This repo contains the code for an interactive web application that allows users to explore, simulate, and visualize a market-neutral statistical arbitrage strategy based on cointegration analysis between stock pairs.
The application is built using Streamlit, providing an intuitive and educational interface for understanding and evaluating pairs trading strategies. Users can dynamically select stock pairs, adjust trading logic parameters, and view detailed backtesting results, including strategy performance metrics and visualizations.
- Help users identify cointegrated stock pairs using rigorous statistical tests (Engle-Granger, Johansen).
- Construct mean-reverting spread portfolios based on a dynamically estimated hedge ratio (OLS or rolling).
- Let users interactively configure and simulate a pairs trading backtest, including:
- Z-score thresholds (entry/exit)
- Maximum holding period
- Take-profit and stop-loss exits
- Cooldown periods between trades
- Transaction cost / slippage estimates
- Provide educational context via tooltips and expandable explanations for all core strategy components.
- Allow users to compare cumulative returns of the strategy against the individual assets.
- Display detailed performance metrics: Total return, CAGR, volatility, Sharpe ratio, drawdowns, win rate, and average holding time.
- Future-proof: Designed for generalization to multi-pair strategies, portfolio backtests, and extended analytics.
- Stock Pair Selection: Choose from predefined or manually entered tickers for backtesting.
- Cointegration Testing: Evaluate long-term relationships using Engle-Granger & Johansen tests.
- Spread & Z-Score Construction:
- Calculate spread via OLS or rolling beta.
- Z-score computed over customizable rolling window.
- Backtest Engine:
- Realistic simulation of trade entries/exits based on signals
- Configurable stop-loss / take-profit bands
- Max holding period and cooldown logic
- Optional transaction costs or slippage built-in
- Cumulative Return Visuals:
- Plot strategy vs. individual stocks (e.g., KO, PEP)
- Log-scale toggle and zoomable charts
- Performance Metrics Dashboard:
- Total Return, CAGR, Annual Volatility, Sharpe Ratio
- Max Drawdown
- Number of Trades, Win Rate, Avg. Holding Period
- Interactive Education:
- Tooltips for key terms (e.g., cointegration, hedge ratio)
- Strategy rationale breakdowns in collapsible sections
- Select Stock Pair
Input two stock tickers (e.g., KO and PEP). - Load Data
Historical price data is fetched (default: 2018–2024) viayfinance. - Test Cointegration
Run Engle-Granger and Johansen tests, display p-values and interpretation. - Compute Spread & Z-Score
Construct spread using OLS or rolling hedge ratio. Normalize via z-score. - Configure Strategy Parameters
Customize thresholds, holding period, stop-loss, take-profit, cooldown, and transaction costs. - Simulate Backtest
Execute strategy logic, manage position states, and track cumulative returns. - Visualize Performance
Compare strategy returns vs. individual asset returns. - Review Metrics
Analyze key performance indicators and trade-level stats.
notebook/
├── app.py # Streamlit web app
├── data_utils.py # Data loading, cleaning, and return generation
├── pairs_analysis.py # Cointegration tests, hedge ratio, spread, z-score logic
├── strategy_simulation.py # Backtest engine, trade tracking, performance metrics
├── README.md # Project documentation
├── requirements.txt # Required Python packages
└── assets/ # Optional folder for logos or saved configs
- Multi-pair portfolio backtesting with ranking logic to simulate strategies across multiple cointegrated pairs.
- Transaction cost and slippage modeling for more realistic performance estimation.
- Dynamic (volatility-adjusted) z-score thresholds to adapt entry/exit logic to market regimes.
- Embedded educational tooltips, a glossary sidebar, and “Explain this” buttons to improve learning and transparency throughout the app.