Underfitting
When a model is too simple to capture the real pattern in the data, so it does poorly even on the examples it trained on. Both training and test performance stay stuck.
Think of It Like This
Drawing a straight line through data that clearly curves.
Underfitting is the mirror image of overfitting: not enough model capacity, too much regularization, or simply too little training. The tell is high error everywhere — the training set included. The fix is a richer model, better features, or just training longer before you reach for anything fancier. Balancing it against overfitting is the whole game of the bias-variance tradeoff.