Total knee arthroplasty does not restore the full range of motion of the knee. Retrospective clinical studies on knee kinematics suffer from multiple biases because the various parameters involved, that is, the posterior condylar offset, the tibial slope, and the condylar roll-back, are not individualized. The present study, based on the geometry of knee flexion, shows that a 3-mm decrease of the posterior condylar offset could reduce knee flexion by 10 degrees before the occurrence of tibiofemoral impingement. In addition, the simultaneous decrease of the tibial slope by 5 degrees could reduce the flexion by a further 5 degrees. These effects could be reinforced if the paradoxical condylar roll-forward was made to exceed 10 mm. Finally, decreasing the condylar offset in a prosthesis with a paradoxical roll-forward and a neutral tibial slope could reduce maximum obtainable knee flexion before impingement by as much as 30 degrees.