We can obtain more information than just a description of `here` and `now` using the variance components and generalizability studies. Variance analysis models give us information about the variance of each chosen facet. The results obtained with the models confirm that the values of the least squares (LS) and maximun likelihood (ML) procedures coincide and therefore, the model can be regarded as linear; in other words, it fulfils the requeriments of homocedasticity, linearity and normality. Moreover, the results of the sample for each facet can be assumed to apply also to the population whence they originate. In both models, the context [C] facet shows the highest percent of variability of the model (round of 22%). The Behavior [B] facet, analyzed independently, represents a high level of the variability of the model [MCB] (21%). When it interacts with the Context [C], it has 51% of the total variability, which indicates that the interaction contexts are nested with the Behaviors. We can interpret this result as a difference in the way the teams play. The Area [A] facet contributes 7% of the variability of the model [MCA], but when it interacts with the Context [C], it has 62% of the total variability, which indicates that the interaction contexts are related to the area where they are more likely to appear. We also observe that most facets and their interactions are significant. The Match [M] facet shows how the play has evolved over the years. When we analyze it individually, the obtained variability percent is very low, near to 0%. Maybe, this means that over the eigth finals of the Eurocoup League, the teams have played almost the same way. In fact, when we combine it with the other facets, in any of the two models, the percent is 1% (similar to those obtained at World Cup study Castellano, Perea & Blanco-Villaseñor, 2007). The value of the coefficient of determination is very high (close to 1) and this indicates that the chosen facets explain almost all the variability of the models [0.90 in MCA and 0.91 in MCB]. It should be noted, however, that the value of the model`s residual error account, respectively, for 8% and 6% of the total. That is, we still do not have all the facets to complete and to explain the variability of the model, but we are optimist. In conclusion, the results were satisfactory. It is necessary to analyze more matches in future to confirm our theory that the teams play this championship in the same way. The estimated patterns (Borrie, Jonsson & Magnusson, 2002; Castellano & Hernández-Mendo, 1999 and 2002; Castellano, Hernández-Mendo, Morales-Sánchez & col., 2007) did not change along the years, which it is a very valuable information about the patterns in soccer.
© Copyright 2008 World Congress of Performance Analysis of Sport VIII. Veröffentlicht von Otto-von-Guericke-Universität, Department of Sports Science. Alle Rechte vorbehalten.