Towards statistical analysis of predictive parameters in competitive speed climbing

Abstract Competitive sport climbing progressed massively within the last quarter century. Development of technology enabling qualitative and quantitative analysis is required to withstand the challenges for athletes and trainers. This paper deals with the statistical study of a data set generated by the application of several image processing algorithms and neural networks on competition recordings. Therefore, calculated parameters are combined with random variables for the implementation of a linear mixed effect model. The resulting model enables the prediction of the end time of different athletes and the determination of its correlation with the input variables. Furthermore, analysis of velocity and path of the centre of gravity in different wall sections is done for all available speed climbing athletes. The observed data set consists of 297 runs in total divided into two subsets of 202 observations of 47 male and 95 of 25 female athletes. Among others, the statistical model was used for the validation of the measured parameters and the review and impact of proven techniques like the Tomoa skip in the start section. Likewise interesting is the high influence of the parameters, measured especially in the middle section of the wall, on the end time.