We address several recent developments in non-cooperative as well as evolutionary game theory, that give a new viewpoint to Complex Systems understanding. In particular, we discuss notions like the anarchy cost, equilibria formation, social costs and evolutionary stability. We indicate how such notions help in understanding Complex Systems behaviour when the system includes selfish, antagonistic entities of varying degrees of rationality. Our main motivation is the Internet, perhaps the most complex artifact to date, as well as large-scale systems such as the high-level P2P systems, where where the interaction is among humans, programmes and machines and centralized approaches cannot apply.