Transparency is a feature. This page explains what the rolling series show and why we compute them — in plain terms. The precise formulas and numeric conventions are documented in our internal methodology; the summary here is deliberate, not an omission.
A rolling series turns a single risk figure into a trailing-window curve that a chart can overlay on the return history. As the window slides forward through time, each point on the curve is the same risk statistic computed only from the most recent stretch of returns. This lets you see how a metric evolved rather than collapsing all of history into one number. Crucially, there is no new math involved: every point on the curve reuses the exact same calculation as the corresponding point-in-time figure, so the curve can never drift away from the headline value.
The curve lines up one-for-one with the return series, so its length always equals the input's length and each point sits directly beneath the return it summarizes. Until enough history has accumulated to fill a full window, the earliest points are left blank rather than reported from a partial window — a short window would produce a different, misleading figure. The first reported value appears as soon as one complete window exists, and the final value is exactly the metric applied to the most recent full window. If a single window happens to be degenerate (for example, a stretch with no variation to measure), only that one point is left blank; the rest of the curve is unaffected.
Reading the blanks. A window must contain at least two observations for the underlying statistic to be meaningful, and the window cannot be longer than the available history. A window exactly as long as the full history is valid — it simply yields a single value at the very end.
This curve shows how the annualized volatility of returns rose and fell over time, each point measuring the dispersion of returns within its trailing window and scaling it to an annual figure. It reuses the same annualized-volatility calculation as elsewhere in the terminal, and is always zero or positive.
This curve shows how the risk-adjusted return — average excess return relative to its own variability — evolved across trailing windows, reusing the same Sharpe ratio calculation used point-in-time. A window in which returns never vary has nothing to divide by, so that point is left blank rather than reported as an undefined value.
This curve shows how the asset's sensitivity to its benchmark shifted over time, each point being the trailing-window slope of the asset's returns against the benchmark's — the same market-model beta computed on a moving window. A window in which either the benchmark or the asset is flat leaves the slope undefined, so that point is left blank. The asset and benchmark must cover the same periods to be paired; a length mismatch is rejected rather than guessed at.
