Data-Driven External-Load Analytics: Integrating Cluster Analysis and ACWR Monitoring in Elite Handball

The aim of this study was to characterize multidimensional external-load profiles obtained from sensor-based tracking data over a four-month competitive period in an elite men's handball team and to investigate their associations with the session type, the playing position, and weekly workload fluctuations, as measured by the acute:chronic workload ratio (ACWR). Data were collected from 23 elite players using an ultra-wideband tracking system. Six variables, i.e., total distance, maximum speed, accumulated acceleration load (AAL), the number of exertions, maximum jump height, and the number of jumps ≥0.30 m, were standardized and clustered using k-means (k = 4). The cluster with the highest composite z-score was defined as a high load. The association between weekly density of high-load sessions and the likelihood of an ACWR spike (≥1.30) was tested using logistic regression. Results showed that match sessions were 1.8 times more likely than training sessions to fall into the high-load cluster. Wings and center-backs were significantly more represented in high-load clusters than goalkeepers and pivots. Additionally, when 15% or more of the previous week's sessions were classified as of the high load, the odds of an ACWR spike in the following week increased by 10.7 times. These findings suggest that data-driven (unsupervised) clustering of external-load variables supports early identification of high-risk workload patterns. Monitoring the weekly distribution of high-load sessions may help mitigate fatigue-related maladaptation by enabling proactive, position-specific load management in elite handball.