Since the 1980s, scientists study the distribution of power in social networks - mostly via conducting experiments on exchange flows in networks. The results of such experiments then allow drawing conclusions about the bargaining power of actors and, hence, give an indication of their power potential. Studying respective results one might get the impression that the distribution of power in social networks is closely related to all actors' structural positions. Classical social network theory provides indices [degree centrality, betweenness centrality etc.J to evaluate the structural position of actors. However, such indices are either not in line with experimental findings or have technical issues, i.e. pre-specification of parameters. Therefore, first, we present an entropy driven approach, which is based on a probabilistic conditional-logical framework, to determine the power potential of actors based on their structural positions - without the aforementioned shortcomings. Second, we show how to create a specific type of dominating set, using only the entropy-based power potential of actors; we call this type of dominating set a 'power dominance network'. Numerical examples demonstrate the gist of the new method.