PURPOSE: To provide training guidance on procedure numbers by assessing how the number of previously performed arthroscopic procedures relate to both competent and expert performance in simulated arthroscopic shoulder tasks. METHODS: A cross-sectional study that assessed simulated shoulder arthroscopic performance was undertaken. A total of 45 participants of varying experience performed 2 validated tasks: a simple diagnostic task and a more complex Bankart labral repair task. All participants provided logbook numbers for previously performed arthroscopies. Performance was assessed with the Global Rating Scale and motion analysis. Receiver operating characteristic curve analyses were conducted to identify optimum cut points for task proficiency at both "competent" and "expert" levels. RESULTS: Increasing surgical experience resulted in significantly better performance for both tasks as assessed by Global Rating Scale or motion analysis (P <.0001). Receiver operating characteristic curve analyses demonstrated 52 previous arthroscopies were needed to perform to a competent level at the diagnostic task and 248 to be competent at the complex task. To perform at an expert level, 290 and 476 previous arthroscopies, respectively, were needed. CONCLUSIONS: This study provides quantified guidance for arthroscopic training and highlights the positive relationship between arthroscopic case load and arthroscopic competency. We have estimated that the number of arthroscopies required to achieve competency in a basic arthroscopic task exceed those recommended in some countries. These estimates provide useful guidance to those responsible for training program. CLINICAL RELEVANCE: The numbers to achieve competent arthroscopic performance in the assessed simulated tasks exceed what is recommended and what is possible during surgical training programs in some countries.