Cryptolean

Bitcoin — What Lookahead Normalisation Costs

Largest distortion +0.198 on 2014-10-06 · typical |error| 0.008 · days off by more than 0.05: 11%

As of 5 Aug 2026Metric riskstat-v1Points 14,496Daily close · not real-time

Scroll or drag to move the time axis · hold shift for the value axis · drag either slider

Description

riskstat-v1 · Both lines are the same metric; only the normalisation window differs. The dashed line scores every past day against extremes set years later, so it draws 2015 using 2021's range — a value nobody could have computed at the time. It renders fine and it is a lie. The two converge once the expanding window has seen both extremes, which is why the error is concentrated in the early record and cannot be checked on recent data.

How to read it

Largest distortion +0.198 on 2014-10-06 · typical |error| 0.008 · days off by more than 0.05: 11%

Every value plotted for a past date could have been computed on that date. Normalised series expand their window rather than scaling against the full history, so nothing here is drawn using information that did not exist yet — which is also why the early years of a normalised series are dropped rather than shown against a range of one.

Metric version
riskstat-v1
Export format
web-v4
Computed
5 Aug 2026
Shape
timeseries
Category
Risk metrics
Data sources
Daily close. Free public endpoints, cached before processing; nothing here is real-time.