Correlated Exposure

Four trades, each risking 1%, feels like 4% at stake and like sensible diversification. If all four are short the dollar, it is one trade risking 4% with extra commission — and the only thing the diversification bought you is the feeling of having it.

Shared currencies, shared driver

Every forex position is two bets: long one currency, short another. Positions that look unrelated on the platform frequently share a leg.

Long EUR/USD, long GBP/USD and long AUD/USD are three different tickets and one position:

Three longs against the same currency. A dollar rally hits all three at once, and it does not matter that the long legs are different — the shared short leg is what is driving them.

Correlation without a shared currency

Some pairs move together with no leg in common, because a third thing drives both.

• AUD/USD and NZD/USD — two commodity currencies with similar exposure to Chinese demand and global risk appetite.
• USD/CAD and oil, inversely. Canada exports it; the pair tends to fall when crude rises.
• XAU/USD and USD/CHF, inversely. Gold and the franc both attract the same safe-haven flow.
• EUR/USD and the S&P 500, loosely and unreliably, through the general appetite for risk.

Aggregating risk across positions

The question to ask of an open book is not "how many trades do I have?" but "what single event closes all of them at a loss?"

A rough method, done by hand in two minutes:

1. List every open position with its risk in percent.
2. Write each one's long and short legs.
3. Total the risk by currency, counting a long leg as positive and a short leg as negative.
4. Read the largest absolute number. That is your real exposure to one currency's move.

A worked book

Four trades, a stated 4% at risk, and the honest reading is a 3% bet that the dollar falls with a 2% euro bet inside it. The yen positions cancel entirely, which means one of those two trades is paying spread and swap to do nothing.

Correlation is rarely perfect, so 3% is an upper bound rather than a certainty — but plan against the upper bound. The cases where correlations reach 1.0 are exactly the days you need the estimate to have been conservative.

Caps per currency and per theme

The fix is a second layer of limits, sitting above the per-trade rule and written down with it.

A theme is a driver rather than a currency: a risk-on move, a commodity story, a central bank meeting. Long AUD/USD, short USD/CAD and long gold share no leg with each other and are all the same bet on a weaker dollar and a stronger commodity complex.

The currency cap then does something useful beyond limiting losses: it forces a choice. When two setups both want the same exposure, take the better one instead of both — which is usually the correct trade anyway.

Correlation that appears only in a crisis

The most expensive property of correlation is that it is unstable in exactly the wrong direction. Pairs that behaved independently for months converge when it matters.

In a genuine risk-off event, the differences between instruments stop mattering and one factor explains almost everything: the rush into dollars, yen and francs, out of everything else. A book that was well diversified on Thursday is one position on Friday.

• Historical correlation is backward-looking. It measures the period it was calculated over, which by definition was not a crisis.
• Liquidity leaves at the same time. Spreads widen, stops fill further away, and the aggregated loss comes in worse than the aggregated risk implied.
• Hedges fail together. The offsetting position you were relying on is priced in the same panic as the position it was offsetting.

What actually helps

1. Assume the worst case in the aggregate. Size the book as though correlations are 1.0 within a theme. If that total is uncomfortable, the book is too big.
2. Cap total open risk regardless of how uncorrelated the positions look on paper.
3. Reduce before scheduled events — central bank decisions, major data, elections — where everything is likely to move on one headline.
4. Do not treat a correlated position as a hedge. A real hedge is the same instrument in the opposite direction; anything else is a second position with its own risk.

None of this requires predicting the crisis. It requires the book to be sized so that a day where everything moves together is a bad day rather than a final one — which is the same rule the whole track has been making, applied to positions in aggregate instead of one at a time.