stat-arb-engine

Statistical Arbitrage Engine built from scratch using Python.

This project demonstrates my interest in quantitative finance and statistical modeling by implementing a pair trading strategy —- from data acquisition and pair selection to signal generation, backtesting, and visualization. It applies mathematical techniques like cointegration, z-score normalization, and time series analysis to financial market data.

---

What It Does

This engine identifies cointegrated pairs and simulates a mean-reversion trading strategy using the spread between their prices.

Key Features:

• Collects and preprocesses stock data using yfinance.
• Uses Engle-Granger cointegration tests to identify statistically viable pairs.
• Constructs a spread between the two assets and generates trading signals based on Z-score thresholds.
• Backtests the strategy with configurable parameters and visualizes the spread with entry/exit markers.

---

Core Concepts

• Cointegration Testing for identifying long-term relationships between assets.
• Z-Score-based Signal Generation for statistical entry/exit rules.
• Backtesting with trade tracking, P&L logging, and optional slippage parameters.
• Data Cleaning, Resampling, and Visualization using Python’s data stack.

---

🗂 Project Structure

File	Description
main.py	Orchestrates the workflow: fetch → analyze → signal → backtest → plot
fetch_data.py	Downloads OHLC data using yfinance
cointegration.py	Computes and filters statistically cointegrated pairs
signals.py	Creates entry/exit signals based on Z-score of spread
backtester.py	Executes trades, tracks P&L, logs strategy performance
pair_analysis.py	Filters and scores pairs based on correlation and stationarity
plotter.py	Visualizes spread and signals on a matplotlib chart
streamlit_app.py	Interactive Streamlit dashboard for parameter tuning and visualization

---

📦 Requirements

Install dependencies using:

pip install -r requirements.txt

requirements.txt includes:

• numpy
• pandas
• matplotlib
• statsmodels
• yfinance
• streamlit

---

🚀 How to Run

python main.py

You can customize:

• The ticker universe (e.g., S&P 500 components)
• Start/end date
• Cointegration p-value threshold
• Z-score entry/exit rules
• Lookback window

---

🌐 Streamlit Dashboard

Launch the interactive dashboard:

streamlit run streamlit_app.py

Customize in the sidebar:

• Tickers (X & Y)
• Start/End Date
• Entry/Exit Z-score thresholds
• Rolling window size
• Notional value per trade

The dashboard outputs:

• Cointegration test results
• Spread & z-score plots
• Trade signals and timing
• Realistic cumulative PnL in terms of dollars

---

📊 Example Output

• Green ▲ = LONG entry
• Red ▼ = SHORT entry
• Blue ● = CLOSE (mean reversion)

---