Background The accuracy of various navigation systems used for total hip arthroplasty has been described, but no publications reported the accuracy of cup orientation in computed tomography (CT)-based 2D-3D (two-dimensional to three-dimensional) matched navigation. Methods In a prospective, randomized controlled study, 80 hips including 44 with developmental dysplasia of the hips were divided into a CT-based 2D-3D matched navigation group (2D-3D group) and a paired-point matched navigation group (PPM group). The accuracy of cup orientation (absolute difference between the intraoperative record and the postoperative measurement) was compared between groups. Additionally, multiple logistic regression analysis was performed to evaluate patient factors affecting the accuracy of cup orientation in each navigation. Results The accuracy of cup inclination was 2.5° ± 2.2° in the 2D-3D group and 4.6° ± 3.3° in the PPM group (P =.0016). The accuracy of cup anteversion was 2.3° ± 1.7° in the 2D-3D group and 4.4° ± 3.3° in the PPM group (P =.0009). In the PPM group, the presence of roof osteophytes decreased the accuracy of cup inclination (odds ratio 8.27, P =.0140) and the absolute value of pelvic tilt had a negative influence on the accuracy of cup anteversion (odds ratio 1.27, P =.0222). In the 2D-3D group, patient factors had no effect on the accuracy of cup orientation. Conclusion The accuracy of cup positioning in CT-based 2D-3D matched navigation was better than in paired-point matched navigation, and was not affected by patient factors. It is a useful system for even severely deformed pelvises such as developmental dysplasia of the hips.