This study aimed to compare the force-velocity mechanical variables derived from high-speed video- and radar-based method during forward skating sprint in ice hockey. Thirteen elite female ice hockey players performed two 40-m forward skating sprints to determine, in the horizontal plane, maximal velocity reached (Vmax), relative maximal theoretical force (F0), maximal theoretical velocity (V0), relative maximal power (Pmax), linear slope of the force-velocity relationship (FV slope), maximal value of the ratio of force (RFmax) and index of force application technique (Drf). Two different high-speed video-based methods adding a time shift (ST-TS) or not (ST) were used and independently compared to the radar-derived method. ST and ST-TS showed significant mean differences (all p < 0.002) compared to radar-derived processing for all variables except for V0 (p = 0.26) and Vmax (p = 0.13) inferred from ST. In reference to radar-derived variables, ST-TS significantly induced larger lower values compared to radar of the main forward skating sprint determinants (Pmax, F0, RFmax and Drf) and moderate-to-large overestimation for velocity variables (V0 and Vmax). Correlations between ST or ST-TS and radar-derived methods ranged from trivial for velocity variables to very large for force and power variables. Consequently, practitioners must be aware that using such high-speed video-based methods would permit to determine mechanical variables at the cost of much lower accuracy and reliability than the radar-derived method.