The tennis serves of one tennis player, fit with upper limb and racket retro-reflective markers, were captured with a Vicon motion analysis system recording at 500 Hz. The raw racket tip marker displacement and velocity were used as criterion data to compare three different endpoint treatments and two different filters. The 2nd-order polynomial proved to be the least erroneous extrapolation technique and the quintic spline filter was the most appropriate filter. The previously performed "smoothing through impact" method, using a quintic spline filter, underestimated the racket velocity (9.1%) at the time of impact. The polynomial extrapolation method remained effective when noise was added to the marker trajectories.