Accuracy and precision of kinematic methods for determining ball release in baseball pitching

Identifying the instant of ball release is critical for analyses of pitching biomechanics. This study evaluated accuracy and precision of three methods for estimating release-Wrist Pass (WP), Metacarpophalangeal Pass (MP), and Metacarpophalangeal Peak Velocity (MV)-across fastballs, curveballs, sliders, changeups, and splitters. Marker motion data for 131 male pitchers captured at 1,000 Hz compared these kinematic methods against a reference release determined from the geometric separation of the ball from the third fingertip. Accuracy was quantified via mean error (systematic bias) and precision via mean absolute deviation (MeanAD). Pitch type significantly influenced detection errors (p < 0.05). The MP method demonstrated highest mean accuracy, with least error relative to the time of reference release. However, WP method exhibited superior precision. This discrepancy was most pronounced in curveballs, where WP (1.1 ms) outperformed MV (1.6 ms) and MP (2.8 ms). Temporal precision must be prioritised, as systematic bias is mathematically correctable while between-pitcher differences in bias are not. At shoulder internal rotation velocities exceeding 5,000 deg/s, a timing discrepancy of 2 ms results in a 10 deg error in shoulder rotation. Therefore, the WP method, adjusted with pitch-specific offsets, provides the most robust approach for research and clinical pitching analysis.