Author(s): LIM, C., TENG, P.S.P., KOMAR, J., Institution: NATIONAL INSTITUTE OF EDUCATION, Country: SINGAPORE, Abstract-ID: 1583

INTRODUCTION:
Movement complexity in climbing is often quantified using global trajectory entropy, but global measures cannot distinguish whether learning reduces complexity uniformly or redistributes it into specific transitions. Moving-window entropy profiling was utilised along with Dynamic Time Warping (DTW) to test whether practice reduces local complexity and whether whole-attempt movement patterns converge despite differences in execution time.
METHODS:
An experienced male climber (height: 1.80m, wingspan 1.90m) completed five attempts on two bouldering problems (easy and hard). Hip trajectories were analysed in frontal (XY), transverse (XZ), and sagittal (YZ) planes, where the Z axis corresponded to the wall-normal (depth) direction. In each plane, local entropy was calculated in sliding windows to illustrate how movement complexity changed over time within each attempt. To compare attempts while allowing for different speeds, DTW was applied to (i) 2D hip trajectories in each plane, and (ii) the time-series of windowed entropy in each plane. DTW produced pairwise similarity scores and an alignment path, with further analysis on the shift in distribution of complexity and concentration of complexity within the trajectory to draw conclusions.
RESULTS:
In the easy route, entropy DTW distances were generally smaller and showed earlier consistency, especially in YZ (Attempt pair 2–5 = 0.025, 63.7% decrease from Attempt pair 1-5). Entropy patterns for the hard route became more consistent across later attempts in the XY (Attempt 4-5 = 0.045, 59.1% decrease) and YZ (Attempt 4-5 = 0.042, 57.1% decrease) planes, but XZ remained less stable (Attempt 4-5 = 0.077, 2.29% decrease), suggesting ongoing re-organisation of transverse-plane control. Trajectory DTW distances showed a stark difference, with high initial distances for the easy route (Attempt 1-5 = 1.93) that dropped from the second attempt onwards (Attempt 2-5 = 0.21, 89.1% decrease), while the hard route saw closer consistency (Attempt 1-5 to Attempt 4-5 = 0.39 to 0.18, 53.9% decrease). For entropy concentration, only the XZ plane (hard route) decreased consistently across attempts (Gini: 0.35 to 0.26), indicating smoothing of movement complexity distribution.
CONCLUSION:
Local entropy and DTW describe different aspects of learning in climbing. Entropy profiles show how complexity is distributed over time within each attempt, while DTW shows how similar attempts are without speed differences and amount of time-warping needed. Across practice, the climber showed clearer stabilisation in some planes (XY and YZ), while the hard route showed continued changes in out-of-plane behaviour (XZ). Trajectory DTW on the easy route also highlighted how rapid reorganisations cause a step-drop in DTW distance. Overall, this combined approach can potentially identify which parts of a movement become stable with practice, offering a practical way to pinpoint how and where reorganisation occurs during complex whole-body skill learning.