How much capital do you need to start pairs trading?

You can start pairs trading with as little as $2,000–$5,000 using ETF pairs through most retail brokerages. Individual stock pairs may require more capital to maintain proper position sizing and meet margin requirements. Our simulator helps you model different capital levels to find the right starting point.

What is the ideal correlation threshold for a pairs trade?

Look for pairs with a rolling correlation above 0.80 over at least 60 trading days. Higher correlation (0.85+) provides more confidence that the relationship will persist. However, correlation alone is not enough—always verify cointegration using the Engle-Granger or Johansen test to confirm mean-reverting behavior.

How do you set entry and exit thresholds?

Common entry thresholds are z-scores of ±2.0 to ±2.5, meaning you enter when the spread deviates 2 to 2.5 standard deviations from its mean. Exit thresholds are typically set at ±0.5 to ±1.0. Tighter thresholds increase trade frequency but may reduce per-trade profit. Wider thresholds reduce whipsaws but require more patience.

Can pairs trading work in a bear market?

Yes—pairs trading is market-neutral by design. Since you go long one asset and short another, overall market direction matters less than the relationship between the pair. In fact, increased volatility during bear markets can create more frequent entry signals. However, correlations can break down during market stress, so risk management is critical.

How often do pairs trading opportunities arise?

For well-chosen pairs, meaningful deviations (z-score > 2) typically occur 3–6 times per year. If you monitor a portfolio of 10–20 pairs, you may have opportunities monthly. ETF pairs like XLE/XOP or XLU/XLF tend to offer more frequent signals than individual stocks due to their higher volatility.

Investment Strategies

Statistical Arbitrage for Retail Investors: Pairs Trading Guide

Learn how statistical arbitrage and pairs trading work, including z-score calculations, entry/exit thresholds, and real examples with KO/PEP. Build market-neutral strategies with our simulator.

Cogent Cash

Research Team

May 13, 2026 13 min read

Statistical arbitrage—often called "pairs trading"—is a market-neutral strategy that exploits temporary price divergences between historically correlated assets. Unlike directional investing, pairs trading profits from the relationship between two assets returning to its mean, regardless of whether the overall market goes up or down.

Key Takeaways

Pairs trading profits from the relationship between two assets returning to its mean, not market direction
Correlation measures directional similarity; cointegration confirms mean-reverting behavior
Entry at z-score ±2.0 captures 95% confidence level deviations from the historical spread
ETF pairs (XLE/XOP, GLD/SLV) offer frequent signals without short borrow costs
Maximum risk per trade should be 3-5% of allocated capital with a stop loss at |Z| > 3.0

Try our interactive Statistical Arbitrage Simulator to backtest pairs trading strategies, calculate z-scores in real-time, and optimize entry/exit thresholds.

What Is Statistical Arbitrage?

Key Insight

Pairs trading profits from the relationship between two assets, not market direction — making it a true market-neutral strategy.

Statistical arbitrage is a quantitative trading strategy that uses statistical and econometric models to identify temporary pricing inefficiencies between related securities. The most common form—pairs trading—was pioneered by Morgan Stanley's quantitative trading desk in the 1980s and has since become one of the most widely used market-neutral strategies.

Statistical Arbitrage (Pairs Trading)

A market-neutral strategy that identifies two historically correlated assets, waits for their price relationship to diverge beyond normal bounds, and bets on convergence back to the historical mean.

Example: Coca-Cola (KO) and Pepsi (PEP) historically move together. When KO underperforms PEP beyond normal levels, you go long KO and short PEP, profiting when the relationship normalizes.

The key insight is that pairs trading does not require predicting market direction. Instead, it exploits the statistical property that correlated assets tend to maintain their relative price relationship over time. When that relationship temporarily breaks down—due to company-specific news, sector rotation, or market inefficiencies—a trading opportunity arises.

How Pairs Trading Works

The pairs trading process follows four systematic steps:

Step 1: Pair Selection

The foundation of any pairs trade is identifying two assets with a strong historical relationship. This relationship is measured using two statistical concepts:

Correlation

A measure (from -1 to +1) of how closely two assets move together. Pairs trading requires high positive correlation, typically above 0.80.

Example: KO and PEP have a 1-year rolling correlation of 0.87, meaning they move in the same direction 87% of the time.

Cointegration

A stronger relationship than correlation. Two cointegrated assets have a spread (price difference or ratio) that is mean-reverting—deviations from the average tend to correct over time.

Example: Even if KO and PEP both drift higher over time, their spread oscillates around a stable mean, creating tradable deviations.

Critical distinction: Correlation measures directional similarity, while cointegration confirms mean-reverting behavior. A pair can be correlated without being cointegrated. Always test for cointegration using the Engle-Granger test or Johansen test before trading.

Step 2: Calculate the Spread Z-Score

Once you have a pair, you monitor the spread between them. The z-score tells you how many standard deviations the current spread is from its historical mean:

Z-Score

The number of standard deviations the current spread deviates from its historical mean. A z-score of 2.0 means the spread is 2 standard deviations above average.

Example: If the mean spread is 20, the standard deviation is 3, and the current spread is 26, then Z = (26 - 20) / 3 = 2.0

Step 3: Entry and Exit Thresholds

Thresholds determine when to enter and exit trades:

Z-Score	Action	Position
Z > +2.0	Entry	Short Asset A, Long Asset B
Z < -2.0	Entry	Long Asset A, Short Asset B
\|Z\| < 0.5	Exit	Close both positions
\|Z\| > 3.0	Stop Loss	Exit—relationship may have broken

The entry threshold of 2.0 is based on the statistical property that approximately 95% of observations fall within ±2 standard deviations in a normal distribution. A z-score beyond 2.0 signals an unusual deviation with a high probability of mean reversion.

Worked Example: KO/PEP Pairs Trade

Let's walk through a complete pairs trade using Coca-Cola (KO) and PepsiCo (PEP)—one of the most classic pairs trading examples.

Step 1: Establish the Historical Relationship

Over the past year, KO and PEP have shown:

Rolling correlation: 0.87
Mean spread: $20.00 (adjusted by hedge ratio β = 0.35)
Standard deviation of spread: $3.00
Cointegration confirmed: Engle-Granger test p-value = 0.012

Step 2: Identify the Signal

On a given trading day, you observe:

KO price: $62.00
PEP price: $175.00
Current spread: $62.00 - (0.35 × $175.00) = $62.00 - $61.25 = $0.75
Z-score: ($0.75 - $20.00) / $3.00 = -6.42

Wait—that z-score is extreme. In this simplified example, let's use more realistic spread values. With a mean spread of $20 and standard deviation of $3, a current spread of $26 gives:

Z-score: ($26 - $20) / $3 = +2.0
Signal: Entry—spread is 2 standard deviations above mean

Step 3: Execute the Trade

Since the spread is above the mean, you expect it to revert lower. This means:

Short the overperforming leg (KO in the spread)
Long the underperforming leg (PEP in the spread)

With a $10,000 portfolio and 20% risk allocation:

Short KO: $1,000 / $62 = 16 shares
Long PEP: $1,000 / $175 = 5.7 shares (round to 6)

Step 4: Exit the Trade

Two weeks later, the spread has converged:

KO price: $61.00 (down $1.00)
PEP price: $176.50 (up $1.50)
New spread: $61.00 - (0.35 × $176.50) = -$0.78
New z-score: (-$0.78 - $20.00) / $3.00 = -6.93

Again, this simplified example shows the mechanics. In practice, your z-score would move from +2.0 toward 0 as the spread converges. When the z-score crosses your exit threshold (e.g., |Z| < 0.5), you close both positions.

Step 5: Calculate Profit

Short KO profit: 16 shares × $1.00 = +$16.00
Long PEP profit: 6 shares × $1.50 = +$9.00
Gross profit: $25.00 (minus commissions and borrow costs)

Retail Approximation: ETF Pairs Trading

Individual stock pairs trading requires a margin account and the ability to short stocks. A simpler retail approximation uses ETF pairs, which often trade with tighter spreads and no short borrow costs.

Recommended ETF Pairs for Retail Traders

ETF Pair	Sector	Typical Correlation	Signal Frequency
XLE / XOP	Energy	0.92	4–6/year
XLU / XLF	Utilities vs Financials	0.78	3–5/year
XLP / XLY	Staples vs Discretionary	0.82	5–8/year
GLD / SLV	Precious Metals	0.85	6–10/year
VNQ / IYR	Real Estate	0.95	2–4/year

Retail advantage: ETF pairs can be traded in cash accounts using inverse ETFs instead of shorting. For example, instead of shorting XLE, you could go long SXY (an inverse energy ETF). This eliminates margin requirements and borrow costs, though it introduces tracking error and decay in inverse products.

Risk Management

Pairs trading is not risk-free. The primary risk is correlation breakdown—when the historical relationship between the pair permanently changes due to structural shifts in one or both assets.

Risk Management Framework

Risk Type	Mitigation	Max Loss
Correlation breakdown	Stop loss at \|Z\| > 3.0	3–5% of capital per trade
Execution risk	Enter both legs simultaneously with limit orders	Slippage of 0.1–0.3%
Borrow costs (shorts)	Use liquid ETFs or inverse products	0.5–2% annualized
Sector-wide shocks	Diversify across 5–10 uncorrelated pairs	Reduce single-pair risk
Extended divergence	Time stop: exit after 30–60 days if no convergence	Opportunity cost
Earnings events	Close positions before earnings of either asset	Gap risk eliminated

Position Sizing Rules

Maximum capital per pair: 15–20% of portfolio
Maximum open pairs: 5–10 (uncorrelated)
Stop loss per trade: 3–5% of allocated capital
Target return per trade: 1–3% (annualized: 15–30%)

Interactive Pairs Trading Calculator

Use this calculator to model pairs trading signals based on current prices and spread parameters:

Pairs Trading Signal Calculator

Adjust values to see current z-score and trading signal

Asset A Price

Asset B Price

Mean Spread

Spread Std Dev

Current Spread

$20.00

Z-Score

0.00

Signal

No Signal

Featured Tool

Statistical Arbitrage Simulator

Apply what you've learned with our interactive tool.

Backtest pairs trading strategies with historical data
Calculate z-scores and cointegration in real-time
Optimize entry/exit thresholds for maximum returns
Monitor multiple pairs with automated signal alerts
Model portfolio risk across uncorrelated pairs

Build Your Pairs Strategy

Frequently asked questions

Disclaimer: This content is for educational purposes only and does not constitute financial advice. Statistical arbitrage involves significant risk, including the potential loss of capital. Past performance does not guarantee future results. Pairs trading requires a margin account for short selling, and inverse ETF products have unique risks including tracking error and decay. Always consult with a qualified financial professional before implementing any trading strategy.