Correlation: why your six positions are really one position
Spreading risk across six trades that all depend on the same thing is not diversification. It is one trade at six times the size, and it only reveals itself on the bad day.
Key takeaways
- – Positions that share a driver are one position for risk purposes, however many tickets they occupy
- – Correlations are unstable and tend toward one exactly when you need them not to
- – Size against the shared driver, not against each individual trade
A trader risking one percent on each of six trades believes they are risking six percent spread across six ideas. If those six trades are long EUR/USD, long GBP/USD, long AUD/USD, short USD/CAD, long gold and long an equity index, they are risking roughly six percent on one idea: the dollar weakens and risk appetite holds up. There is no diversification in that book at all.
Where hidden concentration comes from
Currency pairs share a leg. Every pair with USD on one side is partly a dollar trade, so a portfolio of dollar-denominated pairs positioned the same way is a leveraged directional dollar bet. The same logic applies elsewhere: gold, silver and mining equities respond to overlapping drivers; large-cap altcoins mostly track Bitcoin and liquidity conditions; commodity currencies move together with global growth expectations.
The stability problem
The dangerous property of correlation is not that it exists but that it changes. Relationships that look modest in calm conditions converge toward one during stress, because in a genuine deleveraging everything that is not cash is sold at once. Risk models calibrated on quiet periods therefore understate exposure precisely when the understatement matters.
This asymmetry is the whole problem. Correlation helps you least when you need it most.
A practical test
Before adding a position, ask one question: what single event would cause everything currently open to lose at the same time? If a plausible answer exists – the dollar rallies, risk appetite collapses, the front end reprices hawkishly – then the book is that one bet, and total risk should be sized as though it were a single trade.
A crude but effective rule is to cap aggregate risk on any shared driver at roughly what you would accept on one position. Six correlated trades at one percent each become six trades at 0.15 percent each, or fewer trades.
Genuine diversification is harder than it looks
Trades that are truly independent are rare in a market where global liquidity conditions drive most things. Different instruments are not automatically different bets, and different timeframes on the same instrument are not either. Diversification by strategy – trend following alongside mean reversion, for instance – tends to be more real than diversification by ticker, because the two profit from different market behaviours rather than from different names.
Work out your aggregate exposure with the position size calculator, then check what the total represents rather than what each line does.
