Trial-count estimates for table tennis displacement waveforms vary with action, representation and analytical assumptions

Abstract Repeated-trial mean waveforms often characterise individuals, but the required trial count is not fixed. We reanalysed the public TTMD6 archive (30 archive codes, six fixed English action labels and 9,000 trials), constructed racket and aggregated-body displacement waveforms, and used 5,000 archive-code cluster bootstrap replicates to estimate the number of trials required to reach a relative-reliability reference level of 0.90. For event-unregistered waveforms under negligible residual dependence, minimum integer estimates from the L² trace ratio, which integrates variance across the full waveform, were 6–27 trials for the racket and 4–8 for the aggregated body. For backhand drive (BD), the racket estimate was 27 trials, compared with 17 from the phase-averaged pointwise reliability. The continuous BD threshold was 26.32 (95% bootstrap percentile interval, 17.65–43.48). Estimates were sensitive to assumptions: a Hampel-type local high-value rule reduced the BD racket estimate to 18; a positive trial-dependence scenario (AR(1), φ = 0.30) gave 49 with fixed variance components; and single proxy-peak registration yielded 3–31 trials across conditions. These estimates apply to the specified TTMD6 actions, representations and within-session assumptions and should not be interpreted as a general acquisition prescription.