Opening three different trades does not always mean diversifying. If all three depend on the same thing (say, the dollar weakening), you have really opened one bet spread over three positions, and a single move can trigger all three stops at once. This guide explains correlation, how to calculate the combined risk of several trades, why correlation changes over time and how to set a risk limit for the whole portfolio.
What the correlation coefficient measures
The correlation coefficient runs from −1 to +1. A value near +1 means two instruments tend to move in the same direction; near −1, that they tend to move in opposite directions; near 0, that there is no appreciable linear relationship. It is normally calculated on returns (percentage changes) over a period, not on prices, because trending prices can look related when they are not.
A correlation of +0.8 does not mean they move together 80% of the time; it measures the strength of the linear relationship. It also describes only what happened in the measured window, and does not guarantee what comes next.
Pairs that tend to move together (and pairs that oppose each other)
EUR/USD and GBP/USD usually show a high positive correlation because both have the dollar as the quote currency and the European and British economies share similar cycles. Something similar happens with AUD/USD and NZD/USD. At the other extreme, USD/CHF usually has a negative correlation with EUR/USD, since the franc and the euro often move similarly against the dollar and the dollar sits on opposite sides of the two pairs.
Outside forex, gold often moves inversely to the dollar, and US stock indices move as a bloc (the S&P 500 and Nasdaq 100 usually show very high correlations). The actual values vary over time and with the window chosen, so treat them as tendencies, not constants.
Two correlated trades are almost one
Suppose you buy EUR/USD and buy GBP/USD, each risking 1% of the account. If both pairs rise because the dollar weakens, both win; but if the dollar strengthens across the board, both lose at once and you will have lost close to 2% for a single reason.
What looked like a portfolio of two trades at 1% risk is, in practice, one trade with a risk close to 2%. The problem is not that you trade two pairs, but that you count risk trade by trade and do not add up the trades that depend on the same factor.
Example: calculating combined risk
The worst possible loss, if both stops are hit, is always the sum: 1% + 1% = 2%. When correlation is high, that outcome is far more likely. A simple way to estimate the spread of the combined risk for two trades with risks r1 and r2 is √(r1² + r2² + 2 × ρ × r1 × r2), where ρ is the correlation.
With r1 = r2 = 1%: if ρ = +1, the combined risk is √(1 + 1 + 2) = 2%; if ρ = +0.85, it is √(2 + 1.7) = √3.7 ≈ 1.92%; if ρ = 0, it is √2 ≈ 1.41%; and if ρ = −0.9 (for example, buying EUR/USD and buying USD/CHF, which partly offset each other), it is √(2 − 1.8) = √0.2 ≈ 0.45%. With ρ near +1 diversification disappears, and with negative ρ the positions cancel each other out, but they also cancel their profit and you pay two spreads. This is an approximation, not a guarantee: it does not replace calculating the maximum loss.
Net exposure by currency
Every pair is a bet on two currencies at once: buying EUR/USD means buying euros and selling dollars. To see your real risk, add up exposure by currency. Suppose you are long one lot of EUR/USD (100,000 EUR, about 108,000 USD with the price at 1.08), long one lot of GBP/USD (100,000 GBP, about 127,000 USD with the price at 1.27) and short one lot of USD/CHF (100,000 USD).
All three trades are selling dollars: 108,000 + 127,000 + 100,000 = 335,000 USD of net short exposure, even though they look like three different trades. If you had bought EUR/USD and sold GBP/USD instead, the dollar exposure would almost fully offset (−108,000 + 127,000 = +19,000 USD) and what you would really be trading is the relationship between the euro and the pound.