The general objective of this doctoral thesis was to develop tools to improve the on-ice assessment of young elite ice hockey players. The first part was, through a longitudinal follow-up, to monitor the evolution of the morphological, physiological and skating skills profile that characterize youth players involved in elite ice hockey development leagues. Eighteen (18) players were evaluated off-ice and on-ice at the beginning, end, and at the beginning of the following season. We noticed that the maximal aerobic power was maintained during the hockey season despite little time invested to develop this physical quality. This phenomenon is explained by the fact that the nature of the game as well as the intensity of the on-ice training sessions were enough to maintain a respectable V̇O2max value throughout the season. All skating skills improved during the hockey season, but not during the off-season when some degradation was observed. Specifically, it was observed that the low percentage of common variance (<20%) between physiological variables measured off-ice and skating performance tests assessed on-ice, indicated significant shortcomings in both the choice of conventionally off-ice training methods as well as off-ice assessment tools. For these different reasons, the objective of the second project was to update the V̇O2 values obtained during the Skating Maximal Aerobic Test (SMAT) and to propose a Skating Stride Index (SSI) in order to improve the prediction of oxygen uptake. Indeed, the SMAT, like most field tests, assumes that participants who reach a given level have the same oxygen consumption, which is not usually the case. An SSI was developed, and when associated with the maximum speed achieved during the test, the accuracy of the predicted V̇O2 values was improved (r=0.95, SEE=1.92). When comparing players' SSI, it was noted that the V̇O2 values could vary by more than 5ml∙kg-1∙min-1 within the same stage, suggesting the importance of including the SSI in the prediction of V̇O2max. Subsequently, in a third project, we evaluated the robustness of the O2 cost prediction equations in four different anaerobic skating situations. As such, 24 elite players participated in this study for whom the execution time, heart rate, V̇O2, skating strides and SSI were measured for each test. The four regression equations displayed correlation coefficients ranging from 0.91 to 0.93 and a SEE between 4.5 to 8.4%, suggesting that the execution time alone is a poor predictor of O2 uptake required for this type of effort. The introduction of the SSI will allow coaches to obtain more precise information concerning the development of skating skills of their players, since it has notably been observed that young players generally have a lower skating efficiency than older players do. Finally, the goal of our last project was to examine the predictive value of our algorithms to estimate oxygen cost and lactate level using an on-ice anaerobic lactic capacity test. Thus, 20 elite players were tested, and post-exercise lactate accumulation was measured. The inclusion of the SSI had a major impact in the lactate level prediction by improving the correlation of the multiple regression model from 0.54 to 0.87, while reducing the standard error of the estimate from 10.4% to 6.5%. This finding is consistent with the results presented by Saltin et al. (1972) who reported the importance of considering an index of mechanical efficiency in this type of test. Our results have led to the development of accurate prediction tools to estimate absolute V̇O2 (r=0.87, SEE=0.19) and lactate level values (r=0.87, SEE=0.94). From a practical point of view, the new algorithms developed will help coaches to determine the maximum lactate concentration without requiring the use of blood samples. Thus, our work will have enabled the development of on-ice assessment tools available to both scientists and coaches.