The aim of this study was to determine the reliability of a non-motorised treadmill team-sport simulation for measuring physiological responses and performance demands of team sports. Following familiarisation, 11 team-sport athletes completed a peak sprinting speed assessment followed by a 30-min team-sport simulation on the non-motorised treadmill, on three occasions, 5 days apart. Several performance (total distance, distance covered during each speed category, total work, high-intensity activity, mean maximal sprinting speed and power) and physiological variables (V(O)(2), heart rate and blood measures) were measured. A one-way analysis of variance and ratio limits of agreement were used to compare the results from each trial. Significant differences were established in total sprint distance and high-intensity activity between trials 1-2 and trials 1-3 and 3-s mean maximal sprinting speed for trials 1-3 (p<0.05). No other significant differences were identified. Moderate to high intraclass correlation coefficients (i.e., >0.8) were identified in 11 of the 18 physiological and performance variables measured. Ratio limits of agreement for total distance covered and total work performed during the team-sport simulation were 0.99 (*//1.05) and 0.97 (*//1.09), respectively. Largest measurement error was shown in post-exercise blood lactate concentration with a coefficient of variation of 17.6%. All other measures showed low coefficients of variation of < or = 10%. These results show that the non-motorised treadmill team-sport simulation provides a reliable tool for assessing and monitoring physiological and performance demands of team-sport activity. We recommend the inclusion of two familiarisation sessions prior to testing.