In games of complete information, there are common solution concepts such as Nash equilibrium and correlated equilibrium. The idea is that each player is playing a best response. My question is - Why do we need complete information for these solutions?
If I understand correctly, Mixed Nash equilibrium (and therefore also correlated equilibrium) is guaranteed in every (finite) game, regardless of the information players have on their opponents.
If the information is necessary in order to compute the equilibrium then I'm still not sure why. There are methods that don't require this information. For example, best response dynamics, in which on each turn, one player picks a "beneficial deviation". To do this, you need to know only your own utility function. Another example is no regret algorithms, that are used to compute correlated equilibrium. The Weighted Majority, each player updates his weights only based his cost vector.
Thanks!