Currency Correlation in Forex: Sizing Correlated Trades

Risk rules are applied one trade at a time, and they assume each trade is a separate bet. In currency markets that assumption is often false, because every position involves two currencies and separate positions can rest on the same one.

Correlation is how that overlap is measured. Most guides on the subject present a table of coefficients and stop there, which is the least useful part: a coefficient with no window attached cannot be acted on, and the relationships it describes are least reliable exactly when they matter most. What follows covers the measurement, its instability, and the sizing method that follows from both.

What Currency Correlation Measures

Correlation measures the degree to which two instruments have moved together over a defined period. It is a description of past behaviour, not a prediction of future behaviour.

The distinction matters more here than in most statistics. A correlation figure tells you what the relationship has been; it carries no commitment that the relationship continues, and the sections below cover the conditions under which it reliably does not.

What it is genuinely useful for is answering one question about your open positions: am I holding several trades, or several expressions of the same trade?

Positive, Negative and the Coefficient Scale

The coefficient runs from -1 to +1, and some platforms present the same figure scaled from -100 to +100.

A value near +1 means the two pairs have moved in the same direction, close to in step. A value near -1 means they have moved in opposite directions, equally consistently. A value near 0 means no reliable relationship was present over that period.

The practical reading is about magnitude before sign. A figure of -0.9 and a figure of +0.9 describe relationships of identical strength; only the direction differs. For risk purposes both are strong, and a figure near 0 is the only one that lets you treat two positions as genuinely separate.

Why a Correlation Figure Is Meaningless Without Its Window

This is the omission that makes most published correlation tables unusable, and it is worth stating before any figure is quoted.

A correlation coefficient is computed over a specific number of past periods. That choice is not a detail; it is part of the result. The same two pairs measured over the last twenty days and over the last two hundred can produce materially different figures, and neither is more correct than the other.

A short window is responsive and noisy. It reflects the relationship as it stands now, and it will swing on a few unusual sessions. A long window is stable and slow. It describes a durable structural relationship and will keep showing one for some time after it has stopped holding.

The consequence for the reader is direct. A correlation table published with no window and no date attached is not information you can act on, because you cannot tell which of those two things it is describing. Any figure worth using is one you computed over a window you chose, for a reason you can state, on data ending today.

Match the window to the holding period. A position intended to last a day is not informed by a two-hundred-day relationship, and one intended to last a quarter is not informed by last week’s.

Where Correlations Actually Come From

Correlation between currency pairs is mostly structural, which is why it is more persistent than correlation between unrelated instruments.

The dominant cause is a shared currency. Every pair is a ratio between two currencies, so two pairs holding one in common are partly driven by the same thing. Where the shared currency sits determines the sign: shared in the same position tends toward positive correlation, shared in opposite positions tends toward negative.

That is the mechanism behind the familiar example. Several pairs quoted against the US dollar are each, in part, a position on the dollar. A broad dollar move affects all of them at once, and it does so regardless of what the other currency in each pair is doing.

A second cause is shared exposure. Currencies of economies dependent on a particular export tend to move with that export’s price, and with each other. A third is policy convergence, where central banks facing similar conditions move in similar directions, which also means a divergence in central bank tone can end the relationship.

Knowing the cause is what lets you anticipate the exception. A correlation resting on a shared currency persists as long as that currency is the dominant driver, and weakens when something specific to one of the other currencies takes over.

Positioning is a further cause, and it is the one behind the largest correlated moves. Where many traders hold the same funding currency across different pairs to collect overnight financing, those pairs move together when that currency does, as described under the forex carry trade.

The same reasoning extends beyond currency pairs to bonds, equities and commodities. Our page on intermarket analysis covers which of those cross-market relationships have held and which reversed.

Why Correlations Tighten When Markets Are Stressed

Here is the property that matters most for risk, and that every readable source on this topic omits or mentions only in passing.

Correlations are not stable across market conditions. In calm periods, pairs are driven by their own particular circumstances, and measured correlations are moderate. Under stress, participants tend to reprice broad exposures at once, and the relationships tighten toward each other.

State the consequence plainly, because it inverts the usual reasoning. Diversification across correlated pairs works best when you need it least, and thins exactly when you need it most. Positions that behaved independently for months can move together on the day that decides the account.

This is why a limit derived from calm-market correlations is the loosest version of the constraint rather than the tightest. It is the maximum exposure that was ever defensible, not a safe allowance.

The practical response is not to compute better correlations. It is to set the exposure limit assuming the relationships will be stronger than measured, and to treat any calm-market figure as a floor under the true combined risk rather than an estimate of it.

Sizing Across Correlated Positions

Per-trade risk rules break silently here. A rule that permits one percent per trade permits three percent across three trades, and if those three rest on the same currency, the account is carrying something much closer to a single three percent position on one view.

The fix is to change what the limit applies to. Instead of capping risk per position, cap it per currency.

The method is arithmetic. For every open position, note which two currencies it involves and in which direction. For each individual currency, total the risk of every position exposed to it in the same direction. Then apply your limit to that total rather than to any single trade.

Worked as a hypothetical: three long positions in different pairs, each sized to risk one percent, each short the same currency on the other side of the pair. The per-trade rule reports three separate one percent risks. The per-currency total reports three percent on one currency, which is what the account actually holds.

Two consequences follow. The number of positions you can hold in one direction falls, which is the point rather than a side effect. And the individual position sizes stay computed exactly as before, through the same position size calculator and the same forex lot sizes; what changes is how many of them you may hold at once.

This sits directly under the account-level limits described in forex risk management, and it is the control that the per-trade rule cannot provide on its own.

Why a Negatively Correlated Pair Is Not a Hedge

Negatively correlated positions are repeatedly described as a hedge, and for equal sizes that description is wrong in a way worth spelling out.

Take two positions of the same size in pairs correlated near -1. When one gains, the other loses by approximately the same amount. The combined result is approximately zero before costs.

After costs it is worse than zero. You have paid the spread on both positions, and you continue to pay any overnight financing on both. What has been constructed is a closed position that charges you to hold it.

The general rule: offsetting exposure is not the same as reducing risk. If the intention is to have no exposure, closing the position achieves it at half the cost. A negative correlation is useful for understanding what your open positions add up to, not as an instrument for cancelling them.

There is a narrow legitimate case, and it is worth naming so the rule is not overstated. Unequal sizes in negatively correlated pairs leave deliberate net exposure to one side, which is a position with a view, not a hedge. That is a different decision, taken knowingly.

When Correlation Does Not Apply to You

Correlation is a portfolio property. It describes what happens between positions, so it requires more than one position to mean anything.

If you hold one trade at a time, it does not apply to you at all. There is no combined exposure to compute, and time spent on correlation tables is better spent on the entry and the stop for the single position you hold. This is the most common case among traders learning the method, and no source on the topic says so.

It also matters little for positions held over minutes. Intraday moves are driven by order flow and news arriving in the moment, and a relationship measured across days has limited bearing on the next few minutes, whichever trading sessions you trade.

Where it is decisive is the case in between: several positions held simultaneously over hours or days, sized independently, in pairs that share a currency. That combination is where accounts take losses several times larger than any individual trade was meant to risk.

One further limit is worth stating. Correlation measures the strength of a linear relationship between past moves. It does not establish that one pair causes the other to move, and it does not describe how the two behave in the extreme moves that do the real damage. It is one input to a sizing decision, not the sizing decision, and the broader context is covered in currency trading methods.

Frequently Asked Questions

What is currency correlation in forex?

It is a measure of how closely two currency pairs have moved together over a defined past period, expressed from -1 to +1. It describes what the relationship has been, not what it will be, and its main use is establishing whether several open positions are genuinely separate trades or one exposure held several times.

What does a correlation coefficient of -1 mean?

It means that over the measured period the two pairs moved in opposite directions in near-perfect step. For risk purposes that is a strong relationship, the same strength as +1 with the sign reversed. It does not mean one pair cancels the other safely, because equal-sized offsetting positions still pay costs on both.

Does the correlation between two pairs stay the same?

No. The figure depends on the window it was computed over, so short and long lookbacks on the same two pairs can differ materially. Correlations also tend to tighten under market stress, which means a calm-period figure understates how together those positions will move on a difficult day.

How should I size trades in correlated pairs?

Apply the limit per currency rather than per position. Total the risk of every open position exposed to the same currency in the same direction, and cap that total at the level you would otherwise allow one trade. Individual position sizes are calculated as usual; what changes is how many you may hold at once.

Is trading two negatively correlated pairs a hedge?

Not at equal sizes. The two roughly cancel before costs and lose after them, because spread and any overnight financing are paid on both. If the aim is no exposure, closing the position costs less. Deliberately unequal sizes leave net exposure to one side, which is a position with a view rather than a hedge.