Advertisement IC Markets

When Correlations Break: Diversification Failing in a Crisis

Real-time market analysis, live trade entries, order flow commentary, and daily setups for the London, New York, and Asian session overlaps.
Post Reply
Fairman
Posts: 6955
Joined: Tue Jul 21, 2026 7:11 am
Location: Abuja

When Correlations Break: Diversification Failing in a Crisis

Post by Fairman »

When Correlations Break: Diversification Failing in a Crisis

Correlations between currency pairs, and between currencies and other assets, are not fixed. In crises, they often jump.
diagram.png
diagram.png (37.89 KiB) Viewed 223 times
During normal times, you might hold positions in several pairs and believe they diversify one another. Under stress, investors sell risk assets broadly and buy safe havens. Many positions then move together, as risk sentiment dominates individual fundamentals. The correlations rise toward one, and diversification shrinks exactly when you need it.

In March 2020, for example, the dollar surged against most currencies and stocks fell sharply, with many typical relationships distorted by a scramble for dollar liquidity.

Other patterns also reverse. A currency that was a safe haven in one episode may fail in another, depending on the cause of stress.

Implications for risk management are important. Do not rely on historical correlations to estimate worst-case risk. Stress-test your positions by assuming that correlated pairs move together, and that stops may slip.

Consider the total exposure by theme, such as risk-on versus risk-off, not only by currency.

Use position limits and total risk caps that remain valid when correlations go to one.

Hold more cash, or lower leverage, during periods of heightened uncertainty.

After events, review how your positions behaved, and update your assumptions.

Practical step: calculate your total loss if every open position hit its stop at once, and compare with your risk cap.
It’s Fairman :geek:
Recommended broker for automated trading & scalping IC Markets
Fairman
Posts: 6955
Joined: Tue Jul 21, 2026 7:11 am
Location: Abuja

Re: When Correlations Break: Diversification Failing in a Crisis

Post by Fairman »

The everything-hits-its-stop calculation is a good gut check. If that number scares you, your total risk is too high, no matter how diversified it looks. Add up the risk of every open position as if all stops trigger at once, then add a bit for slippage, maybe 20 percent extra since stops slip in fast markets. With five trades at 1 percent each, that's about 6 percent of the account in a bad hour. If losing that in one afternoon would make you change your plan or panic, reduce the number of open trades or the risk per trade. Diversification shows up most when you don't need it.
It’s Fairman :geek:
Fairman
Posts: 6955
Joined: Tue Jul 21, 2026 7:11 am
Location: Abuja

Re: When Correlations Break: Diversification Failing in a Crisis

Post by Fairman »

Grouping positions by theme instead of by pair is a much better view. Long AUD, long NZD and short JPY are three tickets but one risk-on position. Common themes are risk-on versus risk-off, dollar strength, and commodity currencies. I write each open trade under a theme and cap each theme, say 2 percent of total risk at most. Then a fourth trade that looks like a new idea but sits in an already full theme gets skipped or reduced. During a crisis these themes can merge too. In March 2020 nearly everything became a dollar trade, so check the totals by currency as well.
It’s Fairman :geek:
Shadow Trader
Posts: 5594
Joined: Wed Sep 09, 2026 5:19 pm

Re: When Correlations Break: Diversification Failing in a Crisis

Post by Shadow Trader »

The gap between the everyday estimate and the crisis case can be put in one formula. With equal risk on each trade, combined risk grows with the square root of the count plus a correlation term, so the same five trades look very different at different correlations.
img_13497_r1.png
img_13497_r1.png (63.01 KiB) Viewed 177 times
Say five positions risk 1 percent each. In the chart (illustration), a pairwise correlation of 0.3 gives combined risk of about 3.3 percent, 0.6 gives about 4.1, and at a correlation of one it is the full 5, or 6 with the 20 percent slippage allowance above.

A 4 percent cap set from the calm reading looks comfortable at 0.3. The stress case then sits 50 percent above that cap, and nothing about the trades changed except the correlations.

Correlations measured over the last three months are a poor input for this, because the calm months dominate the window. Sizing the cap against the correlation of one case, and treating anything lower as a bonus, keeps the cap honest.

How do you treat a position in gold or an index alongside the currency trades? In a dollar scramble those can join the same theme too, even though they sit outside the currency list.
FXS Prop Desk
Posts: 1689
Joined: Wed Oct 07, 2026 9:52 am

Re: When Correlations Break: Diversification Failing in a Crisis

Post by FXS Prop Desk »

A practical check for traders running several positions is to look at what the portfolio did during a past stress period, for example the COVID sell-off in spring 2020. Applying current position sizes to price changes from those weeks shows how much the account would lose if correlations moved toward one. It is often a much bigger number than the sum of individual stop losses suggests.
FXS Official Team | The Idea Lab & Strategy
Shadow Trader
Posts: 5594
Joined: Wed Sep 09, 2026 5:19 pm

Re: When Correlations Break: Diversification Failing in a Crisis

Post by Shadow Trader »

For scalpers the same problem shows up inside a single hour. A risk off headline can trigger setups on several related pairs within minutes, and each looks like its own trade.

Say four scalps at 0.5 percent each fire on AUDUSD, NZDUSD, AUDJPY and CADJPY inside 12 minutes. In the chart (illustration), the stops fill at 0.5, 0.6, 0.55 and 0.65 percent, so the four losses add to 2.3 percent and go through a 2 percent daily limit.
img_13497_r2.png
img_13497_r2.png (66.95 KiB) Viewed 151 times
No single trade broke a rule, and a theme cap of 2 percent on open risk would have allowed all four. The cap was set at the daily limit, and slippage did the rest.

Setting the theme cap below the daily limit by the expected slippage closes that gap. With the 20 percent allowance mentioned above, a 2 percent limit points to open risk of about 1.67 percent per theme, which here means three trades, not four.

A time rule helps too. One new entry per theme in any 15 minute window means the second trade is taken knowing how the first went.

How do people group the yen crosses in a dollar scramble? A short AUDJPY is a risk off trade and a yen trade at once, so it can sit in two themes.
Shadow Trader
Posts: 5594
Joined: Wed Sep 09, 2026 5:19 pm

Re: When Correlations Break: Diversification Failing in a Crisis

Post by Shadow Trader »

Currency pairs often show stable correlations in calm periods. EURUSD and GBPUSD, for example, tend to move together, while USDCHF moves opposite to EURUSD. During stress, these relationships can break down or reverse, catching traders who rely on diversification across correlated positions.

A scalper running positions in several pairs may believe the portfolio is diversified, only to find that all positions lose simultaneously when the dollar surges. The apparent diversification was in fact a single bet on dollar direction.

Monitoring rolling correlation over short windows, such as 30 or 60 minutes, reveals changes in real time. Many platforms and add-on tools can display correlation matrices. Sizing positions based on combined exposure to each currency, not on number of trades, offers a more honest picture of risk. Reviewing past losing days to see whether correlated positions contributed helps calibrate limits. Exposure to a single currency across multiple pairs deserves explicit caps.
Post Reply