Your First Backtest¶

Run your first strategy backtest in under 5 minutes.

Prerequisites¶

• TradAI workspace set up (just setup completed)
• Exchange API keys configured (optional, for live data)

Quick Start¶

1. Start Services¶

# Start all services (Backend, Data Collection, Strategy Service, MLflow)
just up

# Verify services are running
just docker ps

Expected output:

NAME                STATUS          PORTS
tradai-backend      Up             0.0.0.0:8000->8000/tcp
tradai-data         Up             0.0.0.0:8002->8002/tcp
tradai-strategy     Up             0.0.0.0:8003->8003/tcp
tradai-mlflow       Up             0.0.0.0:5001->5001/tcp

2. Use Sample Data or Sync Fresh Data¶

Option A: Use Pre-bundled Sample Data (Fastest)

Sample market data is included in the repository for quick testing:

user_data/data/binance/futures/
├── BTC_USDT_USDT-1h-futures.feather
├── BTC_USDT_USDT-4h-futures.feather
├── ETH_USDT_USDT-1h-futures.feather
├── SOL_USDT_USDT-1h-futures.feather
└── ... (multiple timeframes)

This data allows you to run backtests immediately without syncing from an exchange.

Option B: Sync Fresh Data from Exchange

To get the latest data or custom date ranges:

# Sync BTC/USDT futures data for January 2024
tradai data sync BTC/USDT:USDT \
  --start 2024-01-01 \
  --end 2024-01-31 \
  --timeframe 1h

# Verify data is available
tradai data check-freshness BTC/USDT:USDT

3. Run a Backtest¶

Option A: Quick Local Backtest (No Services Required)

For rapid local development without starting any services:

# Quick backtest using freqtrade directly (no backend required)
tradai backtest quick PascalStrategy --local

# With custom period and symbol
tradai backtest quick PascalStrategy --local --period 7d --symbol ETH/USDT:USDT

This runs freqtrade directly as a subprocess, perfect for: - Quick strategy validation during development - Testing when services are not running - CI/CD pipelines

Option B: Full Backtest via Backend Service

For production-quality backtests with full metrics and tracking:

# Start services first
just up

# Run backtest via backend service
tradai backtest quick PascalStrategy --period 30d

# Or use the strategy backtest command for more options
tradai strategy backtest \
  --strategy PascalStrategy \
  --start 2024-01-01 \
  --end 2024-01-31 \
  --symbol BTC/USDT:USDT \
  --timeframe 1h

Example output:

=== Backtest Results ===
Strategy: PascalStrategy
Period: 2024-01-01 to 2024-01-31
Timeframe: 1h
Symbols: BTC/USDT:USDT

=== Performance ===
Total Trades: 15
Winning: 9 | Losing: 6
Win Rate: 60.0%
Total Profit: $127.50 (12.75%)

=== Risk Metrics ===
Sharpe Ratio: 1.45
Sortino Ratio: 2.10
Calmar Ratio: 0.95
Max Drawdown: -3.25%
Profit Factor: 1.85

4. View in MLflow (Optional)¶

Track experiments with MLflow:

# Run backtest with MLflow tracking
tradai strategy backtest \
  --strategy PascalStrategy \
  --start 2024-01-01 \
  --end 2024-01-31 \
  --symbol BTC/USDT:USDT \
  --experiment "my-first-backtest"

Open MLflow UI at http://localhost:5001 to view: - Run metrics (Sharpe, Sortino, Calmar, etc.) - Parameters (stoploss, stake amount) - Artifacts (trade logs)

Using Just Commands¶

For convenience, the justfile provides shortcuts:

# Run backtest for specific strategy
just run-backtest PascalStrategy 2.0.0

# Start strategy service in dev mode
just run-strategy

Backtest Configuration Options¶

Quick Backtest (tradai backtest quick)¶

Full Backtest (tradai strategy backtest)¶

Common Issues¶

"No data available for symbol"¶

Sync the data first:

tradai data sync BTC/USDT:USDT --start 2024-01-01 --end 2024-01-31

"Strategy not found"¶

List available strategies:

tradai strategy list-strategies

"Service connection refused"¶

Ensure services are running:

just up
just docker ps

Next Steps¶

• Create a New Strategy - Build your own strategy
• Collect Market Data - Sync more trading pairs