Stationarity: Ground Rules for Modeling

A stationary series plays by the same statistical rules forever: its mean stays put, its spread stays put, and how today relates to yesterday is the same this year as last. Nothing about when you're looking changes what you expect to see.

Most classical models quietly assume stationarity, because they extrapolate — and extrapolation only works if the rules stay fixed. A series with a trend, growing variance, or seasonality is a series whose rules keep shifting under your feet.