In this paper we study the effect of providing feeback to college students on their relative performance. In a cohort of 1,000 second-year students, a random sample of them received feedback on their position in the distribution of grades. This information was updated every six months during a three-year period. In the meantime, students in the control group could only access information on their own accumulated grade point average (AGPA). Initially, most students are underconfident: they report that, relatively to other students in the same cohort, their AGPA is lower than it is in reality. After the three-year treatment, students exposed to the treatment are correctly informed about their position in the distribution but the control group is still underconfident. However, the treatment was not associated to an improvement in students' academic performance. On the contrary, treated students experience a significant short-term decrease in their educational performance and a short term increase in self-reported satisfaction. The negative impact on performance fades away after six months. At the same time, we observe no significant impact on the perceived easiness of the courses, self-reported effort, or the choice of elective courses. Our results suggest that providing feedback on relative performance does not necessarily lead to higher performance. Its impact might depend crucially on subjects' priors and on their preferences.