Purpose This study aimed to assess the efficacy of critical metabolic power derived from variable-speed movement for classifying intensity in team sport activity. Methods Elite male hockey players (n = 12) completed a series of time trials (100 yards, 400 yards, 1500 yards) and a 3-min all-out test to derive both critical speed (CS) and critical power (CP). Heart rate (HR), blood lactate, and rating of perceived exertion were measured during each protocol. Participants (n = 10) then played two competitive hockey matches. Time spent greater than 85% of maximum HR was compared with time spent above CS (from the time trials) and CP (from the 3-min test). Results Between protocols, there was a moderate and nonsignificant association for CS (r = 0.359, P = 0.252) and a very large association for CP (r = 0.754, P = 0.005); the association was very large for peak HR (r = 0.866, P < 0.001), large for blood lactate (r = 0.506, P = 0.093), and moderate for rating of perceived exertion (rho = 0.441, P = 0.152). Time trials produced higher CS (4.3 vs 2.0 m·s-1, P < 0.001) and CP (18.3 vs 10.5 W·kg-1, P < 0.001) values than did the 3-min test. In matches, there was a very large association between time spent above 85% of maximum HR and time spent above both CS (r = 0.719, P < 0.001) and CP (r = 0.867, P < 0.001). This relationship was stronger for CP compared with CS (Z = 3.29, P = 0.0007). Conclusions Speed is not an appropriate parameter for the classification of team sport activity comprising continual changes in speed and direction; however, critical metabolic power derived from variable-speed activity seems useful for this purpose.