Figure 2

Qualitative description of how the consistency and identifiability of football teams are calculated. First, we obtained the multi-scale pitch passing networks of all matches and teams. Second, the adjacency matrix (see Fig. 1B) of each match and pitch partition is calculated. Third, we calculate the cross-correlation of all matrices for all scales. Each scale corresponds to a different number of nodes m. The scale-consistency C(m) (bottom left) is obtained when only the same team’s adjacency matrices are considered. Finally, we calculate the scale-identifiability I(m) (bottom right) subtracting the correlation with other teams from the scale-consistency C(m). The procedure is repeated for partitions of the pitch of different sizes, leading to a value of C(m) and I(m) that depends on the number of nodes m.