Bias-variance tradeoff
Also: bias–variance tradeoff, bias-variance trade-off, bias-variance decomposition
Expected test error splits into bias², variance and irreducible noise. Increasing model complexity usually lowers bias but raises variance, so the best model sits at the complexity where their sum is smallest.
E[(y − ŷ)²] = Bias² + Variance + σ²_noise
Example
A depth-1 decision tree gives a coarse boundary (high bias, underfits); a depth-20 tree traces every noisy point (high variance, overfits); a depth around 4 is often the sweet spot.