Myth: Three Positions Means Three Trades

The myth: risking 1% per trade is the rule, so three open positions at 1% each is fine — that is 3% at risk, diversified across three instruments. Spreading the risk is the whole point of not putting everything in one trade.

The problem: “three instruments” and “three risks” are different statements, and the difference is a number most traders have never looked up for the instruments they actually trade.

The test: three positions opened together, 1R of risk each, the same 45% win rate on each, run at six levels of return correlation from completely independent to identical. 200,000 simulated sessions per level.

How often does the whole book lose at once?

Three equally-sized positions. The swing column is the standard deviation of the book in units of one position's risk; the loss probabilities come from a standard one-factor Gaussian copula on the stated return correlation.

At zero correlation, all three lose together 16.6% of the time. That is just 0.55 × 0.55 × 0.55 — three independent chances of losing, all landing badly at once — and it is the picture the myth assumes. At 0.85, which is an ordinary daily correlation for US index futures, all three lose together 41.9% of the time. Two days in five, the entire book is red.

The number that actually describes your book

There is a clean way to express this. Take the squared total size of the book and divide it by the book's variance; for positions of equal size that reduces to n ÷ (1 + (n−1)ρ). Three positions at 0.85 correlation is 1.11 independent bets. Not three. Barely more than one.

So the trader who opened ES, NQ and YM at 1% each did not take three 1% trades. They took approximately one 3% trade, at three sets of commissions, with three sets of screens to watch, and with a mental model that told them they were being careful.

The one-day swing tells the same story from the other side: 1.73R at zero correlation, 2.85R at 0.85. The book got 64% more volatile without a single extra contract. Almost all of that extra variance is the shared driver: at ρ = 0.85, covariance between the positions accounts for 63% of the book’s total variance, against zero when they are independent.

What is actually correlated, in practice

You do not need a data feed to spot most of this. The instruments a retail futures trader picks are overwhelmingly the same bet wearing different tickers.

• The US index complex. ES, NQ, YM and RTY (and their micros) are four expressions of one variable. Daily-return correlations routinely sit above 0.85, and in risk-off sessions they converge toward 1 — which is to say, the diversification disappears precisely on the day you needed it.
• Gold and silver. One metals trade with different volatility, not two trades.
• The dollar complex. Short EUR, short GBP and long DXY is one dollar position expressed three times. So is long gold and short DXY, most of the time.
• Energy. WTI and Brent are the same barrel; crude and its products move together.
• The invisible one: time. Three positions opened in the same twenty minutes, ahead of the same release, are correlated through the event even if the instruments are not. An inflation print correlates everything on the board for about four seconds.

Correlation is not a constant, and it moves the wrong way

The genuinely dangerous property is not that correlations are high. It is that they are unstable, and they rise under stress. A pair that has run at 0.3 for a year is not offering 0.3 of protection on a shock day; on a shock day nearly everything with risk in it moves together, and the portfolio you assembled at 0.3 is settled at something much closer to 1.

This is not an abstraction. It is precisely what destroyed Long-Term Capital Management: a book of trades that were uncorrelated in every historical sample and became one trade in the weeks that mattered. The lesson survives the change of scale.

What to do instead

Size the theme, not the ticker. Decide what you are willing to risk on “US equity direction today” — say 1% — and spend it across however many index instruments you like. Three positions at 0.33% each is a real 1% risk. Three at 1% each is a 3% risk with extra commissions.

Count your open risk by driver, not by trade. Before adding a position, name the variable it depends on. If the name matches one already on the book, you are adding, not diversifying.

Assume correlation goes to 1 in the tail. Ask the only question that matters: if everything currently open goes to its stop within the same hour, what is the account down? If that number is unacceptable, the book is too big now — regardless of how uncorrelated it looks in a normal week.

Real diversification is expensive and rare. Genuinely independent risk means different drivers, different time horizons and different sessions. Four screens showing four charts of the same thing is not it, and it is where most over-trading hides — it feels like activity across a portfolio while being one position taken four times.