The aim of this study was to update the metabolic power (MP) algorithm ( , W·kg −1) related to the kinematics data (P GPS, W·kg −1) in a soccer-specific performance model. For this aim, seventeen professional (Serie A) male soccer players ( 55.7 ± 3.4 mL·min −1·kg −1) performed a 6 min run at 10.29 km·h −1 to determine linear-running energy cost (C r). On a separate day, thirteen also performed an 8 min soccer-specific intermittent exercise protocol. For both procedures, a portable Cosmed K4b 2 gas-analyzer and GPS (10 Hz) was used to assess the energy cost above resting (C). From this aim, the MP was estimated through a newly derived C equation (P GPSn) and compared with both the commonly used (P GPSo) equation and direct measurement ( ). Both P GPSn and P GPSo correlated with (r = 0.66, p < 0.05). Estimates of fixed bias were negligible (P GPSn = −0.80 W·kg −1 and P GPSo = −1.59 W·kg −1), and the bounds of the 95% CIs show that they were not statistically significant from 0. Proportional bias estimates were negligible (absolute differences from one being 0.03 W·kg −1 for P GPSn and 0.01 W·kg −1 for P GPSo) and not statistically significant as both 95% CIs span 1. All variables were distributed around the line of unity and resulted in an under- or overestimation of P GPSn, while P GPSo routinely underestimated MP across ranges. Repeated-measures ANOVA showed differences over MP conditions ( F 1,38 = 16.929 and p < 0.001). Following Bonferroni post hoc test significant differences regarding the MP between P GPSo and /P GPSn ( p < 0.001) were established, while no differences were found between and P GPSn ( p = 0.853). The new approach showed it can help the coaches and the soccer trainers to better monitor external training load during the training seasons.