Forex Correlation Matrix
How the major pairs move relative to each other over 30 days, 90 days and a year. The tool for answering a question most traders never ask: are these two positions actually one bet?
By Joey van Diest, founder and editorUpdated
Correlation of daily returns (90 days)
| EUR | GBP | AUD | NZD | USD | USD | USD | EUR | GBP | EUR | AUD | CHF | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| EURUSD | 1 | 0.91 | 0.82 | 0.81 | -0.54 | -0.90 | -0.63 | 0.46 | 0.47 | -0.12 | 0.47 | 0.55 |
| GBPUSD | 0.91 | 1 | 0.80 | 0.80 | -0.54 | -0.86 | -0.65 | 0.36 | 0.58 | -0.53 | 0.45 | 0.51 |
| AUDUSD | 0.82 | 0.80 | 1 | 0.83 | -0.44 | -0.77 | -0.58 | 0.37 | 0.45 | -0.23 | 0.74 | 0.50 |
| NZDUSD | 0.81 | 0.80 | 0.83 | 1 | -0.52 | -0.81 | -0.63 | 0.28 | 0.37 | -0.24 | 0.49 | 0.46 |
| USDJPY | -0.54 | -0.54 | -0.44 | -0.52 | 1 | 0.61 | 0.32 | 0.51 | 0.38 | 0.18 | 0.28 | 0.26 |
| USDCHF | -0.90 | -0.86 | -0.77 | -0.81 | 0.61 | 1 | 0.64 | -0.28 | -0.35 | 0.21 | -0.36 | -0.61 |
| USDCAD | -0.63 | -0.65 | -0.58 | -0.63 | 0.32 | 0.64 | 1 | -0.30 | -0.40 | 0.26 | -0.38 | -0.45 |
| EURJPY | 0.46 | 0.36 | 0.37 | 0.28 | 0.51 | -0.28 | -0.30 | 1 | 0.89 | 0.06 | 0.78 | 0.84 |
| GBPJPY | 0.47 | 0.58 | 0.45 | 0.37 | 0.38 | -0.35 | -0.40 | 0.89 | 1 | -0.41 | 0.76 | 0.81 |
| EURGBP | -0.12 | -0.53 | -0.23 | -0.24 | 0.18 | 0.21 | 0.26 | 0.06 | -0.41 | 1 | -0.11 | -0.08 |
| AUDJPY | 0.47 | 0.45 | 0.74 | 0.49 | 0.28 | -0.36 | -0.38 | 0.78 | 0.76 | -0.11 | 1 | 0.72 |
| CHFJPY | 0.55 | 0.51 | 0.50 | 0.46 | 0.26 | -0.61 | -0.45 | 0.84 | 0.81 | -0.08 | 0.72 | 1 |
Green = moves together, red = moves opposite, pale = little relationship.259 daily observations to 2026-07-23.
Most correlated (90 days)
Most inverse (90 days)
The formula
rt = pricet ÷ pricet−1 − 1
correlation = Σ(x−x̄)(y−ȳ) ÷ √( Σ(x−x̄)² · Σ(y−ȳ)² )
Standard Pearson correlation on daily percentage changes, not on price levels. That distinction matters: correlating raw prices produces impressively high numbers that mostly reflect shared trends rather than day-to-day co-movement, which is the thing that actually determines whether two open positions hurt you at the same moment.
Why this changes position sizing
Suppose you are long EUR/USD and long GBP/USD, sized at 1% risk each, and they are correlated at +0.90. You have not taken two 1% risks; you have taken something close to a single 2% risk on the dollar, because on most days both win or both lose together. Traders discover this the hard way when a broad dollar move takes out several "diversified" positions in one session. The practical rule the matrix supports: when combined exposure is highly correlated, size the group as one idea, then use theposition size calculatorto split that single risk budget across the legs rather than assigning each a full allocation.
Inverse correlation is the mirror image and just as easy to get wrong. Long EUR/USD and long USD/CHF at similar size is close to holding nothing at all, while paying spread on both. If the matrix shows a strongly negative reading between two positions you hold in the same direction, you are probably hedged by accident.
The honest caveats
Correlation is backward-looking, unstable, and says nothing about causation. It also tends to rise toward the extremes precisely during market stress, so the diversification it appears to offer is weakest exactly when it is needed. Compare the 30-day and one-year views before relying on any single number: agreement across horizons suggests a structural relationship, such as two pairs sharing the dollar leg, while disagreement suggests something temporary. For the complementary view of which individual currency is driving these moves, thecurrency strength meterdecomposes pairs into their two legs, and theper-pair pagescarry the underlying one-year history each of these correlations is computed from.
Frequently asked questions
- What does a correlation of +0.90 actually mean?
- That over the window measured, the two pairs moved in the same direction on most days and by proportional amounts. It does not mean one causes the other, and it says nothing about size of move or about tomorrow. In practice a reading above roughly +0.7 is the point at which holding both positions should be treated as one larger position rather than two independent ideas.
- Why are so many pairs correlated?
- Because a currency pair is a ratio of two currencies, and the majors share legs. EUR/USD and GBP/USD both have the dollar on the right, so a broad dollar move pushes both the same way regardless of anything happening in Europe or Britain. That is also why inverse correlations cluster: USD/CHF has the dollar on the left, so it moves opposite to EUR/USD on the same dollar move.
- Is correlation stable?
- No, and treating it as stable is the main way it misleads. Relationships shift with the macro regime and tend to converge toward 1 or -1 during stress, exactly when diversification is supposed to help. Compare the 30-day and one-year columns here: where they disagree sharply, the relationship is regime-dependent and should not be relied on.
- Why daily data rather than intraday?
- Because the source is the ECB's official daily reference rate, which is published once per business day and is free to redistribute with attribution. Intraday feeds that would allow shorter-horizon correlation are commercial products whose terms do not permit republishing on a public site. Daily correlation is the right horizon for position-level risk anyway; intraday correlation mostly measures microstructure.
Method and limitations
Correlations are computed at build time from the same baked ECB daily reference rates behind our converter, using Pearson correlation of daily percentage changes with every pair aligned on a common set of dates. Your visit never queries an upstream API. Because the source publishes one rate per business day, these are daily-horizon correlations; intraday relationships can differ and are not measurable from this data. Correlation is descriptive and unstable, not a forecast. This is an information tool, not trading advice.
Data sources
- European Central Bank reference rates via Frankfurter — 259 daily observations to 2026-07-23
For general information and education only. This is not financial advice and not a recommendation to buy or sell anything. Trading and investing carry risk, including the risk of losing more than your initial outlay. Always verify figures against your broker or the original source before acting on them.
Spotted an error? Email[email protected]and it will be corrected. Maintained byJoey van Diest.