The aim of this study was to develop an upper limb muscle power (PUL) prediction model using the Medicine Ball Throw Test (MBT) in rugby players. Eighteen amateur rugby players underwent the MBT test and the guided bench press exercise at 30, 40, 50 and 60% of 1. Myotest^® accelerometer was positioned on the bench press bar to estimate muscle power. Linear regression was used to derive the upper limb muscle power prediction equation from the MBT distance. The residue analysis estimated the residual error of the predicted values using values obtained by Myotest^®. Bland-Altman plots were used to verify agreement between actual and predicted upper limb muscle power, both in absolute Watts (W) and relative terms (W/kg of fat-free mass). There were significant correlations between actual and predicted upper limb muscle power (r = 0.834, 0.854, and 0.872) for intensities of 30%, 40% and 50%, respectively. Absolute bias of predicted values was -1.87 W (p <0.05). For muscle power predicted relative to fat-free mass, bias was 0.782 W/kg (p <0.05). Conclusion: The MBT test has high correlation with actual PUL values and it was found that the equation developed in this study has high accuracy to predict PUL in rugby players of both sexes.