ldy26098
/
BMAD-METHOD
mirror of https://github.com/bmad-code-org/BMAD-METHOD.git
1
0
Fork 0
Code Packages Projects Releases Wiki Activity

docs: update README to reflect quant-focused approach 

Restructure README with expanded quant-specific content: key
capabilities (bias prevention, statistical rigor), named agent
personas, built-in safeguards, use case examples, project structure
overview, workflow paths as tables, and linked documentation resources.

https://claude.ai/code/session_01HioqMPX4pdZMqSnN3CDEVa
This commit is contained in:
Claude 2026-01-30 07:25:27 +00:00
parent 397d2be16d
commit 1fc477e51c
No known key found for this signature in database
1 changed files with 117 additions and 41 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

158
README.md
View File

@ -1,15 +1,26 @@
# Quant Method # Quant Method
> **v1.0.0-alpha.1** -- Early access. APIs and workflows are evolving.
**AI-Driven Investment Quant Development and Research Framework** -- Specialized AI agents and structured workflows for systematic strategy research, backtesting, risk management, and production deployment. **AI-Driven Investment Quant Development and Research Framework** -- Specialized AI agents and structured workflows for systematic strategy research, backtesting, risk management, and production deployment.
Built on the [BMAD Method](https://github.com/bmad-code-org/BMAD-METHOD), adapted for quantitative finance.
---
## Why Quant Method? ## Why Quant Method?
Quantitative investment research requires rigorous process discipline -- from hypothesis formation through statistical validation to production monitoring. Traditional tools leave gaps between research notebooks and production systems. Quant Method bridges this with AI agents that act as expert collaborators across the entire quant lifecycle. Quantitative investment research demands rigor at every stage -- from hypothesis formation through statistical validation to live monitoring. Traditional tools leave gaps between research notebooks and production systems. Quant Method fills those gaps with AI agents that act as expert collaborators across the entire quant lifecycle.
- **Structured Research Process**: Guided workflows grounded in quantitative finance best practices across research, strategy design, validation, and production ### Key Capabilities
- **Specialized Agents**: Domain experts including Quant Researcher, Portfolio Manager, Risk Analyst, Data Engineer, and more
- **Quant-Adaptive**: Adjusts depth based on strategy complexity -- a simple momentum factor needs different rigor than a multi-asset statistical arbitrage system - **Bias Prevention** -- Agents enforce point-in-time data correctness, flag look-ahead bias in vectorized operations, and require out-of-sample validation before any strategy advances
- **Full Lifecycle**: From alpha research through backtesting, risk analysis, and live monitoring - **Statistical Rigor** -- Walk-forward analysis, significance testing, and reproducibility requirements (random seeds, data versioning, parameter logging) are built into every workflow
- **Structured Research Process** -- Guided workflows grounded in quantitative finance best practices across research, strategy design, validation, and production
- **Quant-Adaptive Depth** -- Adjusts rigor based on strategy complexity: a simple momentum factor needs different treatment than a multi-asset statistical arbitrage system
- **Full Lifecycle Coverage** -- From alpha research and factor discovery through backtesting, risk decomposition, and live performance attribution
---
## Quick Start ## Quick Start
@ -21,44 +32,40 @@ npx quant-method install
Follow the installer prompts, then open your AI IDE (Claude Code, Cursor, Windsurf, etc.) in the project folder. Follow the installer prompts, then open your AI IDE (Claude Code, Cursor, Windsurf, etc.) in the project folder.
---
## Workflow Paths
### Rapid Strategy Path (Quick Flow) ### Rapid Strategy Path (Quick Flow)
Quick hypothesis testing, single-factor strategies, clear signals: For quick hypothesis testing, single-factor strategies, and clear signals:
1. `/strategy-spec` -- analyzes your data and produces a strategy specification with implementation tasks 1. **`/strategy-spec`** -- Analyze your data and produce a strategy specification with implementation tasks
2. `/dev-strategy` -- implements each task (signals, backtest, risk checks) 2. **`/dev-strategy`** -- Implement each task (signals, backtest, risk checks)
3. `/strategy-review` -- validates statistical rigor and code quality 3. **`/strategy-review`** -- Validate statistical rigor and code quality
### Full Research Path (Quant Method) ### Full Research Path (Quant Method)
Multi-factor strategies, portfolio-level research, production deployment: For multi-factor strategies, portfolio-level research, and production deployment:
1. `/research-brief` -- define investment thesis, universe, and data requirements | Step | Command | What Happens |
2. `/create-strategy-design` -- full specification with signal definitions, risk constraints, and performance targets |------|---------|-------------|
3. `/create-architecture` -- technical infrastructure: data pipelines, execution systems, monitoring | 1 | `/research-brief` | Define investment thesis, universe, and data requirements |
4. `/create-research-plan` -- break work into prioritized research and implementation tasks | 2 | `/create-strategy-design` | Full spec: signal definitions, risk constraints, performance targets |
5. `/research-planning` -- initialize research tracking | 3 | `/create-architecture` | Technical infrastructure: data pipelines, execution, monitoring |
6. **Repeat per task:** `/create-task` -> `/dev-task` -> `/task-review` | 4 | `/create-research-plan` | Break work into prioritized research and implementation tasks |
| 5 | `/research-planning` | Initialize research tracking and sequencing |
| 6 | Per task: `/create-task` | Prepare individual research or implementation tasks |
| 7 | Per task: `/dev-task` | Execute tasks (signal implementation, backtesting, risk checks) |
| 8 | Per task: `/task-review` | Validate statistical correctness and code quality |
## Specialized Agents ---
| Agent | Role | Focus | ## Research Lifecycle
|-------|------|-------|
| **Quant Researcher** | Alpha Research + Factor Analysis | Signal discovery, literature review, statistical analysis |
| **Portfolio Manager** | Portfolio Construction + Allocation | Position sizing, rebalancing, benchmark-aware optimization |
| **Quant Architect** | Systems Design + Infrastructure | Data pipelines, execution systems, backtesting frameworks |
| **Quant Developer** | Strategy Implementation | Signal code, backtest harnesses, production adapters |
| **Data Engineer** | Market Data + Alternative Data | Data pipelines, quality validation, feature engineering |
| **Risk Analyst** | Risk Management + Model Validation | Drawdown analysis, stress testing, regime detection |
| **Research Director** | Research Process + Coordination | Research pipeline management, prioritization, tracking |
| **Research Documentarian** | Research Reports + Model Documentation | Strategy documentation, research logs, compliance docs |
| **Strategy Developer** | Rapid Prototyping | Quick hypothesis testing, single-factor research |
## Workflow Phases
### Phase 1: Research ### Phase 1: Research
- Market and academic research - Market and academic literature review
- Factor discovery and screening - Factor discovery and screening
- Data exploration and alternative data evaluation - Data exploration and alternative data evaluation
- Investment thesis development - Investment thesis development
@ -66,31 +73,100 @@ Multi-factor strategies, portfolio-level research, production deployment:
### Phase 2: Strategy Design ### Phase 2: Strategy Design
- Signal specification and universe selection - Signal specification and universe selection
- Risk constraint definition - Risk constraint definition (drawdown limits, VaR/CVaR targets)
- Performance target setting - Performance target setting (Sharpe, information ratio, alpha decay)
- Model specification (statistical, ML, rules-based) - Model specification -- statistical, ML, or rules-based
### Phase 3: Validation ### Phase 3: Validation
- Backtesting with walk-forward analysis - Backtesting with walk-forward analysis
- Out-of-sample testing - Out-of-sample testing (mandatory, not optional)
- Statistical significance validation - Statistical significance validation
- Transaction cost and capacity analysis - Transaction cost, slippage, and capacity analysis
- Risk decomposition and stress testing - Risk decomposition and stress testing
- Regime-conditional performance review
### Phase 4: Production ### Phase 4: Production
- Deployment and integration - Deployment and system integration
- Live monitoring and alerting - Live monitoring and alerting
- Performance attribution - Performance attribution
- Research retrospective and strategy refinement - Research retrospective and strategy refinement
---
## Specialized Agents
Each agent brings domain-specific expertise to the research process:
| Agent | Persona | Role | Focus |
|-------|---------|------|-------|
| **Quant Researcher** | Elena | Alpha Research + Factor Analysis | Signal discovery, literature review, statistical analysis |
| **Portfolio Manager** | Catherine | Portfolio Construction + Allocation | Position sizing, rebalancing, benchmark-aware optimization |
| **Quant Architect** | Victor | Systems Design + Infrastructure | Data pipelines, execution systems, backtesting frameworks |
| **Quant Developer** | Marcus | Strategy Implementation | Signal code, backtest harnesses, production adapters |
| **Data Engineer** | Priya | Market Data + Alternative Data | Data pipelines, quality validation, feature engineering |
| **Risk Analyst** | Quinn | Risk Management + Model Validation | Drawdown analysis, stress testing, regime detection |
| **Research Director** | David | Research Process + Coordination | Pipeline management, prioritization, tracking |
| **Research Documentarian** | Paige | Research Reports + Documentation | Strategy docs, research logs, compliance records |
| **Strategy Developer** | Barry | Rapid Prototyping | Quick hypothesis testing, single-factor research |
### Built-In Safeguards
Agents enforce quant best practices throughout the workflow:
- **No look-ahead bias** -- Vectorized operations must respect temporal ordering
- **Point-in-time data correctness** -- Non-negotiable for backtesting pipelines
- **Reproducibility** -- Random seeds, data versioning, and parameter logging required
- **Out-of-sample validation** -- Every strategy must pass before advancing
- **Realistic assumptions** -- Transaction costs, slippage, and capacity constraints must be modeled
- **Risk metrics** -- Drawdown, VaR, CVaR, and regime-conditional analysis required
---
## What You Can Build
- **Single-factor momentum or value strategies** using the Quick Flow path
- **Multi-factor equity models** with statistical validation and risk decomposition
- **Statistical arbitrage systems** with cointegration testing and regime detection
- **Alternative data signal pipelines** with feature engineering and quality validation
- **Portfolio construction frameworks** with position sizing and rebalancing logic
- **Backtesting infrastructure** with walk-forward analysis and transaction cost modeling
---
## Project Structure
```text
src/
bmm/ Quant Method module (agents, workflows, data)
agents/ 9 specialized quant agents
workflows/ 4-phase research lifecycle + quick flow
teams/ Agent team configurations
data/ Templates and context files
core/ Core framework (shared agents, resources)
utility/ Shared utilities and components
tools/
cli/ Command-line interface
schema/ Agent schema validation
flattener/ Document processing tools
docs/ Tutorials, guides, and reference
```
---
## Documentation ## Documentation
- Getting Started Tutorial | Resource | Description |
- Strategy Research Walkthrough |----------|-------------|
- Backtesting Best Practices | [Getting Started](docs/tutorials/) | Installation and first strategy walkthrough |
- Risk Management Framework | [Quick Flow Guide](docs/explanation/quick-flow.md) | Rapid prototyping methodology |
| [Workflow Map](docs/reference/workflow-map.md) | Complete visual workflow reference |
| [Customization](docs/how-to/customize-bmad.md) | Adapting agents and workflows |
---
## Contributing ## Contributing