Mathematical reasoning involves the activation of frontal and parietal lobes, due to the requirement of working memory, attention, and retrieval mechanisms. Efficient activation of different brain networks, as suggested by previous EEG studies, is related to the level of mathematical ability. We performed a paired comparison of mutual information and power spectrum during math calculations against rest in two groups of participants with different levels of mathematical competence, defined as good and bad counters. Results were displayed through planar head diagrams showing EEG derivations as nodes connected by edges representing significant increases or decreases in mutual information. Power spectrum analysis revealed that good counters presented significant cortical activation in delta, theta, beta and gamma frequency bands and strong alpha blocking during math in comparison to rest. Mutual information analysis showed coupling increases involving more edges, which were topographically widespread in subjects with less arithmetical ability. Decreases in mutual information were prominent in good counters in the beta and alpha frequency bands, while bad counters showed both coupling and uncoupling the alpha band. In both arithmetical ability groups, decoupling was prominent in the prefrontal and parietal areas, particularly in the theta, alpha and beta bands, confirming the importance of frontoparietal activity during mathematical reasoning. Our results are in line with the neural efficiency hypothesis and indicate that those with greater mathematical competence present activation of fewer edges in defined cortical areas, whereas those with lower proficiency require the recruitment of additional edges to complete the task.