What Is Correlation Risk in Forex?

The correlation coefficient compresses the relationship between two currency pairs into a single number between −1 and +1. It is calculated from returns— the bar-to-bar changes of each pair over a chosen lookback window, such as the last 50 daily closes — rather than from the prices themselves.

A reading of +1 means the two return series moved in lockstep over that window; −1 means they moved exactly opposite; 0 means no measurable linear relationship. Real pairs live between the extremes, and the figure depends heavily on how it is measured: the same two pairs can show +0.85 on 50 daily closes and +0.40 on weekly data covering the same months. A correlation is a description of one slice of the past, not a property of the pairs.

Some relationships recur often enough to have become folklore. EUR/USD and GBP/USD have historically tended to be positively correlated: both express a European currency against the dollar, and the underlying economies are closely linked. EUR/USD and USD/CHF have historically tended to be negatively correlated, largely because the dollar sits on opposite sides of the two quotes.

Co-movement is the rule rather than the exception in FX because so many pairs share one leg. In BIS survey data, the US dollar is on one side of roughly nine out of ten trades — so a book of several open positions usually carries a net dollar view, whether or not the trader meant to take one. What drives the dollar on a given day is a separate topic; for risk purposes the point is simply that one driver can move many of your symbols at once.

Consider a $20,000 account with a strict 1% per-trade limit, running three simultaneous positions: long EUR/USD, long GBP/USD and short USD/JPY, each sized so that its stop loses $200. On the order tickets these look like three independent decisions on three different symbols.

Every one of those positions profits if the dollar weakens and loses if it strengthens. When the correlations between the pairs are running high — as they often do when a single dollar-wide move dominates the market — one broad dollar rally can push all three trades to their stops in the same session. The account does not experience three separate 1% events; it experiences something close to one position risking about 3% on a single view: dollar down.

Nothing in the per-trade rule was violated. That is precisely the problem — a per-trade limit caps each ticket in isolation and says nothing about how many tickets express the same underlying exposure.

Portfolio heat is the sum of the open risk across every position: the total percentage of the account lost if every stop were hit at once. It is the natural companion to per-trade sizing. The per-trade figure comes from the position-sizing chain — risk budget, stop distance, pip value, lots — while heat simply adds those figures up across the whole book.

Heat and correlation answer different questions. Heat is the worst-case ceiling; correlation is how likely the worst cases are to arrive together. Three genuinely unrelated trades at 1% each rarely lose simultaneously, so 3% of heat behaves like three small, mostly independent risks. The same 3% on strongly correlated trades behaves like one large risk — the same ceiling, with very different odds of hitting it.

Correlation estimates are least reliable exactly when they matter most. In calm conditions, pairs drift on their own local stories and measured correlations stay moderate. In stress periods — a volatility spike, a sudden flight to perceived safety — many pairs collapse onto a single risk-on/risk-off axis, and co-movements that averaged +0.4 over a quiet quarter can run far higher for days at a time.

The practical consequence is uncomfortable: the diversification benefit between pairs is weakest when protection is needed most, so losses arrive in clusters rather than one at a time. Clustered losses are also the mechanism behind ruin estimates — how streaks compound at a given risk per trade is the subject of the risk-of-ruin guide.

Correlation risk leaves a visible footprint in a trading history. The pattern to look for is simultaneous positions that share a currency and a direction — and account losses that arrived in lumps rather than trade by trade.

• List the periods when several positions were open at once, and note which currencies they shared.
• Net the direction of each cluster: three positions that all gain when the dollar falls are one dollar view, whatever the symbols on the tickets.
• Compare the dates of your worst equity declines with the cluster dates — concentrated exposure tends to show up exactly there.
• Re-express risk per cluster rather than per ticket: a 1% limit on each of four same-direction trades is a 4% cluster.

Reviewing your own MetaTrader positions this way — grouped by overlapping open time and shared currencies — turns an invisible exposure into a measurable one. If your largest peak-to-trough declines line up with your most crowded clusters, the per-trade number was never the figure doing the damage.