Influence of vertical GPS sensor placement on criterion agreement and interchangeability of distance and speed measures.

Abstract

This study evaluated agreement between PlayerTek™ 10 Hz GPS-derived total-distance and peak-speed measurements and practical field-based criterion measures and examined whether vertical sensor placement influenced agreement characteristics and measurement interchangeability. Following quality-control procedures, 758 paired observations were retained for analysis. Total-distance analyses included 533 paired observations from 55 participants collected during a modified team-sport simulation circuit, while peak-speed analyses included 225 paired observations from 68 participants during linear sprint testing. GPS units were positioned simultaneously at upper-back and mid-back locations. Agreement and measurement variability were assessed using established agreement-based statistical procedures, with secondary analyses evaluating placement equivalence and interchangeability. The upper-back placement demonstrated superior agreement with criterion measures compared with the mid-back placement. For total distance, upper-back measurements exhibited a small bias (0.57 m) and low variability (CV = 1.54%), whereas the mid-back placement demonstrated substantial systematic underestimation (-12.69 m), greater variability (CV = 4.18%), and poor agreement between placements. Equivalence testing confirmed that upper-back and mid-back placements were not interchangeable for total distance. For peak speed, upper-back measurements demonstrated stronger agreement with the criterion measure (ICC(A,1) = 0.864; CCC = 0.864) than the mid-back placement (ICC(A,1) = 0.770; CCC = 0.769), with smaller bias and lower variability. GPS-derived total-distance and peak-speed measurements demonstrated superior agreement characteristics when sensors were positioned at the manufacturer-recommended upper-back location. A small 10 cm inferior displacement of the sensor substantially altered total-distance measurements and reduced agreement with criterion measures, whereas peak-speed measurements were less affected. These findings highlight the importance of standardised sensor placement within applied athlete-monitoring environments.