Although Suzuki and Nishijima (2003) examined the construct validity of the soccer defending skill scale (SDSS), it is necessary to examine not only the construct validity, but also the sensitivity and cross validity of the scale to enhance its practicality. The sensitivity and cross validity of scales can be tested simultaneously by structural equation modeling methodologies in analysis of factorial invariance and latent mean structure. Therefore, the purpose of the present study was to examine the sensitivity and cross validity of the SDSS for its practical use using simultaneous analysis and latent mean structure analysis between teams. The samples were 469 defending performances in the final of FIFA World Cup Korea/Japan. Defending game performances were measured by 19 items of 6-point interval scales for each. The procedures for examining applicability of the SDSS were as follows: a) establishing and evaluating a baseline model by conducting single-sample SDSS models for Brazil (BRA) and Germany (GER); b) testing the invariance of the SDSS measurements and structures across the two teams; c) comparing mean values of latent variables between the teams. Results from the single-sample analysis of each team indicated that the SDSS multidimensional confirmatory factor analysis (SDSS Multi-CFA) models, in which three factors of the defending phases were combined with three factors of the defending objects for each team, fit the data well. From the results of simultaneous analysis across the two teams, the model that constrained the factor structure, factor loadings, and error variances to be equal between the teams showed the best fit to the data (Chi-square (42) = 44.103, CFI = .999, RMSEA = .001). The latent mean structure of the SDSS Multi-CFA model was tested between the teams. Comparison between the models for BRA and GER was made first by testing a model that constrained equal latent means in the two teams, then comparing the chi-square of that model with a model in which latent means were allowed to differ (i.e., a BRA model were fixed to zero, and a GER model were relaxed to obtain estimates). The difference between these two chi-square was 12.679, df = 6, P < .05, showing the fit was significantly better when latent means for teams were permitted to differ. Additionally, the parameter estimates in the unconstrained model revealed significant differences (BRA > GER) in latent means with respect to defending objects. In conclusion, the results demonstrated the applicability of the SDSS in team evaluation among high performance levels.Keyword(s): coaching, measurement/evaluation, performance