Three currency pairs, one dollar bet? Correlation and portfolio risk

Count exposures, not just positions

A trader buys EUR/USD, GBP/USD and AUD/USD, allocating $100 of planned stop risk to each. The screens show three pairs and three tickets. Economically, all three positions benefit from their base currency strengthening against USD and can suffer when USD strengthens against it. One dollar-driven event can therefore pressure the whole group.

The sum is still $300 before trading costs, assuming all three stops fill at their planned prices. Correlation does not secretly turn that arithmetic into $500. It matters for how the losses may arrive together, how equity moves before exits and how much diversification the trader actually has. A common exposure and a measured historical correlation are related ideas, not the same quantity.

What correlation does and does not tell you

Correlation describes how two series vary together in a specified sample. Pearson correlation measures a linear relationship and ranges from −1 to +1 when both series have non-zero variation. Positive values indicate that deviations tend to line up; negative values indicate opposite deviations. Near zero means little linear association in that sample, not proof of independence or safety.

For currencies, compare returns over matching intervals rather than the price levels of two trending charts. A daily return can be calculated as the change in the closing price divided by the previous closing price. Correlation neither shows which series caused the other to move nor provides the probability that both stops will be hit. Shared news, other factors and sampling choices can all contribute.

Moving together does not mean moving identically

The chart and table show five fictional daily return intervals, not observations from a real market. Both series often share a direction, but their move sizes differ and the third interval has opposite signs. This is enough to explain co-movement; it is far too little to estimate a useful live trading relationship. No correlation coefficient is being claimed for these pairs.

Even a strongly positive historical coefficient would not require every daily move to match. Look at the scatter of individual outcomes, the observation count and the sample dates, rather than treating a single summary number as a promise about the next candle.

Every pair contains two currency exposures

EUR/USD expresses how many US dollars one euro buys. EUR is the base currency and USD is the quote currency. A rise means the euro has strengthened relative to the dollar; it does not prove that the euro strengthened against every currency. GBP/USD and AUD/USD share USD as the quote currency but have different base currencies and local drivers.

Buying a pair gives long exposure to the base currency and short exposure to the quote currency; selling reverses both. For a leveraged retail Forex or CFD contract, this is an economic exposure description, not a claim that you physically borrow or receive the two currencies. Check the actual contract size and settlement rules.

The direction of the trade changes the dollar bet

Buying EUR/USD and buying GBP/USD both create short USD exposure. Buying USD/JPY creates long USD exposure, so a dollar strengthening against the yen can help that position while a strengthening against the euro can hurt a EUR/USD buy. Selling USD/JPY reverses the dollar exposure. The location of USD in the symbol and the buy/sell direction both matter.

Correlation between pair returns is not automatically correlation between your position profits. Reversing one fixed-size position reverses the sign of its price contribution to profit or loss. Position sizes, currency conversion, timing, fees and active exits then affect the realised strategy results. Opposite dollar legs can partly offset, but they also leave exposure to the other currencies.

Directions describe the isolated relative-currency move, not a guaranteed news outcome. The table makes no claim about historical pair correlations.

The hidden concentration in three $100 risks

A $10,000 account with three simultaneous 1% allocations has $100 + $100 + $100 = $300 of planned initial stop risk: 3% of that same account reference. That budget must be acceptable as a group. Three individually approved 1% trades do not remain a 1% portfolio simply because they use different pair names.

If the positions share short USD exposure, an adverse dollar event can generate several losing trades close together. This is common-factor risk: several outcomes respond to one underlying influence. It does not require the pairs to have equal returns, identical stops or a coefficient of +1. Conversely, independence would not guarantee that simultaneous losses never occur.

Planned stop loss is not the same as expected loss, maximum possible loss or portfolio volatility. Expected loss would need probabilities and outcome sizes; volatility would need position sizes, individual variability and cross-covariances over a defined horizon. Do not substitute a stop-risk amount for a return standard deviation in a portfolio formula.

Worked example: three buys on a $10,000 account

All prices and sizes below are hypothetical. At entry, both balance and equity are $10,000. Assume one standard lot is 100,000 units of the base currency, one pip is 0.0001 and each USD-quoted pair has a $10 pip value per standard lot. Fees, financing and other costs are omitted so that the price-risk mechanism stays visible.

Using lots = dollar risk / (stop distance × pip value per lot), the 50-pip EUR/USD stop gives 0.20 lots, the 100-pip GBP/USD stop gives 0.10 lots and the 25-pip AUD/USD stop gives 0.40 lots. Each planned loss is $100, but their base-currency quantities and dollar notionals are different. Equal stop budgets do not imply equal sensitivities to a price shock.

At the illustrative entry prices, the short USD legs are $22,000, $12,500 and $26,000, totalling $60,500. This is a currency-notional map, not a loss forecast, margin requirement or dollar-index beta. It shows why counting three tickets or three equal risk percentages misses part of the exposure structure.

Total planned price loss: $300, or 3% of the $10,000 reference. Total short USD notional: $60,500. The notional is not the stop loss or the cash margin posted.

Change the direction, keep the stop budget

In this demonstration, change any buy to a sell. The same size and stop distance are retained, with the stop placed on the adverse side of the new trade. Each planned stop risk remains $100, so the portfolio sum remains $300. The long and short dollar notionals change because the currency legs reverse.

A smaller net USD leg does not certify a safer portfolio. Gross exposure, the EUR/GBP/AUD legs, stop timing and currency-specific events remain. This tool does not estimate live correlation or stress losses; it makes direction and notional concentration visible alongside a fixed stop budget.

A US news shock: the planned exit and a worse fill

Suppose a US announcement changes rate expectations and USD strengthens against the three base currencies. The three buys can come under pressure together. If all three planned stop prices are reached and filled exactly, price losses total $300 and the illustrative account falls to $9,700 before costs. No historical correlation estimate is needed to understand this conditional scenario.

For a separate execution stress example, assume every stop fills 20 pips worse than planned. The extra losses are $40, $20 and $80 because the position sizes differ. Total price loss becomes $440, or 4.4% of the initial reference, leaving $9,560 before costs. These invented fills are neither an expected outcome nor a worst-case bound; gaps and liquidity problems can produce other results.

The actual reaction also depends on what was already priced in, euro-area and UK news, Australian developments, risk sentiment and liquidity. Stronger US data need not produce a uniform dollar rally. For an existing position, use the relevant executable bid or ask and current stop rules; a chart midpoint is not automatically an exit price. Margin close-out can also occur before a planned stop.

Planned total: $300. Stress total: $440. The account values of $9,700 and $9,560 exclude all costs. The 20-pip slippage is an assumption, not a typical value or a cap.

Correlation is not identical risk

Two highly correlated return series can have different volatility and ATR, different stop distances, different pip values and very different dollar risk per lot. ATR describes an instrument's recent range; correlation describes co-movement. Neither determines the other. News sensitivity and liquidity can also differ between the pairs.

The statement ‘correlation is 0.8, therefore risk rises by 80%’ is wrong. The number is a dimensionless description of a particular sample, not a cash-risk multiplier. It does not specify leverage, position direction, size, the horizon or stop execution. A portfolio variability model uses correlations together with those sizes and individual volatilities, with its own assumptions and limits. Here, 0.8 is hypothetical, not a measurement of any named pair.

Correlation is not carved in stone

A relationship can be stronger in one sample and weaker or reversed in another. Daily closes, hourly returns and a strategy's actual holding periods answer different questions. Central-bank surprises, changing interest-rate expectations, political events, commodity moves and changes in volatility can shift which factor dominates.

Under stress, a shared factor can dominate and previously useful offsets can weaken. Correlation does not always rise in every crisis, however. Never turn a rolling matrix into a permanent list of ‘safe’ and ‘unsafe’ pairs. A rolling window updates its sample; it still describes the past and its estimate can be noisy.

How to read a correlation matrix responsibly

Record the data source, pair quote convention, start and end dates, interval, return definition and number of matched observations. Align closing times and time zones, handle missing observations consistently and do not replace a missing price with zero. Mixing daily closes from different providers can create an apparent relationship or conceal one.

Check several relevant windows and the strategy's holding horizon rather than selecting the coefficient that looks most reassuring. A near-zero linear coefficient can coexist with nonlinear dependence or clustered losses in extreme events. If a series has no variation, Pearson correlation is undefined. The matrix alone cannot forecast the loss on the next US announcement.

Other currencies and shared market factors

The principle extends beyond USD. Buying EUR/GBP and EUR/JPY gives two long EUR legs. AUD/USD and NZD/USD can share USD exposure and some regional or global risk drivers, while Australian and New Zealand developments can pull them apart. USD/JPY and USD/CHF share a dollar leg but also expose the account to distinct yen and franc influences.

Buying EUR/USD and buying USD/JPY can resemble long EUR/short JPY exposure if the intermediate USD amounts are appropriately matched. Equal lot counts do not guarantee that match. The relationship between cross rates is an accounting clue, not a promise of identical execution, financing or risk. Even assets without a repeated currency can share a funding or risk-sentiment shock.

No numeric market correlation is assigned to these examples. Direction, period, horizon and market regime can change the relationship. The labels identify exposures to inspect, not safe pair combinations.

Three Forex robots are not automatically three independent strategies

Consider Robot A on EUR/USD, Robot B on GBP/USD and Robot C on AUD/USD. Their code, indicators and entry rules may be different. If they are simultaneously long these pairs, they can still carry one shared short-dollar exposure. Different product names do not remove the economic connection; different trading logic can change when that connection is active.

Inspect overlapping positions, direction, lots, stop distances and aggregate equity, rather than pair names alone. Compare strategy returns on aligned time intervals, including floating profit and loss where available. Correlation of closed trades or cumulative balance curves is not a substitute for the risk of concurrent open positions. If a robot adds positions, widens stops or has no finite stop, a $100 initial estimate may cease to describe its exposure.

Strategies can also share leverage, a news filter, funding currency or the same broker and connection. Some are market factors, others are operational risks; a currency correlation matrix does not capture them all. Test a common adverse event and the interaction of the systems, including margin use and close-out rules, without assuming that all robots must always win or lose together.

The position size calculator is not the whole picture

Our current position size calculator uses account balance, account currency, the chosen risk percentage, stop distance and pip value to estimate one position. It has no list of your open trades and no portfolio correlation matrix. Its account-currency selector does not discover the correct pip value automatically; that input must already be expressed in the selected currency.

A correctly sized individual trade therefore does not automatically create a balanced portfolio. Use the calculator for the single-trade arithmetic, then inventory all simultaneous positions. Keep one consistent account reference for the percentage budget; if your policy uses current equity, account for floating losses rather than silently mixing entry balance with a newer equity figure.

Open the Forex position size calculator

An eight-step check before adding a position

Make this account-level check before the next trade and repeat it when an order fills, a stop moves or a robot adds exposure. Pending orders belong in the relevant stress scenario if one event could activate them together.

1. Count all open positions

   Include manual trades and every robot on the account. Record pending orders that could add to the same exposure during the event being assessed.

2. Write down pairs and directions

   Record buy or sell, actual quantity, entry, current price and stop. A symbol without its direction is an incomplete risk description.

3. Identify repeated currencies

   Split each pair into long and short currency legs. Compare amounts in a consistent reference currency instead of simply counting appearances of USD.

4. Name the common factors

   Consider the dollar, interest-rate expectations, shared regional drivers, funding and risk sentiment. A repeated currency is a clue, not a measured coefficient.

5. Describe one shared adverse event

   Ask which positions could lose together and whether offsets survive different move sizes, stop timing and poorer execution. Include relevant margin stress.

6. Add the planned stop losses

   State the money amounts, reference equity or balance and execution assumptions. For open trades, distinguish loss from entry from the additional loss between current price and stop. Do not count an existing floating loss twice. An absent stop leaves this sum incomplete.

7. Compare with the total budget

   Use the portfolio limit you decided in advance, with room for costs and execution uncertainty. An illustrative 1% per trade or 3% total is not a universal safe limit.

8. Reassess after the new trade

   Check how the proposed position changes gross and net currency legs, stop risk and scenario losses. If the group no longer fits your budget, reconsider its size or whether to add it.

Common mistakes and a more useful decision

Keep separate records for planned stop risk, currency notionals and scenario losses. They answer different questions and should not be merged into one unexplained ‘risk score’. A historical hedge can reduce one sensitivity while adding costs and exposure to another currency.

If three positions already fill the same dollar risk budget, adding a fourth needs a fresh group assessment. Reducing position sizes, choosing fewer overlapping trades or waiting for an exposure to close can each change the portfolio. None guarantees diversification or a profitable outcome. Judge the combined account against your stated limits.

Nine errors to avoid

• Counting three different pairs as three independent bets.
• Approving each position without adding their stop-risk amounts.
• Ignoring repeated USD exposure or confusing buy/sell direction.
• Treating high correlation as permanent or guaranteed co-movement.
• Using an old coefficient without its dates, interval and data source.
• Multiplying money risk mechanically by a correlation coefficient.
• Assuming different robots automatically diversify the account.
• Netting opposite currency legs and declaring all risk eliminated.
• Adding exposure because each ticket fits a limit, while the group does not.

Questions worth settling before trading