The Kinect v2 sensor, developed by Microsoft, enables lowcost 3-D motion analysis without the need of markers or other attachments on the human body. Via free Software Development Kit the spatial location of 25 body joints can be extracted at a frame rate of 30 fps. This study evaluates the accuracy of skeleton joint angles obtained by the Kinect in comparison to the golden standard, the marker based Vicon Nexus. A total of 28 subjects and 16 different static postures are analysed using the Bland-Altmann plot. The absolute median differences over all subjects and postures resolve in less than 10◦ per joint angle. The calculation of the upper body rotation along the longitudinal axis as well as the neck angle appear to be inefficient throughout the Kinect device as they result in high deviations (up to -31.0◦ ± 9.1◦ and 24.0◦ ± 3.5◦, respectively) from the Vicon system in the end-range regions. This can be due to the image reconstruction algorithms of the Kinect sensor, which do not expect body segments in those postures evaluated. The remaining upper body inclinations and joint angles resolve in higher accuracies with a median difference of less than 7.2◦. This features the Kinect to be a promising tool for kinematic analyses outside laboratories, where the use of marker-based motion capture systems is difficult.