algoMLTrading

A modular algorithmic Paper trading system for equities using a combination of machine learning, technical strategies, and risk management. Supports historical backtesting and live/paper trading via Alpaca API.

Features

• Fetches and processes live or historical data (via yfinance and Alpaca)
• Calculates technical indicators and fundamental data
• Trains and uses an XGBoost model for directional prediction
• Supports strategies: Mean Reversion, Momentum, Statistical Arbitrage, and Combined
• Risk-managed position sizing and loss limits
• Modular backtester and live trading loop

Limitations

• Performance: Current backtests show negative Sharpe ratio and drawdowns.
• Signal Quality: Combined strategy often produces unprofitable trades.
• Machine Learning: Functional but needs tuning and better features.
• Execution: No real trades are placed unless performance thresholds are passed.
• Pair Arbitrage: Only tested on AAPL and BA.

Still Needed

• Improve feature engineering and model selection
• Tune XGBoost hyperparameters
• Broaden to more tickers
• Log executed trades in live mode
• Refactor signal-weight logic in CombinedStrategy
• Add unit tests

Install

Install all required dependencies:

pip install -r requirements.txt

Output