Including a Technical Factor with Physical and In-Game Movement Factors Improves Model Sensitivity When Evaluating...

Determining characteristics that define talent is critical for recruitment and player development. When developing predictive models, sensitivity is important, as it describes the ability of models to identify players with draft potential (true positives). In the current literature, modelling is limited to a small number of selected variables, and model sensitivity is often poor or unreported. The aim of this study was to determine how a technical factor combined with physical and in-game movement factors affects position-specific model sensitivity when evaluating draft outcome in an elite-junior National Australia Bank (NAB) League population. Physical, in-game movement, and technical involvement data were collated from draft-eligible (18th year) participants in the under 18 boys NAB League competition (n = 465). Factors identified through parallel analysis were used in binomial regression analyses. Models using factor combinations were developed to predict draft success for all-position, nomadic, fixed-position, and fixed&ruck players. Models that best characterised draft success were all-position (physical and technical: specificity = 97.2%, sensitivity = 36.6%, and accuracy = 86.3%), nomadic (physical and technical: specificity = 95.5%, sensitivity = 40.7%, and accuracy = 85.5%), fixed (physical: specificity = 96.4%, sensitivity = 41.7%, and accuracy = 86.6%), and fixed&ruck (physical and in-game movement: specificity = 96.3%, sensitivity = 41.2%, and accuracy = 86.7%). Including a technical factor improved sensitivity in the all-position and nomadic models. Physical factors and physical and in-game movement yielded the best models for fixed-position and fixed&ruck players, respectively. Models with improved sensitivity should be sought to assist practitioners to more confidently identify the players with draft potential.