The authors trained their model on one course and the used the other course for testing and the RMSE was 3.44 and 2.93 for these different courses respectively. In order to explore whether their model was better than the random guessing they generated exam grades randomly and using kernel and uniform distributions. The linear model showed better performance for both of these courses. Additionally, they grouped data into A, B, C, D, F grades and in this case the model performance decreased as it wasn't much better than random guessing.