A data-driven approach to evaluating the use of the periodisation of skill training framework for classifying training in a team-sport context

This study aimed to evaluate the application of the PoST framework in a team sport context and compare two data-driven classification methods to objectively assess the skill training outcomes. Training drills were assessed for information complexity and game-representativeness using variables describing the training environment and athlete performance. Coach classifications of drills were analysed using repeated-measures MANOVAs, while k-means and Gaussian Mixture Model (GMM) clustering objectively grouped drills based on information complexity, and compared with PoST stages. Coach classifications varied, with training intentions not always clearly distinguished, and similar game-representativeness observed across classifications, despite PoST emphasising this as a key differentiator. Both clustering methods identified three distinct training types, aligning with PoST's 'Complex Training', 'Team-based Training' and 'Performance Training' stages. However, the probabilistic overlap of the GMM better captured drills transitioning between stages. Neither method identified drills reflecting PoST's Coordination stage, suggesting misalignment between coach intent and the framework's definition. Coaches should consider both training complexity and representativeness to better align drills with game demands and desired outcomes. A modified framework designed for team-level skill development may support this, while objective classification methods may be used to enhance training design and evaluation.