The green line is an overfitted model, while the black line is a regularized model. The green line tries too hard to fit all the data in the training set, but will therefore likely have a greater error when introduced to the testing set.