Bias–variance

The bias–variance tradeoff describes the two ways a model can get predictions wrong. A model with high bias is too simple to capture the real pattern, like fitting a straight line to a curve, so it’s wrong even on the data it trained on: that’s underfitting. A model with high variance is so flexible that it memorizes the noise in its training data, so it looks great in training but its predictions swing wildly on new data: that’s overfitting. Making a model more complex usually lowers bias and raises variance. You find the sweet spot by checking error on held-out data the model never trained on.