What Is Correlation Risk? Why Diversification Fails

Correlation risk is the exposure created when positions or accounts that look independent are driven by the same factor and therefore lose together. It is why an apparently diversified book produces a single, undiversified drawdown: the risk was counted once per position, but it arrives all at once.

How it works

The pairwise coefficient belongs to forex pair correlation. At book level what matters is how many independent bets it leaves you with.

Independent bets = n ÷ ( 1 + (n − 1) × ρ )

n = number of positions
ρ = average correlation between them

Three positions each risking 1% of the account:

The trader believed they had risked 1% three times. At ρ = 1 they risked 3% once. Nothing on the platform announces the difference.

Correlated exposure arrives through five channels, and most books carry several at once:

• A shared currency leg. Long EUR/USD, GBP/USD and AUD/USD is one short-dollar position in three costumes.
• One strategy across many symbols. A single Expert Advisor trading twelve pairs generates twelve positions from one signal condition.
• One strategy across many accounts. The same EA copied onto four accounts is one bet at four times the size; total capital at risk is what matters, not capital per account.
• A shared regime dependency. Every trend-following system in a book suffers the same choppy market, whatever it trades.
• A shared venue. Same broker, same price feed, same execution. A feed outage or a requote storm hits every position at once.

And ρ is not a constant: in quiet markets the components decouple and the book genuinely looks diversified, then in a liquidity event they converge toward 1.

Why it matters

Correlation risk breaks the arithmetic position sizing depends on. A 1%-per-trade rule is a promise about the worst case, and it only holds if the trades are independent. Ten correlated positions at 1% is a 10% day through a rule followed exactly.

It hits margin level the same way: correlated positions move into floating loss together, so equity falls across the whole book while used margin stays put, and the stop-out threshold approaches faster than any single-position analysis predicted.

It also makes track records misleading. Four accounts each showing a modest drawdown may have printed all four on the same three days: four data points read separately, one in fact.

What the data shows

On ShowMyTrades (August 2026), 699 users run more than one account and 429 run accounts at more than one broker. Across the public accounts with trading history, median autotrading share is 99% and 53.9% run above 90% automated. These describe accounts published here, not traders in general.

Those figures set the scale. More than half of all users hold multiple accounts and roughly a third spread them across brokers, usually to avoid depending on one counterparty. That diversifies the counterparty, not the strategy: if the same automated system runs on both, two brokers deliver one equity curve twice, and a near-total autotrading share makes that the default case.

The outcomes are consistent. Median deepest drawdown is 9.7%, yet 17.6% of accounts have passed 50%. Falls of that depth are rarely one position going wrong; they are a set of positions that turned out to be the same position.

Where you see it on ShowMyTrades

• The Breakdown Statistics module and its By Magic Number tab, which splits the record by strategy rather than by instrument. Two magic numbers with the same good and bad stretches are one system wearing two labels.
• Custom Analysis, which filters by Symbols and Magic Numbers. Recomputing the figures with one magic number excluded shows how much of the record depended on it. The saved preset list is hidden on public pages; the filters are not.
• The Growth view in the charts viewer, read across two account pages. Two curves with the same peaks and troughs on the same dates are one strategy whatever they are called.
• A portfolio, which merges several accounts into one aggregated view — equity curve, statistics and metrics combined. That total is what correlation risk applies to, not the per-account figures.
• The trades table with its Magic Number and Comment columns, showing which system opened what and when.

Common misunderstandings

• "I trade eight pairs, so I am diversified." If seven quote against the dollar, you hold one dollar position sized eight times.
• "Different brokers means diversified risk." It diversifies counterparty and execution, which is worth having. It does not diversify a strategy running identically on both.
• "Correlation is stable enough to plan around." It is measured in normal conditions and converges toward 1 in the conditions that create large drawdowns.
• "Long one pair and short a correlated one cancels the risk." It replaces a directional bet with a cross bet, and you pay two spreads and two swaps to hold it.

For how to see combined exposure rather than four separate pages, read tracking multiple accounts in a portfolio.

Related terms