||This thesis studies two aspects of the complexity in combinatorial games. In the first part, we will look at the complexity to compute the winner of combinatorial games. We will try to determine the complexity of several games like Arc Kayles, Nim on graphs... We will also try to restrict these games to different classes of graphs. The goal of the second part is to define the complexity of a set of rules. Starting from existing concepts like invariant games and universal games, we study these concepts on sets for which the existence of such rules is still open with the objective to produce a classification of games by their rules.