The influence of sample and filter frequencies on baseball pitching data

The throwing motion in baseball pitching is the fastest recorded human motion, demanding higher sample frequencies than most human movements. Previous studies have used sample frequencies from 240 to 500 Hz and a wide range of low-pass filter frequencies, although 13.4 and 18.0 Hz are the most common. This study examined the effects of two common sample frequencies (240 and 480 Hz) and three filter frequencies (13.4, 18.0, 27.2 Hz) on pitching data. Fastballs from 28 professional baseball pitchers were collected at 480 Hz with marker motion capture, then down-sampled to 240 Hz; each data set was filtered at the three filter frequencies. Twenty-six common kinematic, temporal and kinetic variables were computed for each pitcher across all sample-filter frequency combinations and compared using two-way repeated measures ANOVA. Differences were interpreted relative to minimal clinically important differences (MCID). Pitching data were more sensitive to filter frequency (significant differences above MCID in 13 of 26 variables) than to sample frequency (4 of 26 variables). Variables involving differentiation (e.g. velocities and torques) were most impacted. Based on these findings and time-series inspection, we recommend a minimum sample frequency of 240 Hz and a Butterworth low-pass filter of 18.0 Hz for marker-based motion capture pitching mechanics.