Lookback window
Why does the same market week look mild or extreme depending on how far back a measure reaches?
In plain words
A lookback window is the stretch of past data a measure is calculated from: one week, four weeks, twelve weeks, or a year. The market is unchanged, but the number describing it shifts with the window, because a longer window pulls in more varied history.
Why it matters
Risk limits, percentile rankings and volume comparisons all carry an unstated window. A figure computed over a short window reacts quickly to recent conditions; one computed over a long window is steadier but slower to notice a change in regime. Two reports about the same week can therefore disagree without either being wrong.
In a real week: ETH, week of 12 May 2025
In the week of 12 May 2025, ETH's return of -0.67% ranked at the 45th percentile against 378 weeks of history — near the middle — while its realised volatility of 81.2% a year rests on a window the data does not state. Volume of 14.29bn was 1.59 times the prior 12-week average, yet ETH was still 38.88% below its 52-week high. The four preceding weekly returns were -0.65%, 12.85%, 0.98% and 39.01%. Depending on whether the window was one week, four weeks, twelve weeks or a year, the same week reads as quiet, busy or far below past levels.
A common mistake
Reading a percentile or average as a fixed property of the market. Widening or narrowing the window changes the answer, so a figure is meaningless without its stated span.