Bankart lesion is one of the most common lesions of the glenohumeral joint. Several double-row suture methods were reported for Bankart repair, which could provide more stability, yet more motion limitation and complications. Therefore, we introduced a new double-row Bankart repair technique, key point double-row suture which used one anchor in the medial line. The purpose of this article is to investigate the clinical outcomes of this new method and to compare it with single-row suture. Seventy-eight patients receiving key point double-row suture or single-row suture from October 2010 to June 2014 were collected retrospectively. The basic information including gender, age, dominant arm, and number of episodes of instability was collected. Before surgery, the glenoid bone loss was measured from the CT scan. The visual analogue scale, American shoulder and elbow surgeons, the University of California at Los Angeles shoulder scale, and subjective shoulder value were valued before surgery and at the last follow-up. Forty-four patients (24 patients receiving single-row suture and 20 patients receiving key point double-row suture) were followed up successfully. The follow-up period was 9.2 ± 1.1 years (range, 7.8-11.4 years). At the last follow-up, no significant differences were detected for any of the clinical scores. The recurrence rate was 12.5% for the single-row group and 10% for the double-row group, respectively (p = 0.795) 14 patients (31.8%) in the single-row group and nine patients (26.5%) in the double-row group were tested for active range of motion. A statistically significant difference was found only for the internal rotation at 90° abduction (48.9° for single-row and 76.7° for key point double-row, p = 0.033). The key point double-row sutures for Bankart lesions could achieve similar long-term outcomes compared with single-row suture, and one medial anchor did not result in a limited range of motion. The low recurrence rate and previous biomechanical results also indicate the key point double-row suture is a reliable method.