Data-Driven Validation of the Structural Validity World Shooting Para Sport Shotgun Classification: A K-Means Cluster Analysis of 176 Para Trap Athletes.

Abstract

Evidence-based Paralympic classification must limit impairment-related advantages. World Shooting Para Sport classification, based on manual muscle testing, has never been empirically validated. We assessed structural validity using unsupervised machine learning on retrospective data from 176 Para Trap athletes spanning 2017-2025. Z-standardized composites captured upper limb, lower limb and trunk strength, plus trunk stability. K-means clustering, with k selected via elbow and silhouette methods, was compared with official classes using accuracy and information-based metrics. The best four-cluster solution achieved accuracy above 80% and moderate agreement, matching SG-U (Shotgun Upper) and SG-L (Shotgun Lower), but dividing SG-S (Shotgun Sitting) into two distinct groups exposing within-class heterogeneity. Upper limb strength was the main discriminator, followed by trunk stability and lower limb strength. Overall, data-driven clusters largely support existing structure while revealing overlap between SG-L and SG-U and heterogeneity within SG-S, suggesting that multivariate measures contribute to evidence supporting refinement of eligibility thresholds and enhance equity, aligning revisions with The International Paralympic Committee standards.