A trader risking 1% per position opens three: long EURUSD, long GBPUSD, long AUDUSD. On the risk sheet that reads as 3% at stake across a diversified book. In practice a broad dollar rally takes all three toward their stops at the same time, and the account discovers it was carrying one 3% bet on a single idea.
Every FX quote has two currencies in it. The moment you hold more than one pair, the currencies overlap, and the overlap is where the real exposure lives.
Where the co-movement comes from
Three mechanisms produce most of it.
The first is a shared currency. EURUSD and GBPUSD both quote against the dollar, so a dollar-driven session moves them in the same direction. EURUSD and USDCHF quote the dollar on opposite sides, so the same session moves them in opposite directions. Neither relationship is a market insight, it is arithmetic in the quote.
The second is a shared driver. The Australian and New Zealand dollars respond to the same commodity demand and the same regional growth story. The Canadian dollar tracks oil closely enough that an energy move shows up in USDCAD before anything domestic does. Their correlation is not about the quote currency at all, it is about what both economies sell.
The third is risk appetite. In a defensive market the yen and the franc firm while higher-yielding currencies weaken, which pulls a long list of unrelated-looking pairs into a single trade about sentiment. That mechanism is set out in safe haven flows, and it is the one that turns a diversified-looking book into a single position in exactly the sessions where you need diversification.
Correlation is a number with a window attached
The correlation coefficient runs from +1 to -1. Above roughly +0.7 two pairs are behaving as near-substitutes, and below roughly -0.7 they are behaving as near-opposites. Between -0.3 and +0.3 the relationship is weak enough to treat the positions as separate.
The window is what people forget. A coefficient calculated over twelve months and one calculated over five days can disagree completely, and both are correct about different questions. An intraday trader who reads a one-year figure is looking at a relationship that has nothing to do with the next four hours. Match the lookback to your holding period, and read a short window against a long one: a pair that has decoupled from its usual partner over the last week is telling you something specific is driving it right now.
Correlations also break when a domestic story dominates. A Bank of England decision separates the pound from the euro for a few sessions. A commodity shock separates the Canadian dollar from everything else. Treating a coefficient as a fixed property of a pair is how people build hedges that stop hedging at the worst moment. The tools for watching this are covered in correlation matrix tools, most of which let you set the window and the timeframe yourself.
Net your exposure by currency
The practical technique is simple and almost nobody does it. Break every open position into its two currency legs, add them up, and look at what you are actually holding.
| Position | Long leg | Short leg |
|---|---|---|
| Long EURUSD, 1 unit | EUR +1 | USD -1 |
| Long GBPUSD, 1 unit | GBP +1 | USD -1 |
| Long AUDUSD, 1 unit | AUD +1 | USD -1 |
| Net | EUR +1, GBP +1, AUD +1 | USD -3 |
Three units short the dollar, spread across three currencies. If your rules cap risk per idea at 2%, this book breaches the cap even though no individual position does. The fix is not complicated: reduce each position so the netted exposure fits the limit, or drop to the one pair with the cleanest setup and size it properly.
Correlated positions do not average their volatility, they add it. If you size each trade correctly against its own ATR and then hold three of them on the same theme, the drawdown you should plan for is roughly the sum, not the largest.
The hedge that is not a hedge
The other common use of correlation is deliberate offsetting. Long EURUSD and long USDCHF cancels a good part of the dollar risk, because the dollar sits on opposite sides of the two quotes. What remains is an exposure to the euro against the franc, which is a trade you did not choose and probably have no view on.
You also pay for the privilege twice: two spreads on entry, two on exit, and two overnight financing charges, one of which is likely to be negative. For a small account that cost is a slow drain in exchange for a position nobody wanted. If the aim is less dollar risk, take less dollar risk. Cutting the original position is cheaper, cleaner, and requires no assumption about a coefficient holding.
There is one legitimate version. Pair trading, where a trader deliberately buys one currency and sells a close relative expecting the spread between them to converge, is a real strategy with its own logic. It works because the trader has a view on the relationship itself. Bolting a correlated position onto a losing trade to avoid closing it is a different thing entirely, and it is usually the first move in a much larger mistake.
If you are still building the mental map of which currencies sit where, how pairs are quoted is the place to start, because reading the base and quote correctly is what makes the netting table above obvious rather than fiddly.
"Write down the currency legs of every open trade on one line. Most people find they are holding one position, in three places, at three times the size they intended."
— Alex Onta, Executive Director, SINGUARD
Key Takeaways
- Break open positions into currency legs and net them. The total exposure to one currency is what your risk limit should apply to.
- Match the correlation window to your holding period, and compare a short window against a long one to spot a pair that has decoupled.
- Correlations shift when a domestic event or a risk-off session takes over, usually at the moment you were relying on them.
- Offsetting a losing trade with a negatively correlated pair adds costs and an unwanted cross exposure. Reducing size is the cheaper answer.
Frequently Asked Questions
Why are EURUSD and GBPUSD usually correlated?
They share a quote currency and much of their movement is the dollar rather than the euro or the pound. When the dollar strengthens broadly, both pairs fall together regardless of what is happening in Europe or the United Kingdom. The correlation weakens when a domestic story dominates one of them, such as a Bank of England decision or a euro area political event.
Does trading a negatively correlated pair hedge a position?
Only partially, and it is rarely worth the cost. Long EURUSD and long USDCHF offset a large part of the dollar exposure while leaving you exposed to the euro against the franc, which is probably not the trade you wanted. You also pay two spreads and two sets of financing. Reducing the original position is usually cleaner than adding an offsetting one.
What correlation window should I use?
Match it to how long you hold trades. An intraday trader learns little from a twelve month coefficient, and a position trader learns little from a one hour one. Reading two windows together is more useful than either alone, because a short window that has diverged from a long one is telling you that something specific is driving one of the pairs right now.