There are many functional methods for estimating the mean axis of rotation of a joint. However, it is still a controversial issue which method is superior. The purpose of this study was to compare functional methods for estimated axes of rotation from synthetic data. The comparison was made in terms of suitabilities on describing humans in sports. For a more practical situation, the axis error as well as measurement and marker movement error were applied to generated data. Simulations having 1000 times of 80 rotational displacements were performed. The functional methods used in the study were two transformation methods, two fitting methods, and one more transformation method called M. The M method is a combination of S$ddot{o}$derk & Wedin(1993) and Mardia & Jupp(2000). Another factor of the study was angular velocity with levels of .01, .025, .05, .5 and 1 rad/s. The method M resulted in unbiased, stable, and consistent axis of rotation vectors in all levels of angular velocity except .01 rad/s. Therefore, the method M had the highest validity and reliability of all the methods. The fitting methods were very sensitive in small angular velocities and stable only in the velocities of more than .5 rad/s. The most suitable method for analyzing human motion by using marker photogrammetry is M.