Structural break
When does a market stop behaving the way it used to, and how can anyone tell a lasting change from a single bad week?
In plain words
A structural break is a lasting change in the behaviour of a data series: its average level, its swings, or its relationship to other series shift, and the old patterns stop describing the new period. Statisticians locate such breaks by comparing the rules that held before with the rules that hold after.
Why it matters
Risk models, position sizes and hedging ratios are usually built on data from before the break. When the behaviour changes, those settings quietly stop fitting the market, and losses appear without any obvious error in the method that produced them.
In a real week: ETH, week of 4 June 2018
In the week of 4 June 2018, ETH fell 15.42% from 619.66 to 524.11, closing in the lower third of a 24.81% range, 54.78% below its 52-week high. Annualised realised volatility was 86.9% and hourly moves reached -5.01%. With only 16 weeks of history, deciding whether this marked a lasting change was provisional: the sample before it was too short to establish a stable baseline.
A common mistake
Calling any unusually large move a structural break. A break requires evidence that the behaviour persists afterwards, which a single week, however dramatic, cannot show on its own.