Pairs trading strategy: the market-neutral spread trade explained

Pairs trading compares two related instruments, using their spread to identify potential mean-reversion opportunities while managing the risk that the relationship breaks down.

This guide explains how pairs trading works, how traders assess cointegration and hedge ratios, and what risks to consider before using the strategy in CFD trading.

Understanding pairs trading

A pairs trading strategy is a market-neutral approach that involves simultaneously buying one instrument and selling a correlated instrument, with the expectation that the price relationship between the two will revert to its historical norm. Rather than taking a view on the broader market, the trade focuses on the relative performance of one instrument versus another – making it, in theory, independent of whether markets rise or fall.

The strategy is rooted in statistical arbitrage. Two instruments that share a common economic driver – such as two oil majors, two semiconductor stocks, or two related currency pairs – often move together over time. When their prices diverge significantly from their historical relationship, pairs traders may see a potential opportunity: go long the underperformer and short the outperformer, with the expectation that the spread between the two will narrow. The trade is closed when the spread returns to its historical mean, or when the trader’s loss threshold is reached.

What drives pairs trading

The theoretical basis of pairs trading is that two related instruments are ultimately driven by the same underlying economic forces. Understanding those forces helps determine whether a pair is genuinely suitable for the strategy.

Shared economic drivers

The most reliable pairs are those that share a fundamental driver – the same commodity price, interest rate environment, consumer demand cycle, or regulatory framework. Two mining stocks exposed to the same metal, two airlines in the same regional market, or two government bonds with similar maturity and credit quality may all show consistent long-term relationships. When one element of the shared driver affects one instrument more than the other in the short term, the spread can widen – creating a potential pairs trade.

Cointegration versus correlation

Correlation measures how closely two price series move together day to day. Cointegration is stronger and more useful for pairs trading: it means that the difference between two price series, or a linear combination of them, is stationary – reverting to a mean over time rather than drifting indefinitely. Correlation alone is not enough. A pair can be highly correlated in the short term while still trending apart over a longer period. Cointegration testing – using methods such as the Engle-Granger test or the Johansen test – is the standard way to assess whether a pair is suitable for mean-reversion trading.

Spread dynamics

The spread between two cointegrated instruments is the difference, or ratio, between their prices, adjusted by the hedge ratio that makes the pair market-neutral. The spread oscillates around a long-run mean. Traders enter when the spread reaches an extreme – typically a set number of standard deviations from the mean – and exit when it reverts. The speed and reliability of that reversion depend on the strength of the cointegrating relationship and the absence of structural breaks.

Past performance is not a reliable indicator of future results.

How to identify a trading pair

Selecting suitable pairs is the foundation of the strategy. A poorly chosen pair – one that is not genuinely cointegrated, or where the relationship has broken down – is a common source of pairs trading losses.

Qualitative screening

Start by identifying instruments that are logically related: they operate in the same industry, compete directly, are exposed to the same commodity or currency risk, or are part of the same broader index. Common equity pairs include major oil companies, such as ExxonMobil and Chevron, major US bank stocks, or two ETFs tracking the same sector. In forex, commonly traded pairs include AUD/USD vs. NZD/USD, or EUR/USD vs. GBP/USD, which share the same counter-currency and often respond similarly to US dollar movements.

Statistical testing for cointegration

Once a candidate pair has been identified qualitatively, apply a formal cointegration test over a historical window – typically one to three years of daily data. The Engle-Granger two-step method regresses one price series on the other, then tests the residuals for stationarity using an augmented Dickey-Fuller (ADF) test. A significant ADF result, with a p-value below 0.05, indicates that the spread is stationary and the pair is likely cointegrated. This test should be repeated periodically, as cointegration relationships can break down.

Estimating the hedge ratio

The hedge ratio defines how many units of the second instrument to sell for every unit of the first instrument bought, so the position remains market-neutral. It is typically estimated using ordinary least squares (OLS) regression of one price series on the other, or with a Kalman filter for a dynamic hedge ratio that updates as the relationship evolves. An incorrect hedge ratio leaves residual directional exposure, weakening the market-neutral premise of the strategy.

Past performance is not a reliable indicator of future results.

How to execute a pairs trade

Once a suitable pair has been identified and the hedge ratio established, the trade follows a systematic process tied to the spread’s behaviour relative to its mean.

• Step 1. Calculate and monitor the spreadCompute the spread at regular intervals: Spread = Price(A) – (hedge ratio × Price(B)), or use the log-price ratio for percentage-based spread analysis. Plot the spread over time, then calculate its rolling mean and standard deviation. Most implementations use a rolling window of 20–60 trading days to define the current mean and standard deviation, helping the parameters adapt to changing market conditions without overfitting to older history.
• Step 2. Define entry thresholdsThe standard entry rule is to open a pairs position when the spread diverges by more than a set number of standard deviations from its rolling mean. A common threshold is ±2 standard deviations. At +2 standard deviations, traders short the spread by selling A and buying B. At –2 standard deviations, traders go long the spread by buying A and selling B. More conservative traders may use ±2.5 or ±3 to reduce trade frequency and filter for wider divergences, at the cost of fewer opportunities.
• Step 3. Enter the tradeWhen the spread breaches the entry threshold, execute both legs simultaneously, or as close to simultaneously as possible, to reduce leg risk – the risk that one leg is filled and the spread moves against the unfilled second leg. Simultaneous execution is easier for liquid, exchange-traded instruments and more challenging for OTC markets. In CFD trading, pairs can be executed by opening individual positions in both instruments.
• Step 4. Set exit conditionsDefine two exit conditions before entering the trade. The profit exit: close the spread position when the spread returns to within a set band of the mean, commonly 0.5 standard deviations from the mean or the mean itself. The stop-loss exit: close the position if the spread moves further in the adverse direction beyond a defined threshold – typically ±3 or ±3.5 standard deviations – to limit the risk of a structural break in the relationship. Stop-loss orders are not guaranteed. Guaranteed stop-loss orders incur a fee if activated.
• Step 5. Monitor and close the trade Monitor both legs and the spread value throughout the trade. If the spread reverts to the mean, close both legs simultaneously and record the outcome. If it continues to widen beyond the stop threshold, close both legs to crystallise the loss. Some traders also close pairs positions ahead of scheduled earnings announcements or major economic events that could affect one instrument more than the other, reintroducing directional risk.

Past performance is not a reliable indicator of future results.

Types of pairs trading approaches

Pairs trading is not a single method. Several variants apply the same core concept using different instruments, timeframes, and analytical frameworks.

Using pairs trading in practice

Practical implementation involves more than the statistical framework. Execution mechanics, position sizing, and ongoing monitoring all affect real-world outcomes.

Position sizing and capital allocation

Because pairs trading involves two legs, position sizing must account for both. The standard approach is to size positions in value-neutral terms: if you are long £50,000.00 of Instrument A, you sell short £50,000.00 of Instrument B, adjusted by the hedge ratio. This is often called dollar-neutral sizing in industry terminology, but the same principle applies in sterling or any other account currency. The aim is to ensure that a general market move affects both legs in a similar way, leaving profit and loss driven mainly by the spread. Value-neutral sizing also prevents one leg from dominating the trade’s risk profile.

Transaction costs and the break-even spread

Pairs trading generates twice the transaction costs of a single directional trade – one trade for each leg, plus potentially two closing trades. In equity markets, it may also include the cost of borrowing the short position. The spread must move enough from entry to the mean to recover these costs and generate a net profit. This is why pairs trading tends to favour liquid instruments with tight spreads, and why trades with small spread divergence relative to transaction costs may be less practical.

Monitoring for cointegration breakdown

Cointegration is not permanent. A significant event – a major acquisition, regulatory change, or commodity price shock affecting one instrument more than the other – can permanently alter the relationship between two instruments. If a pairs trade is open and the spread continues to widen well beyond historical norms, traders need to reassess whether the relationship still holds. A position that continues to lose on the assumption of mean reversion, when the structural relationship has changed, is no longer a pairs trade – it is a directional loss.

Advanced pairs trading techniques

Beyond the classical two-asset pairs trade, several more sophisticated extensions exist for traders with a quantitative background.

Portfolio cointegration: baskets

Rather than trading a single pair, some systematic traders build portfolios of cointegrated pairs to diversify the idiosyncratic risk of any one relationship breaking down. A basket of 10–20 cointegrated pairs across different sectors reduces the impact of a single cointegration failure. The combined portfolio tends to exhibit lower drawdowns than individual pairs, though it requires more complex monitoring and execution infrastructure.

Dynamic hedge ratio with Kalman filtering

A static hedge ratio – estimated once over a historical window – can become stale as the relationship between instruments evolves. The Kalman filter is a sequential updating algorithm that continuously re-estimates the hedge ratio as new price data arrives. A Kalman filter-based pairs strategy adapts the hedge ratio in real time, potentially maintaining better market neutrality over time. The tradeoff is increased model complexity and sensitivity to filter parameter choices.

Multi-leg statistical arbitrage

An extension of classical pairs trading, multi-leg statistical arbitrage constructs a basket of instruments whose combined price series is stationary, even if no individual pair is cointegrated on its own. For example, a long position in one sector ETF combined with short positions in two different sector ETFs in specific proportions may produce a stationary residual. The Johansen test can be used to identify and estimate such multi-asset cointegrating relationships, providing a statistical basis for the portfolio weights. This method requires more complex portfolio construction but offers more flexibility in identifying tradeable spreads.

Common mistakes in pairs trading

Several recurring errors can reduce the effectiveness of pairs trading strategies in practice.

• Choosing pairs based on correlation alone: high correlation doesn’t mean two assets are cointegrated. Without cointegration, the spread may keep widening instead of reverting.
• Ignoring transaction costs: a full trade cycle involves opening and closing both legs. Spreads, commissions and any shorting costs can reduce or outweigh expected returns.
• Failing to respect the stop-loss: if the spread keeps widening, adding to the position can increase losses. Set a stop-loss in advance and stick to it.
• Using too short a testing window: Cointegration tests are less reliable with small samples. Very short windows can create false positives, making a pair look more stable than it is.

Open a demo account

FAQ