Abstract: The aim of this study was to show that the motion pattern of a well-ridden horse varies less than the motion pattern of an unridden horse. In order to do so, we recorded the motion of two markers, one attached to the dorsal spinous processus of lumbar vertebra L4, the other to the right fore hoof. In total, we measured 21 horses in trot, ridden and unridden, with a fitting and with a non-fitting saddle. After breaking down the entire time series of the three-dimensional motion of the markers into their respective motion cycles, we computed a measure of motion pattern variability for the motion as well as for the derivatives (velocity and acceleration) along each of the three principal dimensions. Two of six variables (velocity and acceleration in the forward direction) displayed a significant discrimination between the ridden and the unridden case, and demonstrated the beneficial effect of a rider on the horse's motion pattern variability. Saddle fit was shown to have also an influence on motion variability: variability of two variables (velocity and of acceleration in forward direction) was significantly lower with a fitting saddle compared to a non-fitting saddle, a third variable (acceleration in the transversal direction) showed a significant difference also. This new method offers an objective evaluation of saddle fit, and a sensitive assessment of the quality of the rider in the moving horse.