(separated by network structure, ordered by # of data-lucky agents)
Each agent's judgment is 1 if data-lucky (its best-supported theory is the optimal one) and 0 otherwise. For each run, p is the proportion of agents who are data-lucky.
Social variance (disagreement within runs): the average over runs of p(1 − p).
Modal variance (unpredictability between runs): the variance of p across runs.
Total variance: the variance of every agent's judgment across every run. It equals social variance plus modal variance.
Each panel shows the runs ordered by number of data-lucky agents, one line per k, as in the Model tab. Rows vary the number of theories; columns vary Var(θ), the spread of the theories' true success rates. Problems generally get easier toward the top right.