Once-Weekly Game-Based (Crossing-And-Shooting) Versus Linear Sprint Training During the Competitive Season: A 12-Week Randomised Controlled Trial on the Effects on Sprint Performance, Match Peak Velocity and Eccentric Hamstring Strength in Amateur Soccer Players

Sprint ability is a key determinant of soccer performance, yet the relative merit of linear versus game-based sprint training during the competitive season remains unclear, and few studies verify whether training-induced gains are expressed in official competition. Thirty-six amateur male soccer players, six from each of six amateur clubs in the same regional league, were randomised with stratification by playing position in permuted blocks of three to a game-based sprint group (GBSG; a standardised crossing-and-shooting drill rather than a conventional small-sided game), a linear sprint group (LSG) or a control group (CG; n = 12 each) and trained once weekly across 12 competitive-season weeks. The trial was registered retrospectively: data collection ran from 1 November 2025 to 24 January 2026 and the registration record was first submitted on 28 August 2026, after all outcome data had been collected and accessed. No dated protocol or statistical analysis plan was established before recruitment, so no analysis reported here was prespecified. The principal outcome, designated at the design stage, was 40 m sprint time at Week 12; T-drill time, bilateral eccentric hamstring strength and match peak velocity were secondary outcomes. The baseline-adjusted (ANCOVA) estimator, the mixed models and all sensitivity analyses were introduced during revision, after the data had been seen, and every between-group contrast in this report is therefore exploratory. The trial was powered for the time × group interaction only (unadjusted pairwise differences detectable at d ≥ 1.20). Match peak velocity was monitored by 10 Hz GPS during 12 consecutive official matches; data were analysed with mixed-design ANOVA, ANCOVA and linear mixed models, referenced to the smallest worthwhile change (SWC). For the principal outcome, GBSG was 0.022 s faster than CG (95% CI −0.039 to −0.005, p = 0.014), whereas LSG did not differ from CG (−0.010 s, 95% CI −0.027 to 0.007, p = 0.256); critically, the direct GBSG−LSG contrast was −0.012 s (95% CI −0.030 to 0.005, p = 0.158), so the two training modalities did not differ from one another, and none of the three adjusted differences reached the SWC of 0.042 s. T-drill performance was unaltered, and the apparent hamstring advantage of GBSG disappeared after adjustment for baseline (adjusted difference 4.5 N, p = 0.139), reflecting baseline non-equivalence rather than training. For match peak velocity, a secondary outcome, the rate of change differed between groups: peak velocity increased 0.098 km·h−1·week−1 faster in GBSG than in CG (95% CI 0.035 to 0.161, p = 0.002), equivalent to 1.08 km·h−1 across the intervention and exceeding the SWC of 0.31 km·h−1, whereas LSG did not differ from CG (−0.034 km·h−1·week−1, p = 0.290); the direct GBSG−LSG contrast, on which any claim of superiority over linear sprinting must rest, was 0.132 km·h−1·week−1 (95% CI 0.069 to 0.194, p < 0.001), equivalent to 1.45 km·h−1 (95% CI 0.76 to 2.14) across the intervention. The statistical significance of this contrast was robust to model specification, to serial correlation and to adjustment for Week-1 value; its practical importance was not, because under the autoregressive and Week-1-adjusted specifications the lower confidence bound corresponds to only 0.22 and 0.26 km·h−1 across the intervention, that is, below the SWC. The trial was not powered for between-group comparisons and these findings are therefore secondary and hypothesis-generating. Attendance at the prescribed sprint sessions was 100% in the two intervention groups, to which sprint sessions alone were prescribed, and no adverse events were reported; prospective feasibility criteria were not defined, so no formal feasibility claim is made. A once-weekly sprint stimulus can be delivered in-season; adaptations in standardised testing were small and below the smallest worthwhile change, whereas the game-based stimulus alone was associated with a rise in peak velocity expressed during official competition that exceeded the smallest worthwhile change under the primary specification, though not under every sensitivity specification. All findings are exploratory (NCT07799324, registered retrospectively).