I have a question about Theorem 1.2 in Game Theory by Fudenberg and Tirole.

Theorem 1.2 in FT(Debreu, Glicksberg, Fan) Consider a strategic-form game whose strategy spaces $S_{i}$ are nonempty compact convex subsets of an Euclidean space. If the payoff functions $u_{i}$ are continuous in $s$ and quasi-concave in $s_{i}$, there exists a pure-strategy Nash Equilibrium.

Theorem 1.1 in FT (Nash): Every finite strategic-form game has a mixed strategy equilibrium.

My question here is that, how do we know that the Nash Equilibrium in theorem 1.2 is a pure-strategy one?

In the proof, they also use Kakutani's fixed point theorem, but I think it can only give us the existence of some Nash Equilibrium, and we know nothing about its characterizations. Additionally, Nash's existence theorem should be a special case of theorem 1.2 since the set of mixed strategies over finite actions is a simplex, which is compact. However, in Nash's theorem, we only know that a mixed-strategy equilibrium will exist.

My understanding of Debreu's theorem is that, the "pure strategy" in this theorem is with respect to the strategy spaces, not to action spaces. To be more specific, take the Nash Existence Theorem example above, the mixed strategy in the theorem is actually a "pure strategy" in the space of mixed strategies, so both theorems don't contradict with each other.

  • 3
    $\begingroup$ The usual proof of Theorem 1.2. uses a setting in which fixed points correspond to Nash equilibria in pure strategies. One can prove Theorem 1.1 from 1.2 by showing that the mixed extension of a finite game has a pure strategy equilibrium. This pure strategy equilibrium of the mixed extension is then a mixed equilibrium of the underlying finite game. $\endgroup$ Dec 27, 2021 at 16:53
  • $\begingroup$ @MichaelGreinecker could you elaborate a bit more on the first sentence? i.e The proof uses a setting in which fixed points correspond to Nash equilibria in pure strategies? And what is the "mixed extension" of a finite game? $\endgroup$
    – Borel
    Dec 27, 2021 at 20:02
  • 2
    $\begingroup$ In the mixed extension, you take your strategy space to be the space of probability distributions over the original strategies. Another way to see the connection: In Nash's proof, one uses the mixed strategy best reply correspondence, while can also work with the pure strategy best reply correspondence under the assumptions of Theorem 1.2. $\endgroup$ Dec 27, 2021 at 22:20

1 Answer 1


I assume you know how the proof goes using Kakutani to prove existence for finite games.

As standard, let $X_{-i} = \times_{j \not = i} X_{j}$.

Define the best response correspondence $BR_i: S_{-i} \rightrightarrows S_i$ by $$ BR_i(s_{-i}) = \arg \max_{s_i \in S_i} u_i(s_i,s_{-i}) $$

Apart from the usual existence and upper hemicontinuity, quasiconcavity of $u_i$ implies the correspondence is convex valued.

As usual, stack them up to define the correspondence $BR: S \rightrightarrows S$.

Apply Kakutani, the desired fixed point is an element of $s^* \in S$, and is thus a pure strategy.

Compare this to the proof for finite games, where we define the correspondence on the space $\times_{i} \Delta (S_i)$!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.