Total knee arthroplasty (TKA) has been shown to improve knee joint function during gait post-operatively. However, there is considerable patient to patient variability, with most gait mechanics metrics not reaching asymptomatic levels. To understand how to target functional improvements with TKA, it is important to identify an optimal set of functional metrics that remain deficient post-TKA. The purpose of this study was to identify which combination of knee joint kinematics and kinetics during gait best discriminate pre-operative gait from postoperative gait, as well as post-operative from asymptomatic. Seventy-three patients scheduled to receive a TKA for severe knee osteoarthritis underwent 3D gait analysis 1 week before and 1 year after surgery. Sixty asymptomatic individuals also underwent analysis. Eleven discrete gait parameters were extracted from the gait kinematic and kinetic waveforms, as previously defined (Astephen et al., J Orthop Res., 2008). Stepwise linear discriminant analyses were used to determine the sets of parameters that optimally separated pre-operative from post-operative gait, and post-operative from asymptomatic gait. Cross-validation was used to quantify group classification error. Knee flexion angle range, knee adduction moment first peak, and gait velocity were included in the optimal discriminant function between the pre- and post-operative groups (P<0.05), with relatively equal standardised canonical coefficients (0.567, −0.501, 0.565 respectively), and a total classification rate of 74%. A number of metrics were included in the discriminant function to optimally separate post-operative and asymptomatic gait function, including the knee flexion angle range, peak stance knee flexion angle, minimum late stance knee extension moment, minimum mid-stance knee adduction moment, and peak knee internal rotation moment (P<0.05). The mid-stance knee adduction moment had the largest standardised canonical coefficients in the function, and 89.5% of cases were correctly classified. Separation of pre and post-operative gait patterns included only three parameters, suggesting that current standard of care TKA significantly improves only walking velocity, knee flexion angle range, and the peak value of the knee adduction moment. A number of gait metrics, which were included in the discriminant function between post-operative and asymptomatic gait, could benefit from further improvement either through rehabilitation or design. With almost 90% classification, separation of post-operative gait function from asymptomatic levels is significant. The consolidation of knee joint function during gait into single, discrete discriminant scores allows for an efficient summary representation of patient-specific (or implant-specific) improvement in gait function from TKA surgery.