Game Theory: Nash Equilibria and Bayesian Nash Equilibria
Game Theory: Nash Equilibria and Bayesian Nash Equilibria
Ayar Portugal
April 19, 2011
1
The expected payoffs of type 1 and type 2 of player 1 are:
τ1 : B τ1 : S
BB BS SB SS BB BS SB SS
B 3 2α + 1 3 − 2α 1 B 3 2α 2 − 2α 1
S 0 2 − 2α 2α 2 S 0 3 − 2α 2α + 1 2
Suppose that player 1’s actions are (B,B), ie B for type τ1 and B for type τ2 .
Then player 2’s best response is (B,B), which is also the best response of player
1’s for these actions. As any player has an incentive to deviate, the strategy
profile ((B, B), (B, B)) is a Nash Equilibrium for any beliefs.
Suppose that player 1’s actions are (B, S), player 2’s best response would
be (B, S) if 2β + 1 > 2 − 2β ⇔ β > 41 and if 3 − 2β > 2β ⇔ β < 43 . In other
words, if player 2’s believes 14 < β < 34 then player 2’s best response is (B, S),
which given the symmetry of this game is also the best response of player 1.
We’ve found a second NE which is the strategy profile ((B, S), (B, S)) for beliefs
1 3 1 3
4 < β < 4 and 4 < α < 4 .
Suppose that player 1’s actions are (S, B). Player 2’s best response would
also be (S, B) iff 3 − 2β < 2β ⇔ β > 34 and 2 − 2β > 2β + 1 ⇔ β < 41 , which is
not possible. We can therefore rule out ((S, B), (S, B)) as a NE. If β < 41 player
2’s best response would be (B, B), in this case player 1 would have an incentive
to deviate from (S, B) to (B, B) and so we find the first NE. If β > 43 player
2’s best response would be (S, S), in this case player 1’s best response would
also be (S, S), therefore we have found the third NE of this game which is the
strategy profile ((S, S), (S, S)).
2
(1 − µ(T ))(1 − µ(R))(1 − µ(r)) + µ(T )(1 − µ(R))(1 − µ(r)) + µ(R)µ(T )(1 −
µ(r)) + µ(T )(1 − µ(R))µ(r) + (1 − µ(T ))µ(R)µ(r) =
= 1−µ(r)−µ(R)+µ(R)µ(r)−µ(T )+µ(T )µ(R)+µ(T )µ(r)−µ(T )µ(R)µ(r)+
µ(T ) − µ(T )µ(r) − µ(T )µ(R) + µ(T )µ(R)µ(r) + µ(T )µ(R) − µ(T )µ(R)µ(r) +
µ(T )µ(r) − µ(T )µ(R)µ(r) + µ(R)µ(r) − µ(T )µ(R)µ(r) =
= 1 − µ(r) − µ(R) + 2µ(R)µ(r) − 3µ(T )µ(R)µ(r) + µ(T )µ(R)∗
b)
(B, T, l) is admissible (but is not a PNE) because for all players there is no
other strategy by which it is weakly dominated. For example, U1 (B, R, r) >
U1 (T, R, r).
Remember: A Perfect Nash Equilibrium need to be a Nash Equi-
librium. Moreover, in a 2-players game if a NE is admissible (i.e.
it is not weakly dominated by any other strategy) then it is also
a PNE.
Exercice 2
a)
As this is a 2 players game, the PNE can be found by considering the admissible
strategies among the Nash Equilibria. There are two Pure Nash Equilibria :
3
(T, L) and (B, R). (T, L) is not weakly dominated, it is therefore admissible
and PNE. Strategies (B, R) are weakly dominated by T and L, it is therefore
not PNE.
We now move to the ε−proper equilibrium. Consider the totally mixed
strategies α1 (T, B) and α2 (L, R).
Both inequalities hold because the more likely player 1 choses T, the greater
the expected the payoff of player 2 will be if he deviates to L, which in turn,
will be advantageous to player 1.
b)
Remember: Since an ε−proper equilibrium is also an ε−perfect
equilibrium, it follows that a proper equilibrium is perfect. Hence,
properness is a refinement of perfection.
Taking into account the reminder here above we will start by finding the proper-
equilibrium (equilibria).
We start with the NE (T, L, l).
Adding one strictly dominated strategy can indeed result in one aditional
proper equilibrium. As this new strategy (r) is strictly dominated player 3 will
never chose to deviate from (l), so we concentrate on player i = 1, 2.
We need to find a sequence εt such that the εt −proper equilibria converges
to what we think is the proper equilibrium.
Which holds for any ε > 0. Consider now the following sequences:
1 t
U1 (T, α2t , α3t ) − U1 (B, α2t , α3t ) = 1 − α2t (R)α3t (r) = 1 − = >0
1+t 1+t
1 t
U2 (L, α1t , α3t ) − U1 (R, α1t , α3t ) = 1 − α1t (B)α3t (r) = 1 − = >0
1+t 1+t
4
We also note that limt→∞ αit = 0. Therefore there is a sequence of ε−proper
equilibria which converges to (T, L, l), the proper equilibrium, as t → ∞.
Similarly, we will show that the profile (B, R, l) is also a proper equilibrium.
Define the following sequences:
1
α1t (T ) =
t+1
t 1
α2 (L) =
t+1
1
α3t (r) =
t
Take the conditions for (B, R, l) to be a ε−proper equilibrium and apply the
sequence as t → ∞:
1 1
U1 (B, α2t , α3t ) − U1 (T, α2t , α3t ) = α3t (r) − α2t (L) = − >0
t 1+t
1 1
U2 (R, α1t , α3t ) − U1 (L, α1t , α3t ) = α3t (r) − α1t (T ) = − >0
t 1+t
As in the previous case, this is a sequence of ε−proper equilibria converging
to (B, R, l) as t → ∞. Therefore (B, R, l) is a proper-equilibrium.
Exercice 3
a)
A profile of strategies is a Perfect Equilibrium if for a given function error µ
b)
qsdf