The provision of non-pecuniary incentives in education is a topic that has received much scholarly attention lately. Our paper contributes to this discussion by investigating the effectiveness of grade incentives in increasing student performance. We perform a direct comparison of the two most commonly used grading practices: the absolute (i.e., criterion-referenced) and the relative (i.e., norm-referenced) grading schemes in a large-scale field experiment at a university. We hypothesize that relative grading, by creating a rank-order tournament in the classroom, provides stronger incentives for male students than absolute grading. In the full sample, we find weak support for our hypothesis. Among the more motivated students we find evidence that men indeed score significantly higher on the test when graded on a curve. Female students, irrespective of their motivation, do not increase their scores under relative grading. Since women slightly outperform men under absolute grading, grading on a curve actually narrows the gender gap in performance.