Gender-based differences in physical fitness profiles of sepak takraw athletes: A multivariate and machine learning analysis

Abstract

Background: Gender differences in physical fitness are well documented in sports science; however, most studies rely on univariate statistical approaches and provide limited insight into multidimensional performance profiles. The application of multivariate and machine learning techniques in sport-specific contexts such as sepak takraw remains limited. Objectives: This study aims to analyse the differences in physical performance between male and female athletes. This study investigates the physical variables that differentiate male and female sepak takraw athletes using a quantitative ex post facto design. Methods: A total of 163 competitive sepak takraw athletes (81 male and 82 female). The sample consists of male and female athletes who underwent a series of physical measurements, including muscle strength, speed, endurance, and agility. Data analysis was conducted using an independent t-test to measure gender differences, Principal Component Analysis (PCA) to explore the  data structure, and a random forest classifier to identify the variables most contributing to gender classification. Results: Significant gender differences were observed in Jump DF, Back Dynamometer, Sprint 30 m, and 1600 m Run (p < 0.01; d = 1.34-2.48). PCA revealed two dominant components (strength–power and endurance), while Random Forest identified back strength as the most influential classification variable (importance = 0.224). The composite index provided a simplified representation of multidimensional performance profiles. Conclusion: The integration of multivariate and machine learning approaches provides a more comprehensive understanding of physical performance profiles in sepak takraw. The proposed composite index offers a practical tool for simplifying complex fitness data and supporting evidence-based training strategies.