The use of normalized performance ratings to valuate players: The case of per-minute statistics in basketball

Abstract

This research empirically tests the hypothesis underlying the use of performance indicators in basketball, when they are normalized using the “per-minute” criterion: There is no association among normalized indicators and the minutes played by each player. This assumption is a necessary requirement in order to compare players using normalized indicators, because if this assumption is not met, then there would be a possible bias in the rankings of each statistical category. We apply several under-used methods in this type of research, such as the analysis of the intra-player variability and the structural equation modelling. Results show that there is no lineal association between the minutes played and the different per-minute normalized categories as: points, rebounds, assists, steals, and blocks. We discuss implications for sport performance and marketing. Therefore, this study contributes to the advance in one of the growing areas of sports research; such as basketball metrics.
Key Words: normalization, basketball, performance, statistics