Validity and Reliability of a 15 Hz GPS Device for Court-Based Sports Movements

Tessaro E., Williams J.H.

Aim: The aim of this study was to determine the criterion validity and reliability of a 15Hz GPS unit during court-based movements that simulate activities encountered during a tennis match.


Twenty GPS units were used (SPI HPU, GPSports, 15 Hz GPS, 100Hz/16g accelerometer, 50Hz magnetometer). A calibrated trundle wheel was used as the criterion device (Meter-Man MK45M, Komelon, 363cm diameter, 1.141m circumference, 0.1m resolution). The wheel was calibrated against a 100m steel tape before and after each day, with no changes observed during the day’s session or between days. Two HPU SPI units (paired units) were attached to the wheel, 8 and 15 cm directly above the axis of the wheel and 2.5 cm lateral to the pivot point of the wheel, enabling comparisons between individual GPS units and the wheel (criterion validity) as well as between paired units (inter-unit reliability). Four trials were conducted to test the validity and reliability of the units. Standard course paths were followed as closely as possible, and the start and stop times of each trial were recorded using a GPS-linked digital watch.

The distance trial (DIST) consisted of four laps of a regulation 400m running track. The GPS units and wheel traveled along an imaginary line 0.3m from the inside of lane one, with the wheel and units passing through 100m intervals with alternating speeds similar to a jog, run, and sprint.

The three court-based trials were performed on a regulation tennis court. The shuttle run trial (SHUT) involved the wheel following a path along the singles baseline (8.2m), pivoting 180°, and then returning to the starting point. Five sets of three laps were performed with a 20 sec pause between sets (246 m total distance). During the change of direction trial (COD), the wheel followed a course consisting of five segments and four turns. The random movement trial (RAND) consisted of five, 10 sec sets of random movements within the singles court.

Cohen’s d was used to calculate effect sizes between unit and wheel distances, and Bland-Altman plots were used to determine the agreement between the GPS- and wheel-measured distances. Root mean square (RMS) errors and intraclass correlation coefficients (ICC) were also computed to evaluate agreement between devices.


Typical GPS position recordings for the four trials showed noticeable differences between the standard paths and the GPS determinations (±0.5m), as high-speed movements and rapid directional changes made it difficult for the wheel to precisely track the specific path. The wheel and GPS units over-estimated the standard distances by 1.0-2.5%. However, Cohen’s d classified differences between the wheel and units as trivial and small. Mean RMS error values between the two devices were < 2% for three of the four trials. Intraclass coefficient (ICC) values were quite high (>0.95). Bland-Altman plots of the total distances for the four trials showed mean bias values were negative, indicating an overestimation of wheel distance by the GPS units, but bias values were small, less than 2.5m and less than 2%. The limits of agreement (LOA) were within an acceptable range for all trials. For all trials combined, 96.25% of the difference measurements fell within the LOA. Inter-unit reliability for paired units was also good, with mean RMS error and bias values within acceptable ranges.

View this research