The Bias-Variance Tradeoff

The bias-variance tradeoff is the single most important idea in supervised learning, and “explain the bias-variance tradeoff” is one of the most common questions in a quant interview. Every model you fit faces the same tension: make it too simple and it misses real structure, make it too flexible and it chases noise. This lesson defines bias, variance, and irreducible error, derives the squared-error decomposition you should be able to write on a whiteboard, works through concrete numbers, and shows how to control the tradeoff in practice. We tie everything back to the interview and to the special danger in finance: overfitting a backtest to noise.