Trading Agent — Automated Three-Stage Trader Learning Pipeline

Production-ready Python framework for autonomous trading agents implementing theory → backtest → paper test → live trading pipeline with continuous learning.

Features

• Event-driven backtester — Realistic OHLCV simulation with slippage, commissions
• Rule engine — Deterministic SMA-based entry/exit logic
• ML scoring model — Logistic regression probability scoring
• Paper broker — Order simulation and position tracking
• PostgreSQL journal — Complete event logging for analysis
• Analytics dashboard — Sharpe ratio, drawdown, win rate, calibration metrics

Quick Start

git clone https://github.com/AE707/trading-agent.git
cd trading-agent
pip install -r requirements.txt

# Run backtest
python src/backtester.py

# Run agent
python src/agent.py

# View analytics
python src/reporting.py

Configuration

Edit config/config.yaml:

trading:
  symbol: BTCUSDT
  risk_per_trade: 0.01
  max_drawdown: 0.15

db:
  uri: postgresql+psycopg2://postgres:postgres@localhost:5432/trading

Architecture

Data → Rules → Scoring Model → Broker → Journal → Analytics

Components

• rules.py — Entry/exit logic
• models.py — ML probability scoring
• broker.py — Order simulation
• journal.py — Event logging
• backtester.py — Historical simulation
• agent.py — Main orchestration
• reporting.py — Metrics & analysis

License

MIT