Introduction
A full distribution can be complicated. Two numbers summarise it surprisingly well: the expected value (where it is centred) and the variance (how spread out it is). Loss functions are expectations, risk is an expectation, and the bias–variance trade-off is literally about these two quantities.