Applying principal component analysis to identify underlying factors of key performance indicators from microsensor data in basketball games

Abstract

The present study applied Principal Component Analysis (PCA) to complex microsensor data from 28 elite professional male basketball players across 122 games (1205 player/game samples) during two basketball seasons to reduce data dimensionality and identify underlying factors related to player load and movement intensity. External load was quantified using T7 Catapult monitoring devices, capturing 18 inertial measurement unit (IMU)-derived variables. PCA with varimax orthogonal rotation successfully extracted four distinct principal components (PC) with eigenvalues >1, collectively explaining 77.95% of the total variance. These components were interpreted as: PC 1 (33.43% variance), dominated by high-intensity acceleration and change of direction and representing high-intensity linear and multi-directional movements; PC 2 (19.27% variance), defined by deceleration measures, reflecting eccentric braking performance; PC 3 (15.49% variance) capturing overall accumulated mechanical workload; and PC 4 (9.76% variance) representing jumping frequency. In conclusion, PCA effectively reduced 18 highly correlated microsensor variables into four distinct, uncorrelated factors that encapsulate the main physical demands of elite basketball game-play, providing a more pragmatic and holistic framework for performance interpretation and subsequent tailoring of training prescription and monitoring over time. The extraction of these four factors provides a diagnostic framework for load profiling in elite basketball; however, as these results originate from a single-team observational design, future research is required to confirm the cross-contextual stability of this mechanical structure.

The authors' abstract, as published at the source. International Journal of Sports Science & Coaching, 2026 · DOI ↗