The pelvis roll angle is an important kinematic parameter, commonly used for gait analysis of horses. We compared pelvis roll angle during walk and trot between two methods: inertial measurement unit (IMU) attached on (a) sacrum and (b) each tubercoxae (TC), and validated these using motion capture. Five horses were equipped with two systems: IMU on both TC and sacrum, reflective markers on both TC and rigid triad of markers (RTM) on sacrum. Measurement was done at walk and trot along straight line. Both systems were frame synchronized. Subsequently, pelvis roll anglewas calculated for each stride using (a) rotation angle (around horse’s sagittal axis) of IMU and RTM, and (b) angle between horse’s horizontal plane and the line crosses both TC (marker and IMU). For method validation and comparison, accuracy of pelvis roll angle (in degree) was calculated between IMU and motion capture within methods and between methods using each system. Root mean square error of pelvis roll angle respectively during walk and trot within (a) were 0.33 and 0.67, within (b) were 0.28 and 0.43, between methods using IMU were 2.09 and 3.59, and between methods using motion capture were 1.57 and 3.55. Results indicated high accuracy of IMU measuring pelvisroll angle, while agreement between methods was low, regardless of system. Differences between methods were related to their centers of rotation, which were (a) below and (b) over sacrum. Although single IMU on sacrum can measure pelvis roll angle accurately, none of the methods indicated the actual pelvis roll angle.