Pairs Trading & Statistical Arbitrage in Python: Market-Neutral Alpha [2026]
Date Published
1. Correlation vs Cointegration: The Mathematical Difference
Why Cointegration is Essential for Statistical Arbitrage
Two assets can be 95% correlated over three years and yet diverge permanently tomorrow if their business models separate. Cointegration guarantees that a linear combination of two price series (Spread = Asset_A - beta * Asset_B) wanders around a stationary mean with finite variance.
2. StatArb Pair Categories and Execution Parameters
Empirical Half-Lives and Z-Score Bounds
Below are typical cointegrated equity and ETF pairs evaluated across a 10-year test window.
Pair Category | Example Tickers | OU Half-Life | Entry Z-Score | Exit Z-Score | Stop Z-Score Oil Majors | XOM vs CVX | 8.4 Days | ±2.0 SD | 0.0 SD | ±3.5 SD Beverage Duopoly | KO vs PEP | 14.2 Days | ±2.2 SD | 0.0 SD | ±3.8 SD Semiconductor ETFs | SMH vs SOXX | 5.1 Days | ±1.8 SD | 0.0 SD | ±3.2 SD Payment Networks | V vs MA | 11.6 Days | ±2.0 SD | 0.0 SD | ±3.5 SD
Was this article helpful?
Your feedback directly guides the research and strategies we publish next.