Using Data Analytics to Make the Scouting and Training of Sports Talents More Effective

Abstract

This research proposes methods to get insights from limited data from sports talents. Since the data is limited, the focus is on comparing talents. Data from Dutch handball talents is used as a case to create and test the methods. The research covers three main phases; data preparation, data analysis and data visualization. Substantiated with theory from the literature, we outline procedures for every phase. For the first phase we propose a way to estimate missing values by using multiple linear regression in combination with clustering. In the second phase, we propose a nearest neighbors regression approach to find the best distance range to compare talents. In the last phase, we show a way of visualizing comparable talents using the spider plot, to present insights to sports scouts and coaches. Based on the results from the methods we tested with the handball data, we conclude that the approach from phase 1 works sufficiently in the case we tested, but that it does not necessarily work for other cases or other variables due to limited data, which is a limitation. Furthermore, we consider both approaches in phases 2 and 3 applicable in the handball case. However, to improve the reliability, the methods could be tested more extensively with other data in a future research.