We prospectively evaluated 61 patients treated arthroscopically for anterior instability of the shoulder at a mean follow-up of 44.5 months (24 to 100) using the Rowe scale. Those with post-operative dislocation or subluxation were considered to be failures. Logistic regression analysis was used to identify patients at increased risk of recurrence in order to develop a suitable selection system. The mean Rowe score improved from 45 pre-operatively to 86 at follow-up (p < 0.001). At least one episode of post-operative instability occurred in 11 patients (18%), although their stability improved (p = 0.018), and only three required revision. Subjectively, eight patients were dissatisfied. Age younger than 28 years, ligamentous laxity, the presence of a fracture of the glenoid rim involving more than 15% of the articular surface, and post-operative participation in contact or overhead sports were associated with a higher risk of recurrence, and scored 1, 1, 5 and 1 point, respectively. Those patients with a total score of two or more points had a relative risk of recurrence of 43% and should be treated by open surgery.