Eighteen games players (9 males, 9 females) performed 30 maximum 6-sprints on a non-motorized treadmill. Each sprint was preceded by a 60-s jog at 40% maximum speed and was followed by a 54-s walk at 20% maximum speed. Thus, the entire test was of 60-min duration. The subjects were then randomly assigned to three groups and repeated the 1-h test 24 h later after consuming either a high, normal or low carbohydrate diet (79 +/- 3, 47 +/- 8, 12 +/- 1% CHO, respectively). During trial 1, mean power output declined from 653 +/- 131 to 600 +/- 158 W during the 30 sprints (P < 0.01) and power output was lower during trial 2 than during trial 1 (n = 18, P < 0.01). During trial 2, there were no differences in sprint performance between the dietary groups for the exercise test as a whole (trial 2 mean power lower than trial 1 by 0.2, 0.5 and 5.0% for the high, normal and low CHO groups, respectively; N.S.), but if only the first nine sprints are considered, then the high CHO group performed better than the low CHO group (P < 0.05). Blood lactate and glucose concentrations were lower during trial 2 than trial 1 by 4.5, 13.8 and 29.0% (lactate) and 14.9, 11.3 and 35.8% (glucose) for the high, normal and low CHO groups, respectively (both P < 0.01). Thus, both the metabolic responses to, and the performance of, maximum intermittent exercise were reduced when the test was repeated after 24 h recovery.(ABSTRACT TRUNCATED AT 250 WORDS)