external/Quant.ai/README.md

metadata

title: Quant AI - Advanced Quant Trading Engine & Backtest Simulator
emoji: 📈
colorFrom: blue
colorTo: indigo
sdk: docker
app_port: 7860
pinned: false

Quant.ai - Advanced Quant Trading Engine & Backtest Simulator

Quant.ai is a production-ready, fully-automated stock trading engine and interactive backtest simulator. Built with a Python FastAPI backend and a React (Vite) frontend, it features real-time K-line pattern recognition, dynamic market regime routing, ATR-based risk sizing, and out-of-sample walk-forward optimization.

---

🚀 Key Features

1. 📊 Advanced K-Line Feature & Pattern Recognition

• 12 K-Line Numerical Features: Computes body ratio, upper/lower shadow ratios, gaps, relative volume (RVOL), and trend context dynamically.
• 22 Quantifiable Candlestick Patterns: Vectorized detection for patterns like Hammer, Shooting Star, Bullish/Bearish Engulfing, Piercing, Dark Cloud Cover, Morning/Evening Star, Three White Soldiers, Rising/Falling Three Methods, Gap Breakout, Exhaustion Gaps, and Neckline breakouts for W-Bottoms and M-Tops.

2. 🚦 Dynamic Market Regime Router

• Dynamically classifies the market into four regimes:
  • trend_up: Strong bullish trend. Activates trend-following strategies (Donchian breakout, EMA crossover).
  • trend_down: Bearish trend. Suspends buy operations and goes into defense.
  • high_volatility: Extreme volatility (ATR/Close in top 10%). Enforces cash preservation.
  • range_bound: Oscillating market. Activates mean reversion (Bollinger Bands oversold) and candlestick reversals.

3. 🛡️ Institutional-Grade Multi-Layer Risk Control

• ATR-Based Sizing: Calculates trade size based on account equity, ATR stop-distance, and risk percentage.
• Soft Drawdown Limit (7%) & Consecutive Losses (5): Triggers a 50% reduction in position size.
• Hard Drawdown Limit (12%): Temporarily locks the trading engine (risk multiplier goes to 0) to prevent capital blowups.

4. 🔄 Walk-Forward Parameter Optimization

• Features a rolling optimization pipeline (walk_forward.py) that divides history into training and test intervals.
• Optimizes parameters (strategy mode, ATR multiplier, RSI) by maximizing the drawdown-penalized net profit (Calmar-like metric) and validates performance out-of-sample.

5. 🌅 Market Open Focus & Opening Range Breakout (ORB) Strategy

• Market Open Focus Mode: Targets the high-volatility market opening (09:30 - 10:15 EST). Restricts buying to this high-momentum window and performs a force liquidation at 10:30 EST to protect capital from the midday choppy sideways trend.
• Opening Range Breakout (ORB): Precomputes the opening high and low from the first 5 minutes of regular hours (09:30 - 09:35 EST) and triggers high-probability breakout buys on high volume (RVOL > 1.2), using the opening range low as a hard failure stop-loss.

6. 🤖 AI Auto-Pilot Parameter Tuning (智能托管)

• Dynamically grid-searches strategy settings over the recent 5 days of 1-minute bar data for the selected ticker.
• Optimizes for the best risk-adjusted performance (Sharpe ratio and max drawdown mitigation) and automatically applies parameters to the active trading dashboard.

7. 🎬 Historical Replay Mode (历史开盘复盘模拟器)

• Granular 1m Simulation: Allows developers and traders to replay the market open sequence step-by-step for any trading day within the last 5 days.
• Interactive Controls: Supports Play, Pause, Single-Step tick progression, Reset, and speed tuning (with simulated intervals down to 50ms per bar).
• Synchronized Portfolio updates: Portfolio equity, cash, holdings, and transactions update dynamically on each step to observe execution points.

8. 🔍 Intraday Trade Inspector (日内成交精细化透视)

• High-Frequency Audit: In daily backtests, clicking any ledger row fetches the 1-minute candlestick data for the execution date and overlays the exact BUY/SELL orders at the market open (9:30 AM EST).
• Auto-scroll focus: In intraday/1m backtests, clicking any ledger row centers the main chart's time axis precisely on the selected transaction bar.

---

📁 Project Structure

├── backend/
│   ├── app/
│   │   ├── config.py           # Trade and risk configurations
│   │   ├── data_manager.py     # YFinance data loading, technical indicators, and regimes
│   │   ├── patterns.py         # 22 K-line patterns and W-Bottom/M-Top detection
│   │   ├── strategy.py         # Strategy routing and evaluation
│   │   ├── simulator.py        # Universal backtesting simulator
│   │   └── trading_engine.py   # Portfolio ledger, execution, and risk gates
│   ├── main.py                 # CLI Backtest interface
│   ├── main_api.py             # FastAPI REST Server
│   └── walk_forward.py         # Walk-Forward rolling optimization engine
├── frontend/                   # React Vite dashboard with TradingView charts
└── README.md

---

🛠️ Installation & Getting Started

Prerequisites

• Python 3.8+
• Node.js 16+

1. Backend Setup

Navigate to the root directory and install dependencies:

pip install pandas numpy yfinance fastapi uvicorn pydantic

Run a CLI backtest simulation:

# Run minute-level day trading simulation for TSLA
python backend/main.py --ticker TSLA --period 5d --interval 1m

# Run daily-level swing trading simulation for TSLA
python backend/main.py --ticker TSLA --period 1y --interval 1d

Run Walk-Forward rolling parameter optimization:

python backend/walk_forward.py --ticker TSLA --period 1y --interval 1d

Start the FastAPI API server:

python backend/main_api.py

2. Frontend Setup

Navigate to the frontend folder, install dependencies, and start the development server:

cd frontend
npm install
npm run dev

---

📊 Backtest Indicators & Performance

Our universal backtest simulator calculates standard trading metrics including:

• Net PnL & Return Percentage
• Max Account Equity Drawdown
• Win Rate & Round Trip Trade Count
• Transaction Commission and Slippage Friction Cost
• Market Regime Distributions

---

📝 License & Disclaimer

This software is provided for educational and research purposes only. Algorithmic trading carries substantial risk, and past performance is not indicative of future results.