OBJECTIVES: Hamstring muscle injury prediction by isokinetic strength ratios is low but could result from the method-depending either on the use of the limbs or of the sportsmen as references. We aimed to establish a predictive model including unilateral and bilateral ratios calculated from the dominant, nondominant, right, and left limb in injured and uninjured professional soccer players. DESIGN: Cohort study. SETTING: Soccer team of the French Professional Premier League. PATIENTS: Ninety-one professional soccer players. INTERVENTIONS: Isokinetic muscle strength was prospectively measured at the beginning of 5 consecutive seasons (2009-2014). MAIN OUTCOME MEASURES: Several bilateral, conventional, and functional ratios were calculated from isokinetic measurements at different angular speeds (60 and 240 degrees/s in concentric mode and 30 degrees/s in eccentric mode). Thirty-one soccer players had a hamstring injury during the seasons and were compared with 60 uninjured players. Four models were tested to predict the occurrence of hamstring injury from isokinetic ratios calculated in accordance with the dominant, nondominant, right, and left limb. RESULTS: No predictive model was found when ratios were calculated from the dominant or the right limb. Two models of prediction were found when ratios were calculated from the nondominant or the left limb. In these 2 models, only the bilateral concentric hamstring-to-hamstring ratio at 60 degrees/s was predictive. The best prediction was found with the left limb. CONCLUSIONS: We identified 2 low predictive models for hamstring muscle injuries depending on the limbs studied. Because of a low prediction, the consensual method used to predict hamstring muscle injury must be defined in future studies.