Aims. Hip arthroplasty does not always restore normal anatomy. This is due to inaccurate surgery or lack of stem sizes. We evaluated the aptitude of four total hip arthroplasty systems to restore an anatomical and medialized hip rotation centre. Methods. Using 3D templating software in 49 CT scans of non-deformed femora, we virtually implanted: 1) small uncemented calcar-guided stems with two offset options (Optimys, Mathys), 2) uncemented straight stems with two offset options (Summit, DePuy Synthes), 3) cemented undersized stems (Exeter philosophy) with three offset options (CPT, ZimmerBiomet), and 4) cemented line-to-line stems (Kerboul philosophy) with proportional offsets (Centris, Mathys). We measured the distance between the templated and the anatomical and 5 mm medialized hip rotation centre. Results. Both rotation centres could be restored within 5 mm in 94% and 92% of cases, respectively. The cemented undersized stem performed best, combining freedom of stem positioning and a large offset range. The uncemented straight stem performed well because of its large and well-chosen offset range, and despite the need for cortical bone contact limiting stem positioning. The cemented line-to-line stem performed less well due to a small range of sizes and offsets. The uncemented calcar-guided stem performed worst, despite 24 sizes and a large and well-chosen offset range. This was attributed to the calcar curvature restricting the stem insertion depth along the femoral axis. Conclusion. In the majority of non-deformed femora, leg length, offset, and anteversion can be restored accurately with non-modular stems during 3D templating. Failure to restore hip biomechanics is mostly due to surgical inaccuracy. Small calcar guided stems offer no advantage to restore hip biomechanics compared to more traditional designs. Cite this article: Bone Jt Open 2021;2(7):476–485