Overfitting is a condition of a machine learning model where the model has learned the patterns from the training data too well, so much so that the model can not generalise to new data.