Currency pair correlations in practice — reading the matrix
Anna runs her account from a Warsaw flat. She has two positions open: long EUR/USD and long GBP/USD, two standard lots each, both behind fifty-pip stops. In her head this is diversification. That afternoon the Fed releases minutes more hawkish than expected. Within forty minutes both stops are taken out in a single dollar move. Damage: close to 8,200 zlotys in half an hour. Anna had one short-dollar bet, opened twice. In this article I explain how to read the correlation matrix, where to check it, and which mistakes cost retail traders money. Anna is illustrative.
What currency pair correlation is and what the Pearson coefficient measures
The formula is simple: take daily percentage changes for each pair, compute means, sum products of deviations, normalise by the two standard deviations. The result sits between minus one and plus one. Pearson built the coefficient for biology; finance picked it up quickly.
What Pearson does not measure matters just as much. Zero means no linear dependency, but does not exclude a non-linear one — two pairs can look uncorrelated for months and then move together during a panic. Correlation is not causation: EUR/USD and GBP/USD correlate at 0.85 because both react to the dollar. Correlation is descriptive — it tells you what already happened, not what will.
Typical correlations of the most traded pairs
The list shows rolling averages for 2020–2025, from daily percentage changes in a ninety-day window. Extreme values (above 0.85 or below minus 0.85) flag pairs with virtually no independent movement. Long-term averages serve as a baseline; re-check current values monthly.
AUD/USD and NZD/USD at 0.90 reflects the proximity of two economies that both export commodities to China and whose central banks move rates in a similar rhythm. EUR/USD and USD/CHF at minus 0.95 is the legacy of the SNB holding the franc close to the euro for decades (the 2011–2015 peg). Gold and the DXY index move opposite because gold is priced in dollars. For the leading pair, see the EUR/USD characteristics overview.
Double-exposure risk — the mistake that cost Anna 8,200 zlotys
Anna's two positions were not independent. At a 0.85 correlation they were one position, taken twice: a hundred-pip dollar move drags eighty-five pips on the other pair in the same direction. Two lots EUR/USD and two lots GBP/USD, from a dollar-exposure standpoint, behave like roughly three and a half lots EUR/USD alone.
The consequence is twofold. Risk is doubled, not reduced — above 0.70 a loss on one pair almost always coincides with a loss on the other. The one-percent rule on a single position is silently broken: Anna, risking 1 percent per position, was in fact risking close to 1.7 percent against the dollar.
Practical rule: never open two positions in the same direction on pairs with |correlation| above 0.70. If you see a strong signal on EUR/USD and GBP/USD at the same time, pick the cleaner technical setup and stick with it. Leaving both open is mistake number one on the retail FX market.
How to compute a rolling 30-day correlation from MT4 and MT5 quotes
Run the calculation once on your own, to understand what sits behind the ready-made matrices. In MT4 or MT5 open the History Center (F2), pick EUR/USD and GBP/USD on the daily interval, save as CSV. In a spreadsheet compute the daily percentage change as closing price divided by previous close, minus one. CORREL on the two series of the last thirty changes returns the rolling 30-day Pearson correlation.
Thirty sessions is about six trading weeks — enough to capture the current regime, short enough to miss a full monetary-policy cycle. Shorter windows are noisier; a ninety-day window is steadier but slower. The compromise: read 30, 60 and 90 days side by side. If the thirty-day window departs sharply from the ninety-day one, a local anomaly is underway — know why before opening a position.
Where to check correlations live — Myfxbook, OANDA and what comes next
Three free tools cover most retail needs. Myfxbook (myfxbook.com/forex-market/correlation) publishes a matrix for around forty pairs in five windows: five minutes, one hour, one day, one week, one month. OANDA's analysis section offers a similar tool that also compares pairs against indices and commodities. Investing.com (investing.com/tools/correlation-calculator) adds six-month and one-year horizons. For broader context, see intermarket analysis on ForexMechanics.com.
Reading the matrix takes three steps. Look at three windows together — 30, 60 and 90 days. If values are close (difference under 0.15), the correlation is stable. Memorise the thresholds: above 0.70 is double-exposure risk, above 0.85 is effectively the same position, below minus 0.70 is potential material for a correlation pair arbitrage strategy. Ignore the five-minute and one-hour windows — noise there is too large outside scalping.
Why everything correlates to one in a crisis
In systemic stress, previously independent assets move together — correlation breakdown, or flight to quality. It shows up in BIS data and Bank of England research on the 2008 crisis. The broader context of how the DXY, oil, gold, and equity indices connect to currency rates is covered in the separate article on intermarket analysis. In a panic, investors cut all risky positions at once: spreads widen, liquidity vanishes, market-making algorithms switch off, and almost everything that is not the dollar or the yen loses value at the same time.
"The greatest contribution of intermarket analysis is the realisation that no market exists in isolation. The dollar's influence on commodities, commodities on bonds, and bonds on stocks means a trader has to watch all four asset classes at once." — John J. Murphy, Intermarket Analysis: Profiting from Global Market Relationships, Wiley, 2004.
Lesson for retail: a 0.3 correlation in calm markets will not protect you in a crisis. On 15 January 2015, when the SNB removed the EUR/CHF floor, the EUR/USD–USD/CHF correlation — minus 0.95 for years — broke down to zero in five minutes, and any hedge built on it produced a double loss. In March 2020, classically uncorrelated assets (gold, treasuries, yen, franc) rose together for days and then fell together. Structural correlations (commodity bloc against dollar bloc) outlast tactical sentiment rotations.
What to do tomorrow — three steps for your own account
Currency pair correlations are not an academic curiosity — they are one of the strongest determinants of real risk in a retail portfolio. A trader without correlation awareness is always taking positions whose true exposure they do not understand.
- Open the correlation matrix in Myfxbook or the OANDA tool today and enter the pairs you actually trade. Check values in three windows — 30, 60 and 90 days. Memorise the thresholds: above 0.70 is doubled exposure, below minus 0.70 is a potential hedge. Write the values into your journal so that a month from now you can compare them and catch any drift.
- Compute a correlation yourself, once, from an MT4 or MT5 export. Pull daily quotes for the two pairs you trade most often over 60 to 90 days, compute daily percentage changes in a spreadsheet, and apply CORREL. You will see how much the value departs from the long-term average and how sensitive it is to the window.
- Adopt the "one position per correlated basket" rule and stress-test the portfolio against a correlation of one. Assume every positively correlated position moves the same direction tomorrow by its average daily range, then check whether the combined loss still fits inside your daily limit. If not, cut a slice before the market does it for you.
Related reading: Risk management basics; correlation pair arbitrage — statistical strategy; DXY and the dollar basket.
Sources & bibliography
-
BIS OTC foreign exchange turnover in April 2022 · oficjalne statystyki Trienialne BIS — struktura globalnych obrotów 7,5 bln USD dziennie www.bis.org ↗
-
BIS Quarterly Review The global foreign exchange market in a higher-volatility environment · analiza zmian w rynku FX po Triennial Survey 2022, kontekst zmienności i korelacji www.bis.org ↗
-
Investing.com Forex correlation calculator · darmowe narzędzie macierzy korelacji par walutowych w pięciu horyzontach czasowych www.investing.com ↗
-
Myfxbook Forex correlation page · aktualne macierze korelacji 30, 60 i 90 dni — referencja branżowa www.myfxbook.com ↗
Frequently asked
How is the Pearson correlation coefficient between two currency pairs actually calculated?
The Pearson correlation coefficient is a classical descriptive statistic invented by Karl Pearson in the late nineteenth century. For two currency pairs the calculation runs like this: take a series of daily percentage changes for each pair over a chosen window (typically 30, 60 or 90 days), compute the arithmetic mean of changes for both pairs, then sum the products of deviations from the mean and normalise the result by the product of the two series' standard deviations. The result always sits between minus one and plus one. Plus one means a perfectly linear positive dependency — when one pair rises by one percent, the other rises by exactly one percent. Minus one is a perfectly linear inverse dependency. Zero means no linear dependency at all. In practice on the FX market the extremes (above 0.95 or below minus 0.95) are rare and typically involve pairs where one currency literally appears on both sides of the quote, such as EUR/USD and USD/CHF (the CHF has historically been pegged in spirit to the euro). You do not have to compute this by hand — tools like Myfxbook and Investing.com refresh their matrices every few minutes. What matters is understanding that Pearson measures a linear relationship, so a low value does not exclude a non-linear dependency (one that only shows up during a market panic, for instance).
Why do EUR/USD and GBP/USD correlate around 0.85 if the economies are different?
Because both pairs carry the US dollar in the denominator — market jargon calls the USD the common leg of the two quotes. When the Federal Reserve telegraphs a hawkish shift, the dollar strengthens against the whole basket, so both the euro and the pound weaken against it simultaneously. EUR/USD drops, GBP/USD drops — hence the positive correlation. The same mechanism runs the other way: when a Non-Farm Payrolls release misses expectations, the dollar weakens on both sides. The second reason is economic: the United Kingdom and the eurozone are two of the most tightly integrated economies in the world. Around 47 percent of UK exports go to the European Union, and 16 percent of eurozone imports come from the UK. When German output slows, UK output tends to slow soon after — and vice versa. The third reason is shared risk-on and risk-off sentiment. In risk-on moments, capital rotates out of the dollar into higher-yielding currencies, and the euro and pound move together. The 0.85 figure is a rolling average for 2020–2025; in individual stretches it has been higher (above 0.90 in the calm months of 2024) or lower (below 0.55 in the first weeks after the Brexit referendum of June 2016, when the pound moved on its own domestic-political logic).
Do currency pair correlations change over time and how often should they be checked?
Correlations are dynamic. The stability you see in headlines like "EUR/USD vs GBP/USD = 0.85" is a long-term average — in individual months the value can run anywhere from 0.95 down to 0.40. Three factors most often disrupt historical correlations. First, political events aimed at one of the currencies: the Brexit referendum of 23 June 2016 dropped the EUR/USD vs GBP/USD correlation from 0.85 to around 0.50 for four months, until the market started trading the pound on the same broad dollar narrative again. Second, central-bank policy divergence: when the Bank of England raises rates on a different cadence than the ECB, the euro–pound correlation weakens. Through 2022–2023, when the BoE and the Fed were better synchronised than the ECB with the Fed, EUR/USD vs GBP/USD correlation occasionally fell below 0.70. Third, geopolitical events that hit asymmetrically — Russia's war in Ukraine in 2022 weighed heavily on the euro because of the eurozone's geographic proximity and energy dependence on Russian commodities, while the pound suffered visibly less. The practical re-check frequency is once a month — open the matrix on Myfxbook or Investing.com across three windows (30, 60 and 90 days) and verify that the values are consistent. When the 30-day window diverges significantly from the 90-day (a difference larger than 0.30), it is a signal that a local anomaly is in progress and you need to identify the cause before opening a new position.
Can correlations be used for simple hedging in a retail portfolio?
In theory yes, in practice with heavy caveats. Hedging through a strongly negative correlation looks tempting: if you are long EUR/USD and worried about a sudden euro depreciation, you could open a second long on USD/CHF (correlation around minus 0.95). When the euro falls, the franc tends to weaken against the dollar as well, so the EUR/USD loss is partly offset by a USD/CHF gain. But this hedge is not free. First, a minus 0.95 correlation does not mean that a hundred-pip move on EUR/USD translates one-for-one into a hundred-pip move on USD/CHF — that ratio is captured by a second parameter, beta, the proportion of volatility between the two pairs. Second, each position carries spread and swap cost, paid on both the primary and the hedge leg. Third — and this is the key point — the hedge stops working exactly when you need it most. During market panics (the SNB peg removal in January 2015, the start of the pandemic in March 2020) historical correlations break down for minutes and the hedge can in extreme moments generate a double loss. For a retail account below the equivalent of 50,000 zlotys, a smarter hedge is cutting the position itself (closing a slice) rather than opening a second, costly leg. Correlation hedging really earns its keep mostly in institutional portfolios with the depth and access to market makers with tight spreads to support it.