Quantitative Methods & RiskPosition Sizing & Capital Allocation
Portfolio Heat & Correlated Risk
Per-trade risk is not portfolio risk
Sizing each position to risk 1% is necessary but not sufficient. If you hold ten positions each risking 1%, your portfolio is not risking 1% — and how much it really risks depends entirely on the correlations between them. Ten independent 1% bets are very different from ten copies of the same 1% bet. The concept that ties this together is portfolio heat: the total risk currently open across all positions, adjusted for how they move together.
Portfolio heat
The sum of risk currently exposed across all open positions — how much the account would lose if every stop were hit at once. A common discipline is to cap total heat (e.g. 4–6% of equity) regardless of how many individual positions are open, so a correlated cluster of stop-outs can't gut the account.
Naive vs. correlation-aware portfolio risk
σ_p = √(wᵀ Σ w) where Σ is the covariance matrix
Hidden concentration: the correlated-cluster trap
Holding long EUR/USD, long GBP/USD, long AUD/USD, and short USD/JPY feels like four diversified positions. It is essentially one large short-dollar bet. If the dollar rallies, all four lose together and your true risk is close to the sum, not the square-root, of the individual risks. Always check correlations BEFORE counting positions as diversification.
A practical portfolio-risk checklist
- 1
Size each position
Apply fixed-fractional risk per trade (e.g. 0.5–1% of equity at the stop).
- 2
Cap total heat
Set a hard ceiling on summed open risk (e.g. 4–6%), independent of position count.
- 3
Group by driver
Bucket positions by their dominant risk factor — dollar, rates, oil, risk-on/off — and limit risk per bucket.
- 4
Stress correlations to +1
Ask: if everything correlated rushed to +1, what is the simultaneous loss? Make sure it is survivable.
- 5
Re-check as positions change
Correlations and concentration shift as you add or close trades; recompute heat on every change.
Cap heat by risk factor, not just by count
A robust rule limits risk per underlying DRIVER, not per ticker. 'No more than 2% of equity at risk to the US dollar across all positions' prevents the correlated-cluster trap that a simple per-trade limit misses entirely. Diversification is about independent drivers, not the number of line items in the book.
Knowledge check
You hold four positions, each sized to risk 1% of equity, and all four are perfectly correlated (ρ = +1). What is the true combined risk if all stops are hit?