Estimator

An estimator is the recipe — a formula or procedure — that turns sample data into an estimate, while an estimate is the actual number it produces this time. “Take the sample average” is an estimator; the value 52,300 it gives on today’s data is the estimate.

We judge estimators by how they behave over many hypothetical samples. A good estimator is roughly unbiased (its average over repeated samples lands on the true value) and low-variance (its answers don’t bounce around wildly). These two goals often trade off, which is the seed of the bias–variance tension you meet later in modelling.