Classical marker-based roentgen stereophotogrammetric analysis (RSA) is an accurate method of measuring in vivo implant migration. A disadvantage of the method is the necessity of placing tantalum markers on the implant, which constitutes additional manufacturing and certification effort. Model-based RSA (MBRSA) is a method by which pose-estimation of geometric surface-models of the implant is used to detect implant migration. The placement of prosthesis markers is thus no longer necessary. The accuracy of the pose-estimation algorithms used depends on the geometry of the prosthesis as well as the accuracy of the surface models used. The goal of this study was thus to evaluate the experimental accuracy and precision of the MBRSA method for four different, but typical prosthesis geometries, that are commonly implanted. Is there a relationship existing between the accuracy of MBRSA and prosthesis geometries? Four different prosthesis geometries were investigated: one femoral and one tibial total knee arthroplasty (TKA) component and two different femoral stem total hip arthroplasty (THA) components. An experimental phantom model was used to simulate two different implant migration protocols, whereby the implant was moved relative to the surrounding bone (relative prosthesis-bone motion (RM)), or, similar to the double-repeated measures performed to assess accuracy clinically, both the prosthesis and the surrounding bone model (zero relative prosthesis-bone motion (ZRM)) were moved. Motions were performed about three translational and three rotational axes, respectively. The maximum 95% confidence interval (CI) for MBRSA of all four prosthesis investigated was better than −0.034 to 0.107 mm for in-plane and −0.217 to 0.069 mm for out-of-plane translation, and from −0.038 deg to 0.162 deg for in-plane and from −1.316 deg to 0.071 deg for out-of-plane rotation, with no clear differences between the ZRM and RM protocols observed. Accuracy in translation was similar between TKA and THA components, whereas rotational accuracy about the long axis of the hip stem THA components was worse than the TKA components. The data suggest that accuracy and precision of MBRSA seem to be equivalent to the classical marker-based RSA method, at least for the nonsymmetric implant geometries investigated in this study. The model-based method thus allows the accurate measurement of implant migration without requiring prosthesis markers, and thus presents new opportunities for measuring implant migration where financial or geometric considerations of marker placement have thus far been prohibitive factors.