Drawdown analysis
Course 04 ยท Portfolio structure

Correlation turns positions into a portfolio.

Measure shared movement, find hidden concentration, and understand why diversification can disappear during stress.

Original correlation map

Different names do not guarantee different risks

Diversification comes from distinct return drivers. A correlation matrix helps expose positions that may respond to the same shock.

US stocks
EU stocks
Gold
Bonds
US stocks
1.00
0.72
0.18
-0.12
EU stocks
0.72
1.00
0.35
-0.08
Gold
0.18
0.35
1.00
0.24
Bonds
-0.12
-0.08
0.24
1.00

Pearson correlation

Measures linear co-movement from -1 to +1. It is useful, but can miss nonlinear dependence.

Rank correlation

Spearman correlation compares ranks and can capture monotonic relationships less sensitive to outliers.

Rolling correlation

Recalculates relationships through time so regime changes and stress convergence become visible.

Stress behaviour

Correlation is conditional

Relationships estimated in calm periods can converge toward +1 when investors sell risk broadly. Always inspect rolling and stressed samples.

Map each position to its economic driver.
Aggregate risk across instruments sharing that driver.
Re-estimate correlation across multiple windows.
Stress the portfolio with higher correlation and volatility together.
Interactive lab

Two-asset diversification lab

See how weights, individual volatility, and correlation combine into portfolio volatility.

Asset B receives the remaining 50% weight.

Portfolio result

14.19%

estimated portfolio volatility

Diversification reduction

3.31 pts

Difference from the simple weighted average of standalone volatilities. Correlation is an estimate, not a permanent property.

Portfolio risk is more than the sum of positions

01

Concentration

Several trades depend on the same factor.

02

Covariance

Volatility and correlation interact with portfolio weights.

03

Breakdown

Relationships change precisely when protection is most valuable.

Lesson complete

Diversify drivers, not labels

Use rolling estimates, factor grouping, and stress assumptions before trusting a low correlation number.

Final lesson: Risk-adjusted returns