Memorizing the practice test
A student who memorizes last year's exact exam aces the practice but bombs the real test with new questions. That's overfitting; splitting data into study material and a fresh test is how we check a model truly learned, not just memorized.
Don't grade your own homework
Split data into a training set (to fit) and a test set (to estimate performance on unseen data). Evaluating on training data flatters the model.
Overfitting
A model overfits when it memorizes noise — high training accuracy but poor test accuracy.