Journal Article

GENDER-SPECIFIC PREDICTORS OF VAULT PERFORMANCE IN GYMNASTICS: A MACHINE LEARNING APPROACH

Marinšek, Miha, Veličković, Saša, Paunović, Miloš, Vodičar, Janez, Đorđević, Dušan, Kreft, Robi, Kolar, Edvard

2025

Overview

This study investigated gender-specific predictors of vault performance in gymnastics by applying machine learning techniques to analyse body composition and run-up dynamics. Data were collected from 27 national-level gymnasts (17 female, 10 male) during the Slovenian Cup competition. The focus on gender-specific predictors stems from fundamental physiological and biomechanical differences between male and female athletes, which influence force production, movement kinematics, and execution mechanics. A deeper understanding of these distinctions enhances the precision of performance modelling and supports the development of targeted, evidence-based training interventions. Spatiotemporal parameters of the run-up were recorded using the OptoGait system, while body composition was assessed with the Tanita DC-360. Principal Component Analysis (PCA) and Boosting regression models were used to identify key predictors of vault execution scores. These methods were selected for their ability to reduce dimensionality and capture complex, nonlinear relationships in performance data. The results revealed clear gender-specific patterns. For female gymnasts, the model explained 74.4% of the variance in execution scores, with Overall Lean Body Mass emerging as the most influential predictor (47.12% relative influence), followed by Overall Contact Phases (25.28%). For male gymnasts, the model demonstrated exceptionally high predictive power, explaining 97.8% of the variance, with Body Fat as the primary predictor (48.44% relative influence), followed by Flight and Contact Dynamics (35.22%). These findings suggest that training strategies should be tailored to gender-specific needs. For women, emphasis on lean muscle development, stride optimisation, and the coordination of rhythm and timing may be beneficial. For men, managing body fat levels, optimising flight and contact dynamics, and adopting an integrated approach to stride mechanics appear essential. Given the potential for misinterpretation of body composition metrics, a holistic approach to athletic conditioning is recommended. However, the study’s limitations, including the small sample size and cross-sectional design-warrant cautious interpretation. This research provides a foundation for future investigations into gender-specific factors affecting vault performance. Larger and longitudinal studies are needed to validate these findings and support the development of more precise training interventions. Ta raziskava je raziskovala napovedne dejavnike uspešnosti preskoka v telovadbi ločeno po spolu z uporabo tehnik strojnega učenja za razčlenitev telesne sestave in dinamike zaleta. Podatki so bili zbrani pri 27 osebah (17 žensk, 10 moških) med tekmovanjem za pokal Slovenije. Poudarek na napovednih dejavnikih uspeha po spolu izhaja iz temeljnih fizioloških in biomehaničnih razlik med moškimi in ženskami, ki vplivajo na proizvodnjo sile, kinematiko gibanja in mehaniko izvedbe. Globlje razumevanje teh razlik povečuje natančnost modeliranja uspešnosti in podpira razvoj ciljno usmerjenih, na dokazih temelječih vadbenih posegov. Prostorsko-časovni parametri zaleta so bili zabeleženi s sistemom OptoGait, medtem ko je bila telesna sestava ocenjena z napravo Tanita DC-360. Za prepoznavo ključnih napovednih dejavnikov rezultatov izvedbe preskoka sta bila uporabljena razčlenitev glavnih sestavin (PCA) in model Boostingove regresije. Ti metodi sta  bili izbrani zaradi njihove sposobnosti zmanjšanja razsežnosti in zajemanja zapletenih, nelinearnih odnosov v podatkih o uspešnosti. Rezultati so razkrili jasne vzorce, ki se točno ločijo op spolu. Pri telovadkah je model pojasnil 74,4 % spremenljivost v rezultatih izvedbe, pri čemer se je kot najpomembnejši napovedovalec izkazala celotna pusta telesna masa (47,12 % sorazmernega  vpliva), sledililo je  trajanje opore (25,28 %). Pri telovadcih je model pokazal izjemno visoko napovedno moč, saj je pojasnil 97,8 % variance, pri čemer je bila telesna maščoba najpomembnejši napovedovalec (48,44 % sorazmernega vpliva), sledila pa sta dinamika leta in opore (35,22 %). Te ugotovitve kažejo, da bi morale biti strategije vadbe prilagojene potrebam posameznega spola. Za ženske je lahko koristen poudarek na razvoju puste mišične mase, optimizaciji korakov ter skladnosti ritma in časa. Za moške se zdi bistveno obvladovanje ravni telesne maščobe, izboljšanje dinamike leta in opre ter sprejetje celostnega pristopa k mehaniki korakov. Glede na možnost napačne razlage meritev telesne sestave je priporočljiv celosten pristop k stanju telesne priprave. Vendar pa omejitve raziskave, vključno z majhno velikostjo vzorca in presečno zasnovo, zahtevajo previdno razlago. Ta raziskava zagotavlja osnovo za prihodnje raziskave dejavnikov, ločeno po spolu, ki vplivajo na uspešnost preskoka. Za potrditev teh ugotovitev in podporo razvoju natančnejših vadbenih ukrepov so potrebne obsežnejše in večletno spremljanje.

• Publisher: University of Ljubljana, Faculty of Sport,University of Ljubljana Press (Založba Univerze v Ljubljani)
• Subject: Biomechanics / Body composition / Body fat / execution score / Females / Gender / Gymnastics / Influence / Integrated approach / Investigations / Machine learning / Measurement techniques / principal component analysis / Principal components analysis / run-up characteristics / Velocity