RELIABILITY AND VALIDITY OF FULLTRACK AI APP TO MEASURE CRICKET BALL SPEED UNDER TRAINING CONDITIONS

Abstract

INTRODUCTION Radar guns are commonly used to accurately and reliably measure ball speed(1), a key cricket bowling performance indicator. App-based approaches, such as Fulltrack AI, are gaining popularity. This study investigated the reliability and validity of Fulltrack AI to measure cricket ball speed compared to a validated radar gun(1). METHODS Ball speed of 1081 deliveries (pace=783; spin=298) from a range of training sessions and conditions (batter, no batter; indoor and outdoor wickets) were recorded simultaneously using a radar gun (Stalker ATS2) and iPad running Fulltrack AI (version 1.13.1). Fulltrack AI data (ball speed (km/hr), line, length (m)) were extracted post-session for tabulation with radar gun data. Statistical analyses were conducted in R Statistical Software independently for bowling type (pace, spin) following exclusion of outliers. Reliability was assessed with standard error of measurement (SEM), coefficient of variation (CV) and intraclass correlation coefficient (ICC). Agreement was assessed using Bland Altman’s, 95% limits of agreement (LOA)(2). Validity was assessed using generalised additive models (GAM), controlling for line, length and interaction of training conditions. RESULTS Whilst reliability coefficients for pace deliveries demonstrated very good agreement (ICC=0.90; SEM=2.61) and lower variability (CV=2.56%) in contrast to spin (ICC=0.76; SEM=2.17; CV=3.08%); LOA demonstrated poor to fair levels of agreement, exceeding maximal allowable differences (>3%). When controlling for line, length and training conditions, GAMs ‘average model’ identified Fulltrack AI significantly (p<0.05) overestimated ball speed (pace: estimate 2.58km/hr, SE=1.24; spin: estimate 3.93km/hr, SE=0.81) when compared to the radar gun. CONCLUSION Fulltrack AI is a reliable method for monitoring ball speed where accuracy is not of paramount importance. Significant overestimation of ball speed in contrast with a radar gun, even after controlling for different training conditions, suggests software refinement is required before such technology is readily adopted for the measurement of speed. 1.Smith & Burke (2021) 2.Bland & Altman (1986)