Students usually have a subjective feeling about their performance after taking an exam so they can imagine the grade they are going to obtain. Besides, students normally have relative opinion of the marks they deserve. We asked students for their self-assessment (expected and deserved grades) after taking some exams in order to compare their estimations with the actual mark. On the other hand, students were asked for the percentage of classes they attended and if the contents of the exam were a reflection of the subject under examination. We use this information to study the relationship between subjective marks and actual exams outcomes. We have used three different statistical procedures in order to explore the relationship between subjective and actual grades. Linear regression, logistic regression and Bayesian nets were used for building models with the interest variables. We show expected marks are related lineally with exam outcomes whereas the percentage of attended classes is related, non-lineally, with the probability of passing the test. In predictive terms, the Bayesian model showed the best fit by predicting the actual mark above 90% of accuracy. These results could be used for weighing actual and subjective marks to get more valid formal assessments.