Agility plays a crucial role in ice hockey training, and it can be developed directly on the ice or by additional off-ice training. Since the effectiveness of on-ice and off-ice training on players’ agility have not been previously described, the purpose of this research is to compare the effects of on-ice and off-ice agility training on skating performance. Fourteen ice hockey players performed agility training on-ice for 4 weeks and off-ice for 4 weeks in a crossover design; they were tested before the agility program, after the first month and after finishing both training programs. The players were randomly assigned into one of two groups (n = 7 in each group), either performing the on-ice training protocol first (Ice1) followed by the off-ice agility training or performing the off-ice protocol first and the on-ice training second (Ice2). The test battery included straight sprints to 6.1 m and 35 m and the S corner test, test with break, weave agility with puck test and reactive agility test. The magnitude based decision showed the effect of agility training in both groups in the weave agility (Ice1, 2.9±2.8% likely improvement; Ice2, 3.1±2.5% possible improvement) and reactive agility tests (Ice1, 3.1 ±2.5% likely improvement; Ice2, 1.7±2.1% possible improvement), where the Ice1 protocol resulted in a likely positive change and Ice2 resulted in a possible positive change. The comparison of the training effect resulted in a possibly harmful change of performance in Ice2 protocol (-0.5 ± 8.9%) compared to Ice1 protocol (-1.0 ± 5.1%). On-ice training is more effective in the development of specific types of agility in adolescent U16 players. However, there is evidence that off-ice agility have motor transfer to on-ice agility. Therefore, we recommend developing on-ice agility with additional off-ice agility training during the ice hockey season.