Preload, which is widely applied in linear motion ball guide (LMBG) to eliminate clearance and increase stiffness, gradually decreases owing to wear, resulting in the degradation of the load-bearing capability and dynamic response of LMBG. However, no solution can be found on the modeling of the preload degradation of LMBG considering raceway profile error, ball diameter error, etc. Therefore, this paper presented a new two-stage degradation model to predict the preload variation of LMBG with considering machining errors. The model was experimentally verified with the prediction accuracy much higher than the model considering no machining errors at either of the two wear stages, which demonstrates the effectiveness of considering machining errors. Additionally, the effects of waviness errors, external load, and feed speed on the preload degradation of LMBG were discussed. The simulation results indicate that the preload loss rate rises with the increase of waviness error, external load and feed speed. For obtaining a longer effective service life of LMBG, it is helpful to select appropriate external load and feed speed conditions and improve the processing technique as well as the machined surface quality.