Disclosed is a method for evaluating an anti-fatigue effect of a composition comprising amino acids. The method comprises steps of: (i) measuring an amount of activity of a biological sample previously obtained from a subject administered with said amino acids as an index of muscle fatigue (ii) measuring concentration of a blood biomarker including cortisol, interferon gamma, ineterleukin-10 of the biological sample of the subject administered with said amino acids as an index of nervous fatigue (iii) determining whether the amount of activity of an administration group is 110 or greater relative to that of a non-administration group, which is set to be 100, and (iv) determining whether the measured biomarker concentration of the administration group is 96 or less relative to that of the non-administration group which is set to be 100. Further disclosed is an amino acid composition which preferably contains 30-200 parts by weight of proline, 60-140 parts by weight of glycine, 50-60 parts by weight of alanine, 50-130 parts by weight of lysine, 30-75 parts by weight of tryptophan and 20-40 parts by weight of histidine, 3-75 parts by weight of tyrosine and 15-45 parts by weight of arginine, 30-55 parts by weight of valine, 35-60 parts by weight of leucine and 25-60 parts by weight of isoleucine. Further disclosed are amino acid compositions containing varying parts by weight of the said amino acids.
