Speed climbing will appear at the Olympic Games in Tokyo. It is an exciting discipline that offers excellent potential for computer-aided processing and analysis. With the recent advances in human pose estimation, it is possible to extract semi-precise 2D skeleton body joint configurations in individual frames of speed climbing videos. However, the estimated skeletons suffer from being extracted at various distances, angles, and zoom settings of the camera, which reduces their inter-comparability. Moreover, these factors change dynamically even during the single run, which hinders data analyzability. This work aims to neutralize these factors by transforming the absolute-view 2D skeletons of speed climbing performance to the virtual reference speed-climbing wall that has the exact proportions of its real-world counterpart. The work will focus on three main directions: i) hold detection, in order to detect salient handholds in the video, ii) hold tracking, to obtain coherent spatio-temporal tracks of individual holds, and iii) point set registration, to find the transformation between the sets of detected and virtual holds at any time moment of the video. Finally, the transformation can be applied to skeleton data to ensure their uniformity.