Diversification is the one free lunch in finance — but only when the positions are actually different bets. Open three shorts on EURUSD, GBPUSD, and AUDUSD and your platform shows three trades, three stop-losses, three tidy 1% risks. The market sees one thing: a large bet that the US dollar rises, taken three times. When the dollar falls, all three stops go in the same hour, and the 1% you planned becomes 3% you didn't.
This is the most common invisible risk in retail portfolios, and automation makes it worse before it makes it better.
Why pairs move together
Currencies trade in blocs. Every pair quoted against the dollar shares the dollar as a leg, so a broad dollar move drags them all at once. Beyond shared legs, economies rhyme: the euro and the pound respond to the same European news flow; the Australian and New Zealand dollars ride the same commodity cycle; gold and the yen both catch safe-haven flows. Correlation is not a constant — it strengthens and fades — but during exactly the events that hurt, the stressed market, correlations tend to snap toward one. Diversification is weakest on the day you need it most.
Robots stack it faster
Run the same strategy on three correlated pairs and it will take the same signal three times — same entry hour, same direction, same losing day. Backtest each pair alone and every report looks fine; nothing in a single-pair backtest can reveal that all three drawdowns happen simultaneously. This is how a portfolio of individually reasonable robots produces one unreasonable account. The fix is not more pairs — it is testing the pairs together and counting their combined exposure as one number.
The same trap wears other costumes: two different strategies that both sell volatility, a grid on EURUSD and a martingale on GBPUSD, ten positions that are all, underneath, short the dollar.
Risk per theme, not per ticket
The repair is bookkeeping, not brilliance. Group every position — manual or automated — by its underlying theme, and cap the risk of the group the way you would cap a single trade. If your per-trade limit is 1%, a cluster of dollar shorts is one trade for risk purposes: 1% across the cluster, not 1% each. Correlation tables are freely available and one glance per week is enough. Blown accounts are considerably more expensive.
Related: position sizing, the arithmetic that keeps you alive →