Determining three-dimensional (3D) acetabular orientation is important for several orthopaedic scenarios, but the complex geometries of both pelvis and acetabulum make measurements of orientation unreliable. Acetabular orientation may also differ between the sexes or racial groups. We aimed to (1) establish and evaluate a novel method for measuring 3D acetabular orientation, (2) apply this new method to a large population of Chinese subjects, and (3) report relevant characteristics of native acetabular orientation in this population. We obtained computed tomography scans taken for non-orthopaedic indications in 100 Chinese subjects (50 male, 50 female). A novel algorithm tailored to segmentation of the hip joint was used to construct 3D pelvic models from these scans. We developed a surface-based method to establish a reliable 3D pelvic coordinate system and software to semi-automatically measure 3D acetabular orientation. Differences in various acetabular orientations were compared within and between subjects, between male and female subjects, and between our subjects and subjects previously reported by another group. The reported method was reliable (intraclass correlation coefficient >0.999). Acetabular orientations were symmetrical within subjects, but ranged widely between subjects. The sexes differed significantly in acetabular anteversion (average difference, 3.0°; p < 0.001) and inclination (1.5°; p < 0.03). Acetabular anteversion and inclination were substantially smaller among our Chinese subjects than previously reported for American subjects. Thus, our method was reliable and sensitive, and we detected sex differences in 3D acetabular orientation. Awareness of differences between the sexes and races is the first step towards better reconstruction of the hip joint for all individuals and could also be applied to other orthopaedic scenarios.