Quantitative Trading Platform

A comprehensive quantitative trading platform for strategy development, backtesting, and parameter optimization.

Tech Stack

• Backend: FastAPI + Backtrader + SQLite + Optuna (optional)
• Frontend: Next.js 14 + TypeScript + Tailwind CSS
• Data Source: Binance public API

Features

Core Features

• Historical data management
• Strategy development and testing
• Backtesting engine with comprehensive metrics
• Parameter optimization

Advanced Optimization Features (New)

• Bayesian Optimization: Efficient parameter search using Optuna
• Composite Scoring: Multi-metric optimization with custom weights
• Stability Analysis: Detect overfitting through parameter variance analysis
• Out-of-Sample Testing: Validate parameter generalization on unseen data

Project Structure

backend/
├── core/                    # Core business logic
│   ├── backtest_runner.py  # Shared backtest execution
│   ├── bayesian_optimizer.py  # Bayesian optimization (Optuna)
│   ├── stability_analyzer.py  # Parameter stability analysis
│   ├── scoring_functions.py  # Composite scoring
│   └── optimizer.py         # Grid search optimizer
├── models/                  # Data models
│   ├── optimization_job.py  # Extended with new fields
│   └── optimization_result.py  # Extended with composite score
├── api/                     # FastAPI endpoints
│   └── backtest.py          # Extended optimization endpoints
├── utils/                   # Utility functions
├── observers/               # Backtrader observers
└── strategies/              # Trading strategies

frontend/                  # Next.js frontend
├── components/
│   ├── OptimizationForm.tsx  # Extended with new options
│   └── OptimizationResults.tsx  # Extended with stability display
└── lib/api/
    └── optimization.ts      # Extended API types

data/                      # SQLite database storage

tests/                     # Test files
├── test_scoring_functions.py
├── test_bayesian_optimizer.py
├── test_stability_analyzer.py
├── test_optimization_api.py
└── test_optimization_integration.py

logs/                      # Application logs

Getting Started

Backend Setup

cd backend
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
pip install -r requirements.txt

# Optional: Install Optuna for Bayesian optimization
pip install optuna>=3.5.0

Frontend Setup

cd frontend
pnpm install
pnpm dev

Development

• Backend runs on http://localhost:8000
• Frontend runs on http://localhost:3000

Usage

Parameter Optimization Methods

1. Grid Search (Default)

Traditional exhaustive search over parameter combinations:

{
  "strategy_name": "MovingAverageCrossover",
  "symbol": "BTCUSDT",
  "interval": "1h",
  "start_time": "2024-01-01T00:00:00",
  "end_time": "2024-01-31T23:59:59",
  "parameter_ranges": {
    "fast_period": [10, 20, 30],
    "slow_period": [30, 40, 50]
  }
}

2. Bayesian Optimization (Recommended)

Intelligent parameter search using Optuna:

{
  "strategy_name": "MovingAverageCrossover",
  "symbol": "BTCUSDT",
  "interval": "1h",
  "start_time": "2024-01-01T00:00:00",
  "end_time": "2024-01-31T23:59:59",
  "optimization_method": "bayesian",
  "n_trials": 100,
  "parameter_ranges": {
    "fast_period": {"type": "int", "min": 5, "max": 50},
    "slow_period": {"type": "int", "min": 20, "max": 100}
  }
}

Benefits:

• More efficient than grid search
• Automatically learns from previous trials
• Better for high-dimensional parameter spaces

Composite Scoring

Optimize using multiple metrics simultaneously:

{
  "scoring_weights": {
    "sharpe_ratio": 0.4,
    "total_return": 0.3,
    "max_drawdown": -0.2,
    "win_rate": 0.1
  }
}

Default weights:

• Sharpe Ratio: 40%
• Total Return: 30%
• Max Drawdown: -20% (lower is better)
• Win Rate: 10%

Stability Analysis

Automatically detect overfitting by testing parameter robustness:

{
  "enable_stability_analysis": true
}

What it does:

• Tests parameter variations around optimal values
• Calculates stability score (0-1, higher is more stable)
• Identifies "parameter plateaus" vs isolated peaks
• Helps distinguish robust parameters from overfitted ones

Stability Report:

{
  "stability_score": 0.85,
  "variance": 0.02,
  "is_stable": true,
  "neighbors": [
    {"parameters": {"fast_period": 19}, "score": 0.82},
    {"parameters": {"fast_period": 21}, "score": 0.83}
  ]
}

Out-of-Sample Testing

Validate parameters on unseen data to detect overfitting:

{
  "enable_out_of_sample": true,
  "test_start_time": "2024-02-01T00:00:00",
  "test_end_time": "2024-02-28T23:59:59"
}

Benefits:

• Detects overfitting to training period
• Validates parameter generalization
• Provides realistic performance expectations

API Endpoints

Optimization

• POST /api/v1/backtest/optimize - Run parameter optimization
• GET /api/v1/backtest/optimization/jobs/{id}/results - Get optimization results
• GET /api/v1/backtest/optimization/jobs/{id}/stability - Get stability report (NEW)
• GET /api/v1/backtest/optimization/scoring-weights - Get default scoring weights (NEW)

Best Practices

1. Use Bayesian optimization for complex parameter spaces
2. Enable stability analysis to detect overfitting
3. Use custom scoring weights to match your investment objectives
4. Always use out-of-sample testing for validation
5. Monitor stability scores - aim for > 0.7

Preventing Overfitting

Overfitting in parameter optimization occurs when parameters are tuned too closely to historical data. Use these techniques:

1. Stability Analysis: Check if small parameter changes cause large performance drops
2. Out-of-Sample Testing: Validate on data not used in optimization
3. Walk-Forward Analysis: Use rolling windows for time series
4. Parameter Constraints: Limit parameter ranges to sensible values
5. Composite Scoring: Don't optimize a single metric in isolation

Testing

# Backend tests
pytest tests/ -v

# With coverage
pytest --cov=backend tests/

# Frontend tests
cd frontend && pnpm test:e2e

License

MIT