The precise implantation of the prosthesis is quite important to have the good long term outcomes. One of the important methods for the implantation is the preoperative templating to determine the selection of implant size, position and alignment. But the preoperative template system sometimes could not give us the reproducibility of the actual size of prostheses which were assessed intra-operatively. One reason for the inaccuracy of the radiograph based templating system is caused by the magnification of the view due to the fixed knee deformity and another is by the rotation view of the prostheses due to tilt shooting radiograph. We are now performing the preoperative CT based 3 dimensional templating system or intra-operative molding template system with virtual bone model, in the addition to the traditional preoperative radiographic templating system. Our aim is to assess the reliability of these templating systems which are traditional radiograph template (2D-template), CT-based template (3D-template) and intra-operative molding template (Free-template), for the selection of correct prosthetic size during operation. We randomly selected 51 patients with 19 osteoarthritis, 30 rheumatoid arthritis, 1 juvenile rheumatoid arthritis and 1 systemic lupus erythematosus, and performed PFC sigma total knee arthroplasty (DePuy). In all 51 knees, the sizes of implants were estimated on the traditional radiograph to get the 2D-template data. The 3D template data were obtained from 25 knees preoperatively by the CT-based navigation system (BrainLab) and Free-template data were collected intra-operatively from 27 knees with the CT-free navigation system (BrainLab). These template data were compared with the actual implanted size of the femur and tibia without any information. The statistical analyses were performed to evaluate the reliability of these templating systems. The precision of 3D-template was not significantly different from that of 2D-template. On the other hand, the reproducibility of Free-template data was significantly higher than the other templating system. In our conclusion, the precise selection of the prostheses could not be achieved with either 2D-template or 3D-template, due to the disappearance of joint space and severe osteophytes. CT-free system could be precisely acquire in-situ bone information of femur and tibia even in the severe deformity knees, and at the same time, CT-free system, which has the kinematics based navigation system, can determine the placement and alignment. Thus CT-free navigation system would be the best assistance during pre- and intra-operation to achieve the long-term good results.