frontmatter 与表格）
- 任何 < 5,000 字的 Day 禁止提交
- 每个核心结论必须有 ≥ 2 个独立数据源交叉验证
- 输出语言：中文
| 序号 | 公司 | 行业 | 完成情况 | 数据源 |
|---|---|---|---|---|
| 001 | 老铺黄金（06181.HK） | 黄金珠宝 | Day01–Day14 ✅ | research-briefs/ |
| 007 | 三花智控（002050.SZ） | 汽车零部件 | Day01–Day26 ✅ | v3.0 26 天完整版方法论标本 |
提示：已完成案例通常涉及作者的私人投资观察，不在本仓库内。请用本方法论对自己的标的做独立研究。
deep-stock-research/
├── SKILL.md # 完整方法论文档（v3.0，89KB）
├── SKILL.md.v1.bak # v1.0 14 天版本（备份）
├── SKILL.md.v2.bak # v2.0 20 天版本（备份）
├── research-briefs/ # 研究背景包（公司基本面）
│ ├── 三花智控.md
│ └── 宁德时代.md
├── README.md # 本文件
├── CHANGELOG.md # 版本演进记录
├── LICENSE # MIT 许可证
└── .gitignore
Deep Stock Research is a systematic equity research methodology for value investors. It takes 26 days of structured analysis (Day01–Day26) to deeply investigate any public company (A-shares / HK / US), producing a complete "full-body scan" of the business.
"Never recommend a stock — only help you see the data behind it."
- Buying a stock is buying a company — view every investment as a business, not a trade.
- The fact base is the foundation of valuation — every multiple must rest on auditable facts (via
neil-fcff). - Moat is the only margin of safety — "cheap" without moat is a trap.
- The $1 Test — does each dollar of retained earnings create at least $1 of market value? The ultimate yardstick of management.
- Numbers tell the truth — provided you face them directly.
- Management is the biggest variable — even a great business with bad capital allocation destroys value.
- When in doubt, pass — don't invest in what you don't understand; you can always wait.
Foundation (Day01–Day20) — Required for every study
| Layer | Days | Topic |
|---|---|---|
| Origin | Day01–08 | Business model, history, segment breakdown, competitive landscape |
| Numbers | Day09–11 | Financial analysis, dividends, management |
| Capital allocation | Day12–14 | Moat depth, five uses of cash, $1 retained-earnings test |
| Judgment | Day15–17 | neil-fcff fact base, multiple assignment, valuation risk |
| Closing | Day18–20 | Investment thesis, master-mindset review, tracking & decision protocol |
Advanced Toolkit (Day21–Day26) — Optional, use as needed
| Day | Topic | Theoretical Source |
|---|---|---|
| Day21 | $1 Retained-Earnings Test | Buffett 1983 shareholder letter |
| Day22 | Red Queen Effect | Van Valen 1973 + Ridley 1999 |
| Day23 | Antifragility | Taleb 2012, Antifragile |
| Day24 | Embedded Real Options | Black-Scholes 1973 + Cox-Ross-Rubinstein 1979 |
| Day25 | Munger's Inversion | Munger 1994 USC speech + 2005 Poor Charlie's Almanack |
| Day26 | Hold-or-Fold Decision Tree | Kelly 1956, A New Interpretation of Information Rate |
cd ~/.qclaw/workspace-xxx/skills/
git clone https://github.com/zhujianneil/deep-stock-research.gitOr via the OpenClaw skill hub:
skillhub install deep-stock-research| Skill | Role | Required |
|---|---|---|
neodata-financial-search |
Real-time financial data queries (quotes, financials, fund flows, ratings) | Yes |
neil-fcff |
Owner-earnings extractor (OCF waterfall → maintenance-capex split → red-flag scan) | Strongly recommended (powers v3.0 Day15) |
moat-multiple-translator |
Moat-coefficient × multiple translator | Strongly recommended (powers v3.0 Day16) |
- Each Day file ≥ 5,000 Chinese characters (excluding frontmatter and tables)
- Any Day < 5,000 chars is not eligible for submission
- Every conclusion must be cross-validated by ≥ 2 independent data sources
- Output language: Chinese (English summary above for international readers)
v3.0 adds a 6-day "philosophy + math" toolkit (Day21–Day26) on top of v2.0's 20-day foundation. Each new Day corresponds to a canonical academic / intellectual source — from Buffett's 1983 retained-earnings test to Kelly's 1956 optimal-betting formula. Together they form a four-layer defense against self-deception:
Layer 1 — Facts (Day15 neil-fcff)
Layer 2 — Judgment (Day16 moat-multiple + Day12 moat depth)
Layer 3 — Master Mindset (Day19 Buffett/Munger)
Layer 4 — Philosophy + Math (Day21-Day26)
MIT — see LICENSE.
Issues and PRs welcome. For substantial changes (new Day, new methodology), please open an issue first to discuss.
If this skill helps your research, please ⭐ the repo. It tells the author the methodology is useful and motivates continued development.
zhujianneil — value investor, equity research enthusiast, lifelong student of Buffett & Munger.
本方法论的所有内容均为教育性质，不构成任何投资建议。投资有风险，决策需谨慎。
All content is for educational purposes only and does not constitute investment advice. Investing involves risk; decisions are your own responsibility.