🏛️ Investment Masters Roundtable

37 AI-Powered Investment Masters • 3-Stage Debate System

14 Value Investors × 12 Technical Analysts × 11 Quantitative Signal Masters

Three-stage analysis with real-time market data → Roundtable debate minutes + Technical charts + Investment recommendations

Quick Start · Architecture · 37 Masters · 中文文档

---

✨ What Makes This Special

Imagine Warren Buffett, George Soros, and 35 other legendary investors sitting around a table, debating whether you should buy a stock — backed by real market data and powered by LLM reasoning.

Feature	Description
🏛️ Three-Stage Architecture	Fundamental debate → Technical trend analysis → Quantitative signal voting
⚔️ Real Adversarial Debate	Bull vs. Bear champions engage in multi-round LLM-driven debate — not just summaries
⚠️ Cross-Stage Conflict Detection	Flags when fundamentals and technicals disagree: "You are trading against Graham"
📊 Signal Matrix Voting	11 quantitative masters independently generate signals → weighted vote → net direction
📈 One-Click Visualization	Candlestick chart with MA/Bollinger/RSI/MACD overlays + support/resistance annotations
🌐 Multi-Market Support	US stocks, Chinese A-shares, Hong Kong stocks — via yfinance + akshare
🔌 Any LLM Backend	Works with GPT-4o, DeepSeek, Qwen, or any OpenAI-compatible API

---

🏗️ Architecture

flowchart TB
    subgraph INPUT["📥 Input"]
        USER["User: Stock Symbol + Market"]
    end

    subgraph DATA["⚙️ Data Engine"]
        direction LR
        OHLCV["OHLCV\nyfinance + akshare"]
        FUND["Fundamentals\nPE/PB/ROE/Revenue"]
        TECH["20+ Indicators\nMA/MACD/RSI/KDJ/BB/..."]
    end

    subgraph STAGE1["🏛️ Stage 1: Fundamental Debate (Weight: 60%)"]
        S1M["14 Investment Masters\nBuffett · Munger · Graham · Lynch\nMarks · Wood · Soros · Dalio\nDuan · Zhang · Li · Feng · Qiu · Lin"]
        S1P["7-Step Process\nIndependent → Camps → 2-Round Debate → Vote"]
    end

    subgraph STAGE2["📈 Stage 2: Technical Analysis (Weight: 20%)"]
        S2M["12 Technical Masters\nDow · Wyckoff · Elliott · Nison\nWilder · Appel · Granville · DeMark\nLane · Williams · Elder · ARBR"]
        S2P["Parallel Analysis\nTrend / Pattern / Volume-Price"]
    end

    subgraph STAGE3["⚡ Stage 3: Signal Voting (Weight: 20%)"]
        S3M["11 Signal Masters\nMACD · Bollinger · MA+OBV\nIchimoku · RSI/ADX · KDJ · CCI\nW%R · CMF · Triple Screen · Chaos"]
        S3P["Signal Matrix → Conflict Debate → Weighted Vote"]
    end

    subgraph OUTPUT["📊 Output"]
        MD["📝 Markdown Report"]
        HTML["🌐 HTML Report"]
        CHART["📈 Technical Chart"]
    end

    USER --> DATA
    DATA --> STAGE1
    STAGE1 -->|"consensus + dissent"| STAGE2
    STAGE2 -->|"trend confirmation"| STAGE3
    STAGE3 --> CROSS["⚠️ Cross-Stage\nConflict Detection"]
    CROSS --> AGG["🎯 60/20/20\nWeighted Aggregation"]
    AGG --> OUTPUT

    style INPUT fill:#1e293b,stroke:#3b82f6,color:#e2e8f0
    style DATA fill:#1e293b,stroke:#f97316,color:#e2e8f0
    style STAGE1 fill:#1e293b,stroke:#22c55e,color:#e2e8f0
    style STAGE2 fill:#1e293b,stroke:#eab308,color:#e2e8f0
    style STAGE3 fill:#1e293b,stroke:#ef4444,color:#e2e8f0
    style OUTPUT fill:#1e293b,stroke:#8b5cf6,color:#e2e8f0

Loading

---

🚀 Quick Start

# 1. Clone
git clone https://github.com/LucasYanzy/investment-masters-roundtable.git
cd investment-masters-roundtable

# 2. Install dependencies
pip install -r requirements.txt

# 3. Configure API key
cp .env.example .env
# Edit .env → set LLM_API_KEY (supports OpenAI / DeepSeek / Qwen)

# 4. Analyze any stock
python run.py --symbol AAPL --market US       # Apple (US)
python run.py --symbol 000001 --market CN     # Ping An Bank (China A-share)
python run.py --symbol 0700.HK --market HK    # Tencent (Hong Kong)

# 5. Launch Web UI
python run.py --web

📋 Output

Each analysis produces:

• 📝 {symbol}_report.md — Full roundtable debate minutes
• 🌐 {symbol}_report.html — Dark-themed interactive HTML report
• 📈 {symbol}_chart.png — Candlestick + indicators chart

---

🤖 37 Masters Across 3 Stages

🏛️ Stage 1: 14 Investment Masters (Qualitative Debate)

#	Master	School	Core Framework
1	Warren Buffett	Value Investing	Moat, margin of safety, long-term holding
2	Charlie Munger	Value Investing	Mental models, inversion thinking
3	Benjamin Graham	Value Investing	Intrinsic value, Mr. Market, safety margin
4	Peter Lynch	Growth Investing	PEG valuation, invest in what you know
5	Howard Marks	Contrarian	Second-level thinking, cycle positioning
6	Cathie Wood	Disruptive Innovation	ARK innovation framework
7	George Soros	Macro Hedge	Reflexivity theory, trend trading
8	Ray Dalio	Macro Hedge	Economic machine, debt cycles, All Weather
9	Duan Yongping 段永平	Pragmatist	Business model first, "ben fen" philosophy
10	Zhang Lei 张磊	Growth Investing	Structural value, dynamic moats
11	Li Lu 李录	Value Investing	Circle of competence, Civilization 3.0
12	Feng Liu 冯柳	Contrarian	Weak-form system, odds thinking
13	Qiu Guolu 邱国鹭	Value Investing	"Count moons, not stars"
14	Lin Yuan 林园	Consumer Investing	Monopoly brands, "mouth stocks"

📈 Stage 2: 12 Technical Analysis Masters (Trend / Pattern / Volume)

#	Master	Core Theory	Focus Area
1	Charles Dow	Dow Theory	Primary / secondary / minor trends
2	Richard Wyckoff	Wyckoff Method	Volume-price, accumulation/distribution
3	Ralph Elliott	Wave Theory	Wave counting, Fibonacci
4	Steve Nison	Japanese Candlesticks	Candlestick pattern recognition
5	J. Welles Wilder	RSI/DMI/ADX/ATR	Trend strength, volatility
6	Gerald Appel	MACD	Moving average convergence divergence
7	Joseph Granville	Volume 8 Rules	OBV volume-price validation
8	Thomas DeMark	TD Sequential	Precise buy/sell points
9	George Lane	Stochastic/KDJ	Overbought/oversold zones
10	Larry Williams	Williams %R	Extreme price levels
11	Alexander Elder	Triple Screen	Multi-timeframe analysis
12	ARBR Team	Sentiment Indicators	Market sentiment gauges

⚡ Stage 3: 11 Quantitative Signal Masters (Pure Numerical Signals + Debate)

#	Master	Indicators	Signal Output
1	Gerald Appel	MACD	Golden/death cross + target levels
2	John Bollinger	Bollinger Bands	Band breakout/reversion
3	Joseph Granville	MA + OBV	Moving average cross + volume validation
4	Goichi Hosoda	Ichimoku Cloud	Cloud support/resistance
5	J. Welles Wilder	RSI/ADX/ATR/PSAR	Overbought/oversold + trend strength
6	George Lane	KDJ	%K/%D crossover
7	Donald Lambert	CCI	Channel breakout
8	Larry Williams	W%R	Overbought/oversold / divergence
9	Marc Chaikin	CMF	Money flow direction
10	Alexander Elder	Triple Screen	Multi-timeframe signals
11	Bill Williams	Chaos Trading	Alligator/Fractal/AO/AC

---

📁 Project Structure

investment-masters-roundtable/
├── run.py                        # CLI entry + Web launcher
├── config.py                     # Global config (env-driven)
│
├── data_engine/                  # Data layer
│   ├── fetcher.py                #   OHLCV (yfinance + akshare)
│   ├── fundamental_data.py       #   Fundamentals (valuation/financials)
│   ├── technical_indicators.py   #   20+ technical indicators
│   ├── cache.py                  #   2-level cache (LRU + file)
│   └── schema.py                 #   Unified data structures
│
├── masters/                      # 37 masters analysis layer
│   ├── base_master.py            #   Base class + LLM call + MasterOpinion
│   ├── registry.py               #   37-master registry
│   ├── investment_master.py      #   Stage 1: 14 investment masters
│   ├── technical_master.py       #   Stage 2: 12 technical masters
│   ├── signal_master.py          #   Stage 3: 11 signal masters
│   └── skill_references/         #   Master frameworks (.md)
│       ├── investment_masters/   #     14 value/growth frameworks
│       ├── technical_masters/    #     12 technical analysis frameworks
│       └── signal_masters/       #     11 quantitative signal frameworks
│
├── debate_orchestrator/          # Debate coordination layer
│   ├── orchestrator.py           #   Main orchestrator (3-stage)
│   ├── stage1_debate.py          #   Stage 1: 7-step debate
│   ├── stage2_analysis.py        #   Stage 2: parallel analysis
│   ├── stage3_signals.py         #   Stage 3: signal matrix + debate
│   ├── cross_stage_conflict.py   #   Cross-stage conflict detection
│   └── aggregator.py             #   60/20/20 weighted aggregation
│
├── report_generator/             # Report generation layer
│   ├── markdown_report.py        #   Markdown roundtable minutes
│   ├── html_report.py            #   Dark-themed HTML report
│   └── chart_renderer.py         #   mplfinance technical charts
│
├── web_app/                      # Web interface (Flask)
│   ├── app.py                    #   Flask application
│   └── templates/index.html      #   Dark-themed SPA
│
├── tests/                        # Tests
│   └── test_smoke.py             #   Smoke tests
│
├── docs/                         # Documentation
│   └── architecture.md           #   System architecture
│
└── examples/                     # Sample outputs

---

🔧 Tech Stack

Layer	Technology
Data Fetching	yfinance (US/HK) + akshare (China A-shares)
Technical Indicators	pandas-ta + pure pandas/numpy fallback
LLM Integration	OpenAI-compatible API (GPT-4o / DeepSeek / Qwen)
Concurrency	asyncio (parallel within stage, serial between stages)
Visualization	mplfinance + matplotlib
Web Interface	Flask + dark-themed SPA
Logging	Python logging with --verbose mode

---

⚙️ Configuration

Copy .env.example → .env:

# LLM (any OpenAI-compatible API)
LLM_API_BASE=https://api.openai.com/v1
LLM_API_KEY=sk-xxx
LLM_MODEL=gpt-4o

# Data
DEFAULT_PERIOD=2y          # Historical data length
FMP_API_KEY=               # Optional: financialmodelingprep

# Stage Weights (must sum to 1.0)
STAGE1_WEIGHT=0.6          # Fundamentals
STAGE2_WEIGHT=0.2          # Technical confirmation
STAGE3_WEIGHT=0.2          # Signal timing

---

📈 Performance

Metric	Value
LLM Calls	37 base + up to 6 debate rounds ≈ ~43 calls
Analysis Time	30–120 seconds (depends on LLM response speed)
Concurrency	All masters parallel within each stage (14 / 12 / 11)
Cache TTL	Technical data: 1h, Fundamentals: 24h

---

🤝 Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'feat: add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

---

📄 License

MIT © 2026 Lucas Yan

---

⭐ If this project helps you, please give it a Star!

中文文档 →