Pairs trading is a statistical arbitrage strategy that involves identifying two securities with historically correlated prices, then taking offsetting long and short positions when their spread widens beyond a threshold. The trade profits when prices converge back to their historical relationship. It is one of the simplest and most widely studied forms of statistical arbitrage.
Pairs trading is a market-neutral trading strategy that involves taking simultaneous long and short positions in two historically correlated securities. When the price relationship between the two securities temporarily diverges from its historical norm, the trader goes long the relatively cheaper security and short the relatively expensive one, betting that the spread will revert to its mean.
The strategy was pioneered by Nunzio Tartaglia's quant group at Morgan Stanley in the mid-1980s and quickly became one of the most widely practiced forms of statistical arbitrage. Its appeal lies in its conceptual simplicity and its natural hedge: because you are simultaneously long and short, the portfolio is largely insulated from broad market movements.
Pairs trading is built on the concept of mean reversion β the tendency of the spread between two correlated instruments to return to its historical average. When the spread is unusually wide, it's likely to narrow; when it's unusually narrow, it's likely to widen.
A pairs trading strategy is built in three phases:
Phase 1 β Pair Selection: Identify pairs of securities with a stable, mean-reverting spread. Common approaches include:
Phase 2 β Signal Generation: Once a pair is identified, compute the spread (typically the price ratio or regression residual) and its z-score. A z-score above +2 (spread is unusually wide) triggers a short-the-outperformer / long-the-underperformer trade. A z-score below -2 triggers the opposite.
Phase 3 β Execution and Risk Management: Enter positions when the z-score exceeds a threshold, exit when it reverts near zero, and stop out if the z-score reaches an extreme level (e.g., |z| > 4) suggesting a structural break.
Mock interviews, resume guides, and 500+ practice questions β straight to your inbox.
Visa (V) and Mastercard (MA) are classic pairs trading candidates β both are payment network duopolies with similar business models, revenue drivers, and competitive dynamics.
Setup: Over the past 12 months, the price ratio V/MA has averaged 0.55 with a standard deviation of 0.015.
Signal: Today, V is at $280 and MA is at $540. The ratio is 280/540 = 0.519. The z-score is (0.519 - 0.55) / 0.015 = -2.07. This means Visa is unusually cheap relative to Mastercard.
Trade: Buy $100,000 of Visa (357 shares at $280) and short $100,000 of Mastercard (185 shares at $540). The portfolio is dollar-neutral.
Outcome scenario β convergence: Over two weeks, V rises to $290 and MA stays at $540. The ratio is now 290/540 = 0.537, z-score = -0.87. The spread has partially converged. P&L: Long V gain = 357 Γ $10 = +$3,571. Short MA gain = $0. Total: +$3,571.
Outcome scenario β divergence: V drops to $270 and MA rises to $560. The ratio is 270/560 = 0.482, z-score = -4.53. The spread has widened further. P&L: Long V loss = 357 Γ (-$10) = -$3,571. Short MA loss = 185 Γ (-$20) = -$3,700. Total: -$7,271. This is why stop-losses are essential.
Want personalized guidance from a quant?
Speak with a quant trader or researcher whoβs worked at a top firm.
Book a Free ConsultWhile simple pairs trading is well-known and widely taught, modern stat arb firms have extended the concept far beyond two-stock pairs:
Pairs trading remains a popular backtesting exercise and is often used as a homework problem in quant interviews. Understanding the mechanics, risks, and statistical foundations of pairs trading is excellent preparation for more advanced stat arb discussions at firms like Citadel and Two Sigma.
Z-score of the price ratio: measures how far the current ratio has deviated from its historical mean. Entry signals are typically triggered at |z| > 2.
Regression-based spread: stock A is regressed against stock B, and the residual is the spread to trade. Beta (the hedge ratio) determines the relative position sizing.
Pairs trading is one of the most common topics in quant trading interviews, especially at hedge funds that run statistical arbitrage strategies. Interviewers may ask you to design a pairs trading strategy, discuss cointegration vs. correlation, or walk through the risks. Understanding pairs trading demonstrates your grasp of market-neutral strategies and mean reversion β core concepts at firms like Citadel and Point72.
See our Citadel interview questions for examples. Book a free consultation to prepare for quant research interviews.
Statistical arbitrage (stat arb) uses quantitative models to identify and exploit temporary pricing inefficiencies between related securities, typically holding diversified portfolios of long and short positions.
Mean reversion is the tendency of asset prices, returns, or other financial metrics to move back toward their long-term average after deviating significantly, forming the basis for many systematic trading strategies.
Backtesting is the process of testing a trading strategy against historical market data to assess how it would have performed, helping quants evaluate strategies before deploying real capital.
The Sharpe ratio measures risk-adjusted return by dividing a portfolio's excess return over the risk-free rate by its standard deviation, making it the gold standard for comparing strategy performance.
Simple two-stock pairs trading has become less profitable over time as the strategy has become widely known and competition has increased. However, more sophisticated versions β using cointegration, machine learning, alternative data, and multi-asset approaches β remain profitable at well-resourced quant firms. The key insight is that the basic concept (mean reversion between related securities) is sound, but the implementation must be increasingly sophisticated to maintain an edge.
Correlation measures how closely two price series move together in the short term. Cointegration measures whether two price series maintain a stable long-term equilibrium relationship. Two stocks can be highly correlated but not cointegrated (they move together but can drift apart permanently). Cointegration is the more appropriate test for pairs trading because it captures the mean-reverting property of the spread.
The primary risks are: (1) Structural break β the historical relationship permanently changes due to mergers, industry disruption, or fundamental shifts. (2) Liquidity risk β inability to exit positions during market stress. (3) Model risk β the statistical model is wrong or overfitted to historical data. (4) Execution risk β slippage and transaction costs eroding thin margins. (5) Crowding β too many traders running the same pairs can reduce profitability and increase crash risk.
Start with a fundamental rationale β companies in the same industry with similar business models (e.g., Coca-Cola and PepsiCo, or Visa and Mastercard). Then apply statistical tests: check that the price spread is stationary (mean-reverting) using the Augmented Dickey-Fuller test or Engle-Granger cointegration test. A pair should have both an economic rationale and statistical evidence of mean reversion.
Our bootcamp covers probability, statistics, trading intuition, and 500+ real interview questions from top quant firms.
Book a Free Consult