The purposes of this study were to determine the anaerobic running capacity (ARC) and to study the relationship between statistic ARC values and real exercise performance. Subjects were 16 female runners. These runners were tested on five short-distances running with balanced order. The ARC data were obtained from the calculations using non-linear velocity-time model, linear model of velocity-1/time model, linear model of distance-time and non-linear model of three parameters. These ARC data were then analyzed by ANOVA with one-way repeated measures to study the significance of ARC values among those different models. The correlation between these ARC values and other anaerobic capacity tests, the measurements of vertical jump, Margaria-Kalamen test, and Wingate test on runners were also conducted. Pearson's product-moment statistic model was then employed to determine the correlation between the ARC values obtained from the statistical model and anaerobic capacity tests. The ARC values calculated from each model reached significant difference in this study, especially, the ARC determined by the 3P model was the better representive of all the anaerobic running distance tests. The correlation between ARC of two non-linear model and anaerobic tests showed significant correlations, especially, the ARC determined by the 3P model revealed a high correlation (r=.810~.830). It is concluded that the 3P non-linear model is the best model for evaluating ARC in female short-distance runners.