Open a long in EURUSD, a long in GBPUSD, and a short in USDJPY, each risking 1 percent, and your platform reports three positions. The market sees one: short the US dollar, at 3 percent. When a strong US inflation print lands, all three stops fill within the same minute, and a trader who believed he was diversified has lost $300 of a $10,000 account on a single headline. Nothing was mispriced and no rule was consciously broken. The risk was simply counted per ticket when it should have been counted per idea.
Where the overlap hides
The dollar is the biggest source. It sits on one side of six of the seven major pairs, so most forex portfolios are secretly a single dollar opinion; EURUSD and GBPUSD alone commonly track each other with intraday correlation above 0.8. Indices are the second cluster: S&P, Nasdaq and DAX move together on risk appetite, and crypto frequently joins them, so two index longs plus a bitcoin long is one sentiment trade in three costumes. Then the commodity links: gold and silver travel together, and oil pushes USDCAD around, so short USDCAD and long oil largely repeat each other. None of this needs a statistics package to manage, though a monthly glance at a correlation matrix is worth the five minutes it takes. The big clusters are stable enough to treat as standing rules, and the rest yields to one honest question per trade.
What stacking does to your numbers
Your risk framework was designed assuming trades are independent. One percent per trade feels safe because losses are not supposed to synchronize. Correlation voids that assumption: at a correlation around 0.85, three 1 percent positions behave like a single position risking nearly 3 percent, while pretending on screen to be diversified. The knock-on effects are worse than the headline number. A $300 daily loss limit can be consumed by one macro event before you can react, turning your circuit breaker into a formality. And the hit arrives as one emotional event, not three small ones, which is exactly the trigger profile for revenge trading. Stacked correlated losses are how disciplined traders have undisciplined afternoons. The same logic applies across time, not just across open tickets: five dollar-theme losses in a row on five different pairs is not five setups failing, it is one thesis failing five times while the journal spreads the blame.
Cap exposure by theme, not ticker
The fix is bookkeeping. Define your themes: dollar direction, risk appetite (indices plus crypto), energy, metals. Cap aggregate open risk per theme at 2 percent, $200 on the $10,000 account. Before every entry, ask the sorting question: what has to happen in the world for this trade to win? If the answer matches an open position, it is the same trade, and there are two honest options. Take only the stronger chart at full size, or split the budget, say 0.5 percent each on EURUSD and GBPUSD, so the theme still totals 1 percent. What is not on the menu is full size on both because they have different names. The 1 percent rule sizes a trade; the theme cap sizes an opinion. Run the count again before every addition, not just at the open: a book that started diversified drifts toward one theme as orders trigger through a session, and the third correlated fill is usually the one that breaks the budget.
Correlation spikes exactly when it hurts
The number you measured in calm markets understates the one that shows up in stress. On quiet days, pairs and indices wander semi-independently; during a risk-off shock, nearly everything converges into one trade as positioning unwinds together. So average correlation flatters your diversification precisely until the day it matters. The practical rule: around scheduled events that touch a whole theme, US CPI, FOMC, nonfarm payrolls for the dollar complex, treat every position in that theme as one position, and reduce total theme exposure to what you would risk on a single trade through news. Diversification measured on calm days is a fair-weather instrument.