Power index and sharing centrality are highly correlated with the economic benefit in a sharing network. a Graph-theoretic power index (GPI) captures the probability the focal node will be engaged in a successful mutual selection with one of his or her alters. This in turn, gives the node a structural advantage in bargaining in strategic interactions. The easiest way to understand how the geodesic position benefits a node is the following simple scenario. Suppose node A randomly selects one of her neighbors (say node C) in order to allocate all its resource; then if C has also selected node A as its unique collaborator, asymmetric and fully reciprocal relationship can be established and both node C and A will stop seeking collaborators. Otherwise, node A can proceed and select another node as a potential collaborator. This iterative process continues until a mutual selection is achieved for every possible pair of nodes. It is easy to see that each dyad has a different probability of mutual selection throughout this iterative process. Namely, the probability of mutual selection (by concurrent nomination in an repetitive process) is 0.6 for the edges a, b and c, d, and 0.2 for a–c and a–d. In total, this means that node A can find a full reciprocator with probability 1, while for node B the probability is 0.6 and for nodes C and D it is 0.8. Sharing centrality is the sum of reciprocal network degree among alters. The centrality approximates the structural advantage, which GPI represents, but using only local topology information. b–d Dots indicate several measures of network centrality for each subject and his or her total received units at the end of game (out of N = 600 subjects who participated). Dot color indicates the network’s connection radii; red for radius = 0.25, gray for radius = 0.30, and blue for radius = 0.35. GPI and sharing centrality show clear correlation with their economic benefit from a sharing game (b, c). This correlation is higher than all the other centrality measures (d). The “r” values indicate Pearson correlation coefficients (P < 0.01 for all the correlation coefficients)