💹 Kalman Filter-Based Statistical Arbitrage Strategy

This project implements a dynamic pairs trading strategy using Kalman Filters to estimate time-varying hedge ratios in cointegrated stock pairs from the Consumer Staples sector. Developed for the Data-Driven Portfolio Optimisation (IEDA3180) course at HKUST.

🧠 Project Objectives

• Detect mean-reverting opportunities in stock pairs
• Estimate dynamic hedge ratios using Kalman Filters
• Backtest and compare strategy performance against benchmarks

⚙️ Techniques & Tools

• Languages & Libraries: Python, pykalman, statsmodels, matplotlib, seaborn, Yahoo Finance API
• Pair Selection:
  • Engle-Granger test
  • Augmented Dickey-Fuller (ADF)
  • Half-life analysis
• Signal Generation:
  • Rolling Z-score of Kalman-estimated spread
  • Entry: ±2.0 | Exit: ±0.5 thresholds

📊 Strategy Performance (Out-of-Sample: 2022–2024)

• CAGR: 6.7%
• Sharpe Ratio: 1.26
• Max Drawdown: -3.5%
• Outperformed SPY and static OLS strategies in both returns and risk-adjusted metrics

📂 Files

• code/: Backtesting and signal generation scripts
• graphs/: All the graphs generated
• slides.pdf: Presentation slides

🧰 Skills Demonstrated

Quantitative Research · Kalman Filtering · Portfolio Optimisation · Time Series Analysis · Backtesting

🔗 Connect

LinkedIn • Email