The generation of correct query execution plans relies on up-to-date statistics.
For database tables that contain millions of rows, frequent statistics runs might not be necessary if the distribution does not change much.