Trading strategies · Advanced

Pairs Trading And Statistical Arbitrage

A directional trade has one way to lose. A pairs trade has a way to lose on both legs at once, exactly when the relationship stops holding.

Last reviewed 15 September 2026 by the MarketPro research desk · Editorial policy

In short

Pairs trading takes a long position in one instrument and a short position in a related one, aiming to profit from the gap between them narrowing rather than from market direction. If the two converge, both legs contribute. If they rise together the legs cancel, which is the market-neutral property. If the divergence widens, both legs lose at the same time.

What to remember

Market neutral removes directional risk, not risk. Relationship risk, double execution and double financing all remain.
The losing case hits both legs simultaneously, which is the outcome that descriptions of this strategy tend to skip.
Correlation is not enough. The property the strategy needs is a spread with a stable average to return to, checked across several windows.
Costs are doubled on entry, exit, commission and overnight financing, against a target that may only be a couple of percent.

Lesson 16 of 24 in Trading strategies

PreviousTrading Around News And Economic Events NextVolume Analysis For Confirmation

A spread trade written out as two tickets

Pairs trading buys one instrument and sells a related one at the same time, aiming to profit from the gap between them narrowing rather than from the market going up or down. It is worth seeing the arithmetic before the theory, because the theory makes it sound cleaner than it is.

Two correlated instruments, A and B, have historically moved together. Over three weeks A rises 4% while B rises 1%, a divergence of three percentage points against a historical average of well under one. A pairs trader sells A and buys B in sizes chosen so both legs carry the same monetary exposure, then waits for the gap to close.

Convergence to the average: the gap narrows to 0.8 points. The short leg gains, the long leg gains, and the profit is the change in the spread rather than in either instrument.
Both rise together: A gains another 5% and B gains 5%. The legs cancel. The position is flat, which is the market-neutral property working as intended.
Divergence widens: A gains 6% while B falls 1%. Both legs now lose, which is the case that people who describe this as low risk tend to skip.

That third line is the whole risk of the strategy. A single directional trade has one way to lose. A pairs trade has a way to lose on both legs simultaneously, and it happens precisely when the relationship you relied on stops holding.

What market neutral does and does not remove

Market neutral means the position has little exposure to the overall direction of the market. It does not mean low risk, and the distinction gets blurred constantly.

What it removes is broad directional risk, so a general sell-off affects both legs roughly equally. What it leaves entirely intact is relationship risk, which is the risk that the specific link between these two instruments changes. It also introduces execution risk on two positions instead of one, and financing costs on both.

When the relationship simply ends

Correlation is a description of the past, and it is measured over a chosen window. Two instruments that moved together for two years can stop for reasons that have nothing to do with statistics:

12 more sections in MarketPro

One central bank shifts policy and the other does not, which permanently repositions a currency pair relationship.
A company in a stock pair announces a strategic change, a legal problem or an earnings miss that the other does not share.
A commodity relationship breaks on a supply disruption specific to one side.

In every one of those cases the divergence that looks like a trading opportunity is actually information. The market is repricing one instrument for a reason, and a trader betting on convergence is betting against that reason without necessarily knowing what it is. This is the central difficulty: the setup looks identical whether the gap is noise or news.

Measuring a relationship before trusting it

A correlation figure alone is insufficient, because correlation measures whether two things move together, not whether they return to a stable relationship. The property pairs trading actually needs is that the spread between them is mean-reverting.

Check over multiple windows. A relationship that holds over 30 days, 90 days and a year is more credible than one visible only in the most recent month.
Plot the spread itself, not the two price lines. If the spread has no stable average to return to, there is nothing to trade.
Measure typical deviation so you know what an unusual gap actually is in this pair, rather than judging it by eye.
Ask what economically connects them. A relationship with a mechanism behind it, such as two currencies with linked economies, is more durable than one found by searching historical data for correlations.

That last step is the defence against a common failure: search enough combinations and you will find pairs that correlated strongly by coincidence, and those relationships do not persist.

Two legs, two spreads, two sets of financing

The costs deserve their own accounting, because they are doubled and easily underestimated.

Cost	Single trade	Pairs trade
Spread paid	Once	Twice, on entry and exit
Commission	One round turn	Two
Overnight financing	One position	Both legs, often for weeks
Margin	One requirement	Two, unless netted by the broker

Since the profit target is a convergence that may be a couple of percent, doubled costs consume a meaningful share of it. Check what margin each leg requires with the margin calculator, because holding two positions for several weeks ties up more capital than the headline exposure suggests.

Why this sits at the advanced end

Pairs trading requires managing two positions, tracking a relationship rather than a price, sizing both legs so the exposures actually balance, and accepting that the loss case hits both sides at once. None of that is exotic, but all of it assumes a solid grasp of correlation and its limits first. Statistical relationships describe what has happened. They do not constrain what happens next, and the trades that hurt most are the ones where a relationship ended for a reason that only became obvious afterwards. The glossary covers correlation and the related terms if any of that is unfamiliar.

Finish this lesson in MarketPro

The rest of Pairs Trading And Statistical Arbitrage, a quiz that checks it landed, and your place kept in the curriculum.

12 more sections of this lesson
All 320 lessons, with no cap and no paywall
Free on any account, including a free one

In your browserContinue on webor continue on web

Same account as the app · No install · Free to start

Or scan with your phone camera to get the MarketPro app.

Download on theApp Store Get it onGoogle Play

Free account, no card needed.

Not investment advice. Past performance is not indicative of future results.

Run the numbers while this is fresh

Work it through on your own position size. Nothing leaves your browser and there is no sign-up: Position size calculator, Profit & loss calculator, Compounding calculator.

Written and reviewed by the MarketPro research desk

MarketPro is a trading-signal and trading-education app operated by Harajuku Holdings LTD (Cyprus). Our signal desk publishes and tracks every trade idea in-app, and every page here is checked against the app's live behaviour before publishing. Last reviewed 15 September 2026. How we produce signals and content · About MarketPro

FAQ

Pairs Trading And Statistical Arbitrage: questions

What is a market neutral trading strategy?

It is a position constructed so that the overall direction of the market has little effect on the outcome, usually by holding an equal monetary exposure long in one instrument and short in a related one. A broad rally or sell-off then moves both legs roughly together and cancels out. What remains is exposure to the specific relationship between the two instruments, which can still produce a substantial loss.

How do you choose pairs for pairs trading?

Start from an economic connection rather than a statistical search, such as two currencies with closely linked economies or two companies in the same sector facing the same inputs. Then check that the spread between them, not just their correlation, has a stable average across several time windows. Pairs found purely by scanning historical data for high correlation frequently turn out to have been coincidental.

Is pairs trading lower risk than directional trading?

It removes one source of risk and adds others, so it is different rather than safer. Direction matters less, but the trade depends on a historical relationship continuing to hold, and when a divergence reflects genuine news about one instrument rather than temporary noise, both legs can lose together. It also doubles the costs and requires managing two positions instead of one.

Trade · Learn · Earn

The whole academy is free in MarketPro

All 320 lessons, 12 guided paths, quizzes and a daily streak. Free on every account, in the app or your browser, with vetted free signals alongside it.

In your browserContinue on webor continue on web Download on theApp Store Get it onGoogle Play

Same account as the app · No install · Free to start

Not investment advice. Past performance is not indicative of future results.