Statistical Arbitrage

StatArb is a quantitative strategy that uses data-driven models to identify temporary price deviations between two or more related assets. The most common form is "Pairs Trading"—where you find two stocks that are highly correlated (e.g., Coke and Pepsi, or two major Banks). When one stock temporarily outperforms or underperforms the other, you buy the "weak" one and short the "strong" one, betting that their prices will eventually return to their historical correlation.

The beauty of StatArb is that it can be market-neutral. By simultaneously holding a long and a short position, you eliminate "beta" (market risk). If the entire market crashes tomorrow, your long and short positions should theoretically move together, canceling each other out. Your profit comes only from the *convergence* of the specific pair you are trading.

• Selection: Identify pairs with a high historical correlation coefficient (e.g., > 0.90).
• Spread Calculation: Create a "spread" chart: $Spread = Price_A - (HedgeRatio \times Price_B)$.
• Z-Score Analysis: Monitor the Z-Score of the spread. When the Z-Score exceeds a threshold (e.g., +2.0 or -2.0), the spread is statistically "stretched" and likely to revert.
• Execution: Enter the trade when the Z-Score is extreme; close the trade when the Z-Score reverts toward zero.