Correlation between pairs

Every currency pair is a ratio of two economies, which means pairs sharing a currency share a driver. Long EUR/USD and long GBP/USD look like two positions on two different pairs, but both are substantially short the dollar. If the dollar rallies on a strong jobs number, both lose together, and your carefully-set 1% risk was quietly a 2% bet on one macro event.

Correlation is usually expressed between minus one and plus one. Near plus one, two pairs move together; near minus one, they move oppositely; near zero, they are broadly independent for the period measured. EUR/USD and USD/CHF have historically been strongly negatively correlated, because one has the dollar in the quote and the other has it in the base — that is not a mystery, it is arithmetic.

The trap is that correlation is not constant. Pairs that behave independently for months can snap to near-perfect correlation during a risk-off shock, exactly when your positions are largest and your stops are being tested. Diversification measured in calm markets tends to evaporate in the markets where you needed it.

Diversification measured in calm markets has a habit of evaporating in precisely the markets where you needed it.

Practically: before opening a second position, write down what the trade is actually a bet on. "Long EUR/USD" and "short USD/JPY" are both "short the dollar". If you would not put 2% on a single dollar-weakness trade, do not arrive at 2% by accident through two tickets. Count exposure by currency, not by pair.

Correlation can also be used deliberately. Deliberately negatively-correlated positions can hedge an exposure you cannot exit, and cross-pairs like EUR/GBP let you express a view on two European economies without any dollar exposure at all. The point is not to avoid correlation but to be the one choosing it.