Statistical Arbitrage Trading Strategy

Description

This project implements a statistical arbitrage trading strategy using cointegrated stock pairs. By leveraging linear regression and Z-scores, the algorithm identifies mean-reverting opportunities and generates trading signals. Backtested on historical stock data, the portfolio grew from $24,050 to $25,489.50 over a 2-year period.

Features

• Pair Selection:

  Identifies cointegrated pairs using the Engle-Granger method. Validates pairs with a 4-year in-sample and 2-year out-of-sample analysis.

• Trading Algorithm:

  Uses Z-scores of spreads for entry/exit signals. Implements long-short trades based on mean-reversion.

• Performance Metrics:

  Portfolio growth tracked with key metrics such as total return and Sharpe ratio.

  Installation

1. Clone the repository:

bash
git clone https://github.com/yourusername/statistical-arbitrage.git
cd statistical-arbitrage

2. Install dependencies:

bash
pip install -r requirements.txt

Usage

1. Data Preparation:

   Use Yahoo Finance or other APIs to download historical price data for stocks.

2. Run Pair Selection:

   Identify and validate cointegrated pairs using the provided scripts.

3. Backtesting:

   Simulate trades based on Z-score thresholds and track portfolio performance.

4. Visualization:

   View spread plots, Z-scores, and portfolio growth over time.

Project Structure

bash
├── main.py               # Main script for pair selection and backtesting
├── functions.py          # Helper functions for cointegration, Z-scores, and trading logic
├── data/                 # Folder for historical stock data
├── screenshots/          # Output visuals for Z-scores and portfolio growth
│   ├── zscore_plot.jpg
│   ├── portfolio_growth.jpg
├── requirements.txt      # Project dependencies
├── LICENSE               # License information
├── README.md             # Project documentation (this file)

Key Results

Initial Portfolio Value: $24,050 Final Portfolio Value: $25,000 Time Frame: 2 Years Notable Metrics: Sharpe Ratio: Win Rate:

Dependencies

• Python 3.8+

• Libraries:

  pandas numpy matplotlib seaborn statsmodels yfinance