The present article uses agent-based social simulation to study rational behaviour in networked innovation. A simulation model that includes network characteristics and network participant's characteristics is run using parameter sweeping, yielding 1450 simulation cases. The notion of coalitions was used to denote partnerships in networked innovation. Coalitions compete against each other and several variables were observed for winning coalitions. Close analysis of the variations and their influence on the average power per winning coalition was analysed using stepwise multiple regression analysis. The analysis brought forward two main conclusions. First, as average betweenness centrality per winning coalition increases, the average power per winning coalition decreases. This implies that having high betweenness centrality as a network participant makes it easier to build a successful coalition, as a coalition needs lower average power to succeed. Second, as the number of network participants increases, the average power per winning coalition decreases. This implies that in a larger network, it may be easier to form a successful coalition. The results form the basis for the development of a utility-based recommendation system that helps people choose optimal partners in an innovation network.