Biomechanical Variables Associated with the First 20 m of the Start Phase in Speed Skaters

This exploratory study examined segmental biomechanical variables associated with 20 m start performance in 24 juvenile inline speed skaters (14–16 years). Athletes performed three maximal 20 m starts; the best time was recorded with photoelectric timing gates, and the corresponding trial was analyzed from two-dimensional video using Tracker. Segmental kinematics and indirect inertial estimates were derived using De Leva-adjusted anthropometric parameters. Candidate predictors were selected with LASSO, and the complete preprocessing and selection pipeline was evaluated using nested leave-one-out cross-validation. Post-selection ordinary least squares, bootstrap, and Bayesian models were used as secondary analyses of coefficient sensitivity and uncertainty. LASSO retained forearm segmental inertial force estimate, thigh velocity, forearm velocity, thigh angular velocity, and torso angular acceleration. The nested pipeline achieved RMSE = 0.12 s, MAE = 0.08 s, and R2_pred = 0.485. Greater thigh velocity, forearm inertial force estimate, thigh angular velocity, and torso angular acceleration were associated with shorter times, whereas greater forearm velocity was associated with longer times. These findings identify candidate within-sample associations, not stable biomechanical determinants. Because of the small sample size, cross-sectional design, indirect two-dimensional measurements, absent reliability analysis, and lack of external validation, the variables should be used only for hypothesis generation and provisional monitoring.