Pairs Trading in Forex: When Two Legs Are Really One Cross

Pairs trading is described almost everywhere as a market-neutral structure: buy one instrument, sell a related one, and profit when the gap between them returns to its usual size. The published treatments are written around equities, where the two legs are two separate companies and the arithmetic stops there.

A currency quote is a ratio, and that changes what happens when you hold two of them at once. Two forex legs that share a currency do not stay two positions. They combine into a single exposure with a name of its own, and the trade you are actually carrying may already be quoted as one symbol on the same platform.

What a Pairs Trade Is Trying to Do

The structure has one moving part. Two instruments that normally track each other drift apart; the trader buys the one that has fallen behind and sells the one that has run ahead, then closes both when the gap narrows again. Direction of the wider market is meant to drop out, because whatever lifts or sinks both legs affects them roughly equally.

That is why the strategy is presented as neutral. The bet is on the relationship, not on either instrument, and the payoff comes from the distance between them rather than from a price level.

The idea was built for equities, where a leg is a share in one company and the two legs stay two things however long they are held.

A currency leg is a ratio between two national currencies, and the platform lists ratios between many of those currencies as symbols in their own right, a convention described under cross rates and how they are quoted. That difference is not cosmetic, and it decides what the structure is before any relationship is measured.

Everything in that description depends on one condition holding: that the gap has a level it tends to return to. If the two instruments simply drifted together for a while and then stopped, there is nothing to converge on, and the trade becomes two ordinary directional positions facing opposite ways.

In forex a second condition sits in front of that one, and it is arithmetic rather than statistical. Before asking whether a relationship holds, ask what the two legs add up to.

Before Anything Else: Does the Pair Reduce to a Cross Rate?

Every currency quote is one currency divided by another. Buying EURUSD is holding euros against dollars. Selling GBPUSD is holding dollars against pounds. Hold both and the dollar appears once on each side of the calculation, where it cancels.

What remains is euros against pounds, which the platform already lists as EURGBP. The two-leg position and the single symbol describe the same exposure, and they respond to the same events. Nothing about the pairing removes market direction, because there is only one market involved and it is the euro against the pound.

This is worth checking before any statistical work, since it decides whether the trade is a spread at all. The same cancellation runs through most of the combinations a screening tool will suggest, because the majors are quoted against one currency and that currency is usually the dollar.

The check itself takes seconds and needs no data. Write the four currency codes in the two legs, in the order each symbol quotes them, and see whether any code appears twice. A repeated code means the position collapses; the direction of each leg then tells you which of the two remaining currencies ends up on the numerator side.

The last row is the case where the two-leg structure survives the test. Four distinct currencies cannot be collapsed into one quote, so the position genuinely holds two exposures and the relationship between them is a real question.

The first three rows are not failures of the strategy. They are a reason to hold the cross directly: one symbol, one spread, one financing line, one margin calculation, and a chart of the actual thing being traded rather than two charts that have to be read together.

Correlation Is Not Cointegration, and Only One Supports the Trade

Where four currencies really are involved, the next question is whether the gap between the legs returns to a level. The published guides answer it with a correlation figure, and correlation answers a different question.

Correlation measures whether two series moved in the same direction over a chosen window. Two instruments can move together perfectly while the distance between them widens without limit, and that combination produces a high correlation reading alongside a spread that never converges.

How much the chosen window changes the figure is set out under currency correlation and the window it is measured over, and that page carries the measurement side of this in full.

Cointegration is the property the trade actually needs. It asks whether a particular combination of the two series stays within a stable range rather than wandering, which is another way of saying the gap has a level to come back to. Two series can be cointegrated while showing modest correlation, and can be strongly correlated while not being cointegrated at all.

Getting the two confused changes the meaning of a losing position. If the pair was selected on correlation alone, a spread that keeps widening is not an unlucky outcome to be waited out; it is what an unbounded relationship does, and there is no mechanism obliging it to reverse.

Testing for the property is a statistical exercise on historical data, and its result holds for the sample it was run on. It is a reason to consider a pair, not evidence about what the pair will do next. This is the same limit that applies to the wider family of relative-value approaches described under how statistical arbitrage differs from latency and triangular arbitrage.

Two Legs Means Two Spreads and Two Swap Lines

None of the treatments read while preparing this page put a cost figure against the structure. In forex the cost side is where a two-leg trade differs most from a single position, and it is arithmetic that anyone can run on their own account.

Opening two positions crosses two spreads on entry and two more on exit. The spread on a cross-rate symbol is not the sum of the spreads on the two majors it is built from, so a structure that reduces to a cross can cost more to hold as two legs than as one, purely through the number of times the bid-ask distance is paid.

Financing behaves the same way. MetaTrader 5 stores a long swap value and a short swap value as two separate properties of each symbol, alongside a swap calculation model and the weekday on which the triple rollover is applied. The two legs draw on two different symbols, so the platform applies each symbol’s own figures independently and there is no netting between them.

The consequence is that both legs can be charged. A long position in one symbol and a short position in another can each fall on the negative side of that symbol’s swap table, and the combined financing on a position held for weeks is then a running cost rather than the near-zero that a neutral structure implies. How the charge is applied to a single held position is covered under how swap is credited on a held position.

Both values sit in the contract specification the broker publishes for each symbol. That is where to read them, in the direction each leg will be held, and before the position is sized rather than after.

Whether Your Account Can Hold the Structure at All

Margin is the part most likely to differ from what the equity-focused treatments assume, and it depends on a setting the broker chose rather than on the trade.

MetaTrader 5 supports two position accounting systems, netting and hedging, and the documentation states that which one applies is set by the broker on the account. Under netting there is one position per symbol, so a second deal in the same symbol changes the existing position instead of creating another.

The platform does define a reduced margin treatment for exactly this kind of structure. Its margin documentation sets out a preferential charge for a spread, which it defines as two positions taken in related symbols facing opposite ways, and then restricts that treatment to accounts using netting. On a hedging account the relief is unavailable.

Hedged margin is a separate mechanism and is often mistaken for the same thing. It covers a buy and a sell held in one and the same symbol, with the broker choosing between margining the larger of the two sides and charging a set amount for each covered lot. Two related but different symbols fall outside it, however closely they track. The full behaviour of both accounting modes is set out under netting and hedging position accounting.

For a pairs trade this produces a concrete question with a concrete answer. Ask the broker which accounting system the account uses, and whether any correlated-symbol margin configuration is enabled on it. If the answer is a hedging account with no such configuration, the two legs are margined as two full positions, and the capital the structure ties up is roughly double what a single cross-rate position would need.

The Thresholds Every Guide States and No Source Supports

Published treatments of this strategy converge on a set of numbers: a minimum correlation a candidate pair should clear, a spread deviation at which to enter, a wider one at which to stop out.

Not one of those figures arrived with a citation attached. No study behind it, no sample, no market, no period, no data source. Two of the four treatments examined carry the identical set, which makes the numbers conventional rather than measured.

The dates on those pages belong beside the numbers. Of the four treatments examined, one was last revised in 2022, one was published in 2024 with no revision date shown, and one states no date at all. A threshold repeated across pages that have not been revisited is evidence of copying between them, not of a result that has held.

Numbers of that kind cannot be repeated here. They also cannot be checked, which is the more useful point: a threshold with no stated sample carries no information about the market it came from, and the same figure applied to equities and to currencies is unlikely to mean the same thing in both. Nowhere above will you find a correlation minimum, an entry band or a stop distance, and their absence was chosen deliberately.

Checking a Candidate Pair Before You Size It

Five checks settle whether a candidate is a spread trade, a disguised cross, or a structure the account cannot carry efficiently. Each has a definite answer available before any capital is committed.

First, write out the four currencies in the two legs. If any currency appears twice, work out which symbol the combination reduces to and look for it in the platform symbol list.

Second, if the cross exists, compare its spread against the two spreads the legs would cost, and treat the two-leg version as needing a reason beyond convenience.

Third, read the long and short swap values for both symbols in the contract specification, in the direction each leg will be held, and note whether both are negative.

Fourth, confirm the position accounting system on the account, and ask whether correlated-symbol margin relief is configured, since the netting requirement is documented rather than optional.

Fifth, establish whether the gap between the legs has a level to return to, rather than whether the legs moved together. Correlation over a window does not answer that question and is not a substitute for it.