The aims of this study were to describe differences in the acceleration-speed (A-S) profile in-situ and to assess the week-to-week reliability of the A-S profile in-situ over a given training cycle of elite youth soccer players, in relation to the number of sessions included and analyse the effect of the inclusion or not of a specific sprint session. In this retrospective study, 18 male elite U19 football players (179.4 ± 7.1 cm;​ 69.0 ± 9.5 kg) participated. GPS data collected from three consecutive typical training weeks were used to calculate different combinations of A-S profile in-situ variables (theoretical maximal acceleration [A₀], theoretical maximal speed [S₀] and the slope of the acceleration-speed [ASslope]). The number (and content) of sessions affected mainly S₀ while A₀ remained similar with or without a sprint session. The reliability of the A-S profile in-situ is more related to the spread of points rather than a specific number of sessions (and thus points) and was improved when a high percentage of maximum speed (i.e. ≥ 95%) was reached. The present study showed low week-to-week variability for A₀, S₀ and AS_slope. However, practitioners need to make sure that the values cover a sufficient range of raw data [20–95% of maximum speed] to build a clear and consistent linear regression, and in turn extrapolate meaningful A-S profile values.