Bayesian Games and the Smoothness Framework

Abstract

We consider a general class of Bayesian Games where each players utility depends on his type (possibly multidimensional) and on the strategy profile and where players' types are distributed independently. We show that if their full information version for any fixed instance of the type profile is a smooth game then the Price of Anarchy bound implied by the smoothness property, carries over to the Bayes-Nash Price of Anarchy. We show how some proofs from the literature (item bidding auctions, greedy auctions) can be cast as smoothness proofs or be simplified using smoothness. For first price item bidding with fractionally subadditive bidders we actually manage to improve by much the existing result \cite{Hassidim2011a} from 4 to $\frac{e}{e-1}\approx 1.58$. This also shows a very interesting separation between first and second price item bidding since second price item bidding has PoA at least 2 even under complete information. For a larger class of Bayesian Games where the strategy space of a player also changes with his type we are able to show that a slightly stronger definition of smoothness also implies a Bayes-Nash PoA bound. We show how weighted congestion games actually satisfy this stronger definition of smoothness. This allows us to show that the inefficiency bounds of weighted congestion games known in the literature carry over to incomplete versions where the weights of the players are private information. We also show how an incomplete version of a natural class of monotone valid utility games, called effort market games are universally $(1,1)$-smooth. Hence, we show that incomplete versions of effort market games where the abilities of the players and their budgets are private information has Bayes-Nash PoA at most 2.