Training with simulation systems has become a primary alternative for learning the fundamental skills of robotic surgery. However, there exists no consensus regarding a standard training curriculum: sessions defined a priori by expert trainers or self-directed by the trainees feature lack of consistency. This study proposes an adaptive approach that structures the curriculum on the basis of an objective assessment of the trainee's performance. The work comprised an experimental session with 12 participants performing training on virtual reality tasks with the da Vinci Research Kit surgical console. Half of the subjects self-managed their training session, while the others underwent the adaptive training. The final performance of the latter trainees was found to be higher compared to the former (p=0.002), showing how outcome-based, dynamic designs could constitute a promising advance in robotic surgical training.