Self-assessment activities are increasingly popular in the classrooms. But regarding self-assessment accuracy, mixed results are reported without clear reasons for this. This article addresses the impact of gender and performance on self-assessment accuracy. To evaluate self-assessment accuracy, peer and professor assessments are used as reference. A research project, relating to the assessment of oral presentations using scoring rubrics, is conducted in two university degrees. Data from 155 self-assessments and more than 12,000 peer and professor assessments are gathered. The analysis differentiates by degree, gender and level of performance in the assessed competency. The results show that self-assessment accuracy is low and related to the student's gender, that men rate themselves higher than women do, and that even using a scoring rubric, students receiving higher marks from professors are more accurate than students receiving lower marks.