Figure 2

Single-strategy model. (a) Predictive accuracy of the single-strategy model as a function of the game aggregation factor α. Each point represents the average of a 5-fold cross-validation; error bars indicate the standard error of the mean. The solid black line represents the accuracy of the baseline model (see text and Ref. [13]). (b) Average number of groups of players \(\langle N_{g}\rangle\) as a function of the game aggregation factor α. Each point represents the average of the number of groups for players identified for each of the 5 folds; error bars indicate the standard error of the mean. (c)–(e) Groups of games in the TS-plane (top) and cooperation matrix p (bottom) for aggregation factors α equal to: (c) 0, (d) 1, (e) 2, (f) 4. In the game plots, each color indicates a different group of games in the most plausible partition as obtained from Eq. (6) for each value of α. In the cooperation matrices, each row corresponds to a group of players k (with height proportional to the number of players in the group), and each column to a group of games ℓ as indicated by the colors at the top (with width proportional to the number of games in the group). Each element \(p_{k\ell}\) represents the probability that an individual in group k cooperates when playing a game in group ℓ, with dark red meaning always cooperate and dark blue always defect. The game groups and the cooperation matrices correspond to one of the folds in the 5-fold cross-validation, but are very consistent across folds (see Fig. S5)