Rolling correlation

What it adds to ordinary correlation

Rolling Correlation (Rolling Correlation) measures the relationship between two assets over a sliding window (thirty or sixty sessions). It turns correlation from a fixed number into a line that breathes over time: you see it tighten, loosen and flip. That movement is the information.

Why the window is the point

A full year of correlation buries shifts. A pair that moved together and split two months ago still looks “correlated” in the long average. A short window shows the breakup as it happens. Changes in correlation often come with changes in the macro regime, before the story shows up in prices.

Practical uses

Checking whether hedges still work (is the inverse still inverse?). Catching moments when a sector breaks away from its market (leadership or a broken relationship). Checking statistical-arbitrage pairs that live on a stable relationship and die when it flips.

Hypothetical example

Two mining stocks had a steady correlation of 0.85 for two years. Within a month the rolling line slid to 0.3 after an acquisition changed the story for one of them. The pairs strategy built on them was dead, and the line announced it before earnings were calculated.

Common mistake

A window that is too short turns the line into noise that jumps every week. Pick the window to match the age of the relationship you study, and use short ones for alerts, not for verdicts.

What the rolling version shows and the fixed one doesn't

It shows when the relationship changed, not just that it changed. A jump in correlation within two weeks says one driver now leads both assets, often monetary-policy news. That timing is what a hedging decision is built on.

Where to see it on the platform

The asset relationships page draws it for any pair you pick, so you see when the relationship was born and when it aged, not just its value today.

The correlation matrix shows the rolling windows and their path.