IESDS

Itereated Elimination and Nash Equilibria
We consider finite two player games–though all of these will generalize to

any finite game. We let A denote the set of strategies for Player 1 and B denote
the strategies for Player 2.
Recall that when we say a dominates a0 we mean that it weakly dominates
0
a.
Proposition 1 Any game as at most one dominant solution.
Proof It is impossible for a to dominate a1 and a1 to dominate a.
Proposition 2 If (a∗ , b∗ ) is a dominant solution, then (a∗ , b∗ ) is a Nash equi-

librium.
Proof The strategy a∗ dominates every other strategy in A. Thus
v1 (a∗ , b∗ ) ≥ v(a, b∗ )
for all a ∈ A and a∗ is the unique best response to b∗ . Similarly, b∗ is the unique
best response to a∗ . Thus (a∗ , b∗ ) is a Nash equilibrium.
The game Battle of the Sexes shows that the coverse fails as there are games
with Nash equilibria which are not strictly dominant.
Proposition 3 If (a∗ , b∗ ) is a Nash equilibrium, then (a∗ , b∗ ) is not eliminated

by IESDS.
Proof We prove this by contradiction. Suppose (a∗ , b∗ ) is eliminated during

IESDS. Then one of the strategies is removed at some stage of the construction.
Let’s suppose that a∗ is removed before b∗ (the other case is similar). Consider
the stage when a∗ is eliminated. At this stage of the construction, we have a
game where a∗ and b∗ are possible strategies and, because it is about to be
eliminated, there is a strategy a0 ∈ A1 such that a0 strictly dominates a∗ . But
then
v1 (a0 , b∗ ) > v1 (a∗ , b∗ )
and a∗ is not a best response for Player 1 to b∗ . This contradicts our assumption
that (a∗ , b∗ ) is a Nash equilibrium.
The converse fails. For example, in Battle of the Sexes, (F, O) and (O, F ) are
not eliminated by IESDS (or even IEDS) but are not Nash equilibria. Similarly,
in Mathching Coins no strategies are elimated by IESDS (or IEDS) but no
strategy profile is a Nash equilibrium.
The converse is true in the special case when there is IEDS solution–i.e.
some IESDS procedure ends in a unique solution.
1
Proposition 4 If (a∗ , b∗ ) is an IEDS solution, then (a∗ , b∗ ) is a Nash equilib-
rium.
Proof For purposes of contradiction suppose a∗ is not a best response to b∗ –the

other case is similar. Let X = {a ∈ A : v1 (a, b∗ ) > v1 (a∗ , b)}. By assumption
A 6= ∅. All of the strategies in X must be eliminated in the process. Look at
the last stage where a strategy a ∈ X is elimiated. For it to be eliminated, there
must be a strategy a0 ∈ A such that a0 dominates a. But then
v1 (a0 , b∗ ) ≥ v1 (a, b∗ ) > v1 (a∗ , b∗ )
and a0 ∈ X. But we must eventually eliminate a0 and this contradicts the fact
that a was the last element of X eliminated.
Corollary 5 If there is an IESDS solution, it is the unique pure strategy Nash

equilibrium.
Proof By Proposition 4 the unique IESDS equilibrium is a Nash equilibrium.

By Proposition 3, if there was a second Nash equilibrium it would also be an
IESDS equilibrium.
Corollary 6 If there is a strongly dominant strategy equilibrium, it is the unique

Nash equilibrium.
Proof If (a∗ , b∗ ) is a strictly dominant strategy equilibrium, then in the IESDS

process at stage 1 would eliminate all strategies except a∗ and b∗ , so (a∗ , b∗ ) is
the unique IESDS-equilibrium and hence the unique Nash-equilibrium.
Example 2 below shows that a game may have a dominant solution and
several Nash equilibria.
Corollary 7 There can only be at most one IESDS solution. In particular, in

IESDS the order that we eliminate strategies does not matter.
Proof If there were two IESDS solutions, then would both be Nash equilibria
contradicting Corollary 4.
More generally, the set of strategies that survive IESDS elimination does not
depend on the order of elimination.
Example 1 In IEDS the order of elmination may matter. We also note that
this is a game solvable by IEDS with two Nash equilibria.
Consider the game:
L R
T 1,1 0,0
M 3,2 2,2
B 0,0 1,1
Process 1: Since M dominates T, we can eliminate T to get
2
L R
M 3,2 2,2
B 0,0 1,1
Now R dominates L. Eliminating L we get
R
M 2,2
B 1,1
and (M,R) is an IEDS solution.

Process 2:
M dominate B
L R
T 1,1 0,0
M 3,2 2,2
Now L dominates R
L
T 1,1
M 3,2
Thus (M,L) is an IEDS solution.
Example 2 The coordination game below is a game with two Nash equilibria
only one of which is an IEDS solution–and no IESDS solution
L R
T 1,1 0,0
B 0,0 0,0
In this game (T,L) is the unique IEDS solution, indeed it is a dominant

solution, but (B,R) is also a Nash equilibrium.
IESDS does not simplify this game at all.
Example 3 The following game has a unique pure strategy Nash equilibrium
but can not be simplified by IEDS
L C R
T 2,2 -1,1 1,0
M 1,-1 0,0 1,-2
B 0,1 2,-1 -2,3

IESDS

Uploaded by

Copyright:

Available Formats

IESDS

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

IESDS

Uploaded by

Copyright:

Available Formats

Itereated Elimination and Nash Equilibria

We consider finite two player games–though all of these will generalize to

Proposition 1 Any game as at most one dominant solution.

Proof It is impossible for a to dominate a1 and a1 to dominate a.

Proposition 2 If (a∗ , b∗ ) is a dominant solution, then (a∗ , b∗ ) is a Nash equi-

Proof The strategy a∗ dominates every other strategy in A. Thus

Proposition 3 If (a∗ , b∗ ) is a Nash equilibrium, then (a∗ , b∗ ) is not eliminated

Proof We prove this by contradiction. Suppose (a∗ , b∗ ) is eliminated during

Proof For purposes of contradiction suppose a∗ is not a best response to b∗ –the

v1 (a0 , b∗ ) ≥ v1 (a, b∗ ) > v1 (a∗ , b∗ )

Corollary 5 If there is an IESDS solution, it is the unique pure strategy Nash

Proof By Proposition 4 the unique IESDS equilibrium is a Nash equilibrium.

Corollary 6 If there is a strongly dominant strategy equilibrium, it is the unique

Proof If (a∗ , b∗ ) is a strictly dominant strategy equilibrium, then in the IESDS

Corollary 7 There can only be at most one IESDS solution. In particular, in

Process 1: Since M dominates T, we can eliminate T to get

Now R dominates L. Eliminating L we get

and (M,R) is an IEDS solution.

Thus (M,L) is an IEDS solution.

In this game (T,L) is the unique IEDS solution, indeed it is a dominant

You might also like