Lecture 8: Practice: Nash Equilibria in Mixed

Strategies
Vera Sharunova

This time we will dedicate the entire lecture to solving various practice problems so
that you can get more comfortable with finding mixed-strategy Nash equilibria. We will
revisit the crush game and consider examples of games when players have more than
two strategies.

1 Meeting Your Crush at a Dining Hall

Let’s recall our crush coordination game from two weeks ago. The game is represented in
its normal form below. There are two pure-strategy Nash equilibria (McElroy, McElroy)
and (Corcoran, Corcoran). But which of the two is going to be played in reality? With
multiple pure Nash equilibria, it is hard to make a prediction of the outcome of the
game. You may play McElroy because you think that (McElroy, McElroy) is going to
be realized, and your crush may choose Corcoran thinking that (Corcoran, Corcoran) is
going to be realized. Should you both think that way, the outcome of the game would
be (McElroy, Corcoran) which is not even a Nash equilibrium. So it is fair to say that
you are fundamentally uncertain about the actions of your crush. What should you do
in this case?

Your Crush
q 1−q
McElroy Corcoran
McElroy p 3, 2+ 1, 1
You
Corcoran 1−p 0, 0 2, 3+

(a) If you and your crush were trying to play it “safe” and choose strategies that maxi-
mize your payoffs in the worst-case scenario, which strategies would you choose? What
would the outcome be in this case?

You may want to choose a strategy that maximizes the payoff you will get in the worst-
case scenario. For instance, if your crush goes to McElroy, the worst-case scenario is
for you to end up at Corcoran and get a payoff of 0. However, if your crush goes to
Corcoran, the worst-case scenario is for you to end up at McElroy and get a payoff of

1
ECON3308.01: Game Theory in Economics Summer 2021

1. In this case, you may want to choose McElroy and guarantee yourself a minimum
payoff of 1, as opposed to going to Corcoran and risking to get a payoff of zero. If your
crush acts as cautiously as you are, they will choose to go to Corcoran and guarantee
themselves a payoff of 1.

(b) Find the mixed-strategy Nash equilibrium of this game.

Let us start by finding expected payoff from your pure strategies McElroy and Corcoran,
given that the crush is choosing a mixture q between their pure strategies.

EUY (McElroy, q) = 3q + (1 − q) = 2q + 1
EUY (Corcoran, q) = 2(1 − q) = 2 − 2q

Recall that maximization of a player’s expected payoff from the entire game is equivalent
to setting the two payoffs from pure strategies, given the actions of the opponent, equal
to each other. This is because the FOCs that come out of the maximization problem
can be rearranged to represent the opponent’s indifference property.

So, by the opponent’s indifference property, the crush’s optimal mixture should make
you indifferent between going to McElroy or Corcoran.
1
EUY (McElroy, q) = EUY (Corcoran, q) = 2q + 1 = 2 − 2q ⇒ q ∗ =
4
You could think of the opponent’s indifference property this way: For the mixture be-
tween two pure strategies to be a best response to the other player’s strategy, both of the
pure strategies have to be best responses to the other player’s strategy and, therefore,
deliver equal payoffs. If they do not deliver equal payoffs, one of the strategies is not a
best response and should not be used. But if that is the case, the player should choose
a pure strategy, and not a mixture. Thus, the opponent’s indifference property
is an intrinsic characteristic of mixed-strategy Nash equilibria.

Similarly for your crush

EUC (McElroy, p) = 2p + 0(1 − p) = 2p

EUC (Corcoran, p) = p + 3(1 − p) = 3 − 2p
3
EUC (McElroy, p) = EUY (Corcoran, q) = 2p = 3 − 2p ⇒ p∗ =
4

So the mixed-strategy NE of this game is ( 34 McElroy+ 14 Corcoran, 14 McElroy+ 43 Corcoran).

(c) Calculate the expected payoffs in the mixed-strategy equilibrium. Are you better off
playing the “safe” equilibrium from (a) or the mixed-strategy Nash equilibrium? What

2
ECON3308.01: Game Theory in Economics Summer 2021

is the probability of you two meeting?

Let us calculate the expected payoffs in the equilibrium. There are two ways of doing
this: (i) calculating the probability that the game ends up in a given outcome and
multiplying it by the corresponding payoff, or (ii) simply plugging q ∗ and p∗ in one of
the pure strategy payoffs, since the opponent’s indifference property holds in equilibrium.
3 1 3 3 1 1 1 3
EUY (p∗ , q ∗ ) = · ·3+ · ·1+ · ·0+ · · 2 = 1.5
4 4 4 4 4 4 4 4
3 1 3 3 1 1 1 3
EUC (p∗ , q ∗ ) = · ·2+ · ·1+ · ·0+ · · 3 = 1.5
4 4 4 4 4 4 4 4
or

1 1
EUY (p∗ , q ∗ ) = EUY (McElroy, q ∗ ) = EUY (Corcoran, q ∗ ) = 2 · + 1 = 2 − 2 · = 1.5
4 4
3 3
EUC (p∗ , q ∗ ) = EUC (McElroy, p∗ ) = EUY (Corcoran, p∗ ) = 2 · = 3 − 2 · = 1.5
4 4
As you can see, it is very hard to meet your crush at a dining hall, since the probability
of you two meeting is only 2 · 43 · 14 = 83 . But you guarantee yourself a higher expected
payoff of 1.5 by playing the MSNE relative to the “safe” choice of always going to McEl-
roy.

(d) Graph yours and crush’s best responses on a p-q coordinate plane. Label all Nash
equilibria.

See Figure ??. The two best responses intersect in three points: (i) (0,0) which represents
the pure-strategy Nash equilibrium (Corcoran, Corcoran); (ii) (1,1) which represents the
pure-strategy Nash equilibrium (McElroy, McElroy); and (iii) (1/4, 3/4) which represents
the mixed-strategy Nash equilibrium we have just found.

2 Another Example of a 3x2 Game

This game is discussed in Section 7.6 of the textbook in a setting of a tennis point. I
know nothing about tennis, so I will use the same payoffs, but abstract strategies and
players.
(a) Does this game have a pure-strategy Nash equilibrium? Does Player 1 have a strategy
that is never a best response to a pure strategy chosen by Player 2?

In this game, there are no pure-strategy Nash equilibria and strategy E is never a best
response to any of the pure actions of Player 2. However, it may be a best response
to a mixed strategy of Player 2. Let us verify this fact by graphing Player 1’s expected

3
ECON3308.01: Game Theory in Economics Summer 2021

p = Pr(You go to McElroy)
(McElroy, McElroy)
1
Your best response

3
(p∗ , q ∗ )
4
Crush’s best response

(Corcoran, Corcoran) q = Pr(Crush goes to McElroy)

1 1
4
Figure 1: Your and your crush’s best responses as a function of each other’s mixtures.

Player 2
q 1−q
A B
|
C 50, 50 80, 20−
−
Player 1 D 90, 10 20, 80|
E 70, 30 60, 40|

payoffs as a function of the opponent’s mixture q.

(b) Suppose that Player 2 chooses to play a mixture q between her pure strategies A and
B, assigning a probability of q towards playing A. What are Player 1’s expected payoffs
from his pure strategies as a function of q?

EU1 (C, q) = 50q + 80(1 − q) = 80 − 30q

EU1 (D, q) = 90q + 20(1 − q) = 20 + 70q
EU1 (E, q) = 70q + 60(1 − q) = 60 + 10q

(c) Graph Player 1’s expected payoffs in a EU1 -q plane. Does Player 1 have a strategy
that is never a best response to any mixture of the opponent q?

See Figure 2 below. There is no strategy of Player 1 that would not be a best response
to a potential mixture q of Player 2. In fact, the best response of Player 1 is given as

4
ECON3308.01: Game Theory in Economics Summer 2021

follows: 

 C if q < q̃




C, E if q = q̃
BR1 = E if q̃ < q < qb



 E, D if q = qb


D if q > qb

EU1 EU1

C
E

q
0 1 2 1
2 3

Figure 2: Expected payoffs of Player 1 from choosing either C, D, or E as a function of

(d) Calculate
their the expected
belief about payoffs
Player 2’s of Player 1 under Player 2’s optimal mixture between
actions.
C and E, call it q̃. Calculate the expected payoffs of Player 1 under Player 2’s optimal
mixture between E and D, call it qb.

Let’s start by finding the values of the two thresholds q̃ and qb. Mixture q̃ involves pure
strategies C and E, therefore set EU1 (C, q) equal to EU1 (E, q):
1
EU1 (C, q̃) = EU1 (E, q̃) ⇒ 80 − 30q = 60 + 10q ⇒ q̃ =
2

Mixture qb involves pure strategies E and D, therefore set EU1 (E, q) equal to EU1 (D, q):
2
EU1 (E, qb) = EU1 (E, qb) ⇒ 20 + 70q = 60 + 10q ⇒ qb =
3

5
ECON3308.01: Game Theory in Economics Summer 2021

Now we can calculate Player 1’s expected payoffs from choosing their pure strategies if
Player 2 chooses q̃ or qb:
1 1
EU1 (C, q̃) = EU1 (E, q̃) = 80 − 30q̃ = 60 + 10q̃ = 80 − 30 · = 60 + 10 · = 65
2 2
2 2
EU1 (E, qb) = EU1 (D, qb) ⇒ 20 + 70b q = 20 + 70 ·
q = 60 + 10b = 60 + 10 · ≈ 66.67
3 3
(e) Given the zero-sum nature of the game, which mixture will be optimal for Player 2,
q̃ or qb? Given this observation, which pure strategy of Player 1 will not be used in his
optimal mixture p∗ ?

The choice between mixtures q̃ and qb is under Player 2’s control. They will choose
whichever mixture gives them a higher payoff. Given that the game is basically a zero-
sum game (a constant-sum game summing up to 100), we can calculate the payoffs of
Player 2 as:
EU2 (·, q̃) = 100 − EU1 (·, q̃) = 100 − 65 = 35
EU2 (·, qb) = 100 − EU1 (·, qb) = 100 − 66.67 = 33.33

Therefore, Player 2 will choose mixture q̃, and Player 1 will not be using D in their
optimal mixture, since q̃ only has them mixing between C and E.

(f ) Using your answer in (e), reduce the game to 2×2. What is the optimal mixture p∗
of Player 1 in this reduced game?

The reduced game is given by

Player 2
q 1−q
A B
C p 50, 50| 80, 20−
Player 1
E 1−p 70, 30− 60, 40|

By the opponent’s indifference property, Player 1’s optimal mixture will make Player 2
indifferent between playing pure strategies A or B:
1
EU2 (p, A) = EU2 (p, B) = 50p+30(1−p) = 20p+40(1−p) ⇒ 30+20p = 40−20p ⇒ p∗ =
4

Player 2’s best response is given by:



B if p < 1
4
1
BR2 = A, B if p =


4
1
A if p > 4

6
ECON3308.01: Game Theory in Economics Summer 2021

(g) For the reduced game, graph the best responses of the two players in a p-q plane.
Mark the MSNE.

We already know that Player 2’s optimal mixture q̃ involves mixing between C and E
with probability 1/2. Therefore, the best response of Player 2 in the reduced game is:


C if q < 12
BR1 = C, E if q = 12


E if q > 12
Now we are ready to graph the two best responses in the reduced game. See Figure 3
below.

p
(C,A)
1

3
(p∗ , q ∗ )
4

(E,B) q
1 1
2
Figure 3: Player 1’s and Player 2’s best responses as a function of each other’s mixtures.

(h) Does Player 1 have an incentive to deviate from p∗ if Player 2 uses her optimal
mixture? Verify if there are profitable pure-strategy deviations.

In order to see whether Player 1 has an incentive to deviate from mixture p∗ (when Player
2 keeps using q̃), we need to see if deviating to any of the pure strategies available to
Player 1 achieves a higher payoff than the mixture p∗ . If there are no profitable deviations
to pure strategies, there can be no profitable deviations to any mixed strategy, because
alternative mixtures will involve mixing over pure strategies that we have shown to
achieve a lower payoff than p∗ .
1 1
EU1 (C, q̃) = 50 · + 80 · = 65
2 2
7
ECON3308.01: Game Theory in Economics Summer 2021

1 1
EU1 (E, q̃) = 70 · + 60 · = 65
2 2
1 3
EU1 (p∗ , q̃) = p∗ EU1 (C, q̃) + (1 − p∗ )EU1 (E, q̃) = · 65 + · 65 = 65
4 4
1 1
EU1 (D, q̃) = 90 · + 20 · = 55
2 2
Therefore, there is no incentive for Player 1 to deviate to any of the pure strategies and,
as a consequence, to any alternative mixture. Assuming that Player 2 keeps playing q̃,
deviating to a different mixture between C and E will achieve the payoff of 65 as well.
Similarly, choosing a mixture that assigns a positive probability to strategy D, can only
decrease the payoff.

3 An Example of a 3 × 3 Game (Optional)

Now, let us consider a 3 × 3 game, using the game of penalty kicks in soccer as an
example. Consider the following normal form with success percentages as payoffs for
each player. In general, it is possible that players will use all of the three pure strategies
in their optimal mixtures, but not in this particular example. See Section 7.7A of the
textbook for such an example.

Goalie
qL qC qR
Left Center Right
Left pL 45, 55 90, 10 90, 10
Kicker Center pC 70, 30 0, 100 70, 30
Right pR 95, 5 95, 5 60, 40

(a) Find the goalie’s equilibrium mixture, assuming that he mixes over all of his pure
strategies.

Notice that we cannot easily graph the expected payoffs of any of the players in this
game, since it would require us to use a three-dimensional coordinate plane: one di-
mension for the expected payoffs and two dimensions for probabilities qL and qC in the
mixture. Therefore, we must start by assuming that both players will use all three
strategies in their equilibrium mixtures.

In order to find the goalie’s equilibrium mixture, we must make the kicker indifferent
between choosing Left, Center, or Right, given that the goalie chooses mixture q =
(qL , qC , qR ). Substitute qC = 1 − qL − qR .

EUK (Left, q) = 45qL + 90qC + 90qR = 45qL + 90(1 − qL − qR ) + 90qR = 90 − 45qL

8
ECON3308.01: Game Theory in Economics Summer 2021

EUK (Center, q) = 70qL + 0qC + 70qR = 70qL + 70qR

EUK (Right, q) = 95qL + 95qC + 60qR = 95qL + 95(1 − qL − qR ) + 60qR = 95 − 35qR

In the mixed-strategy equilibrium,

EUK (Left, q) = EUK (Center, q) = EUK (Right, q)

{ { {
90 − 45qL = 95 − 35qR qR = 5+45q
35
L
qR = 5+45q
35
L

90 − 45qL = 70qL + 70qR 115qL + 70qR = 90 115qL + 70 5+45q

35
L
= 90
{
qR = 5+45q
35
L 185
= 287
qL = 16
41

This means that qC = 1 − 185

287
− 41
16
= − 287
10
, which violates non-negativity of probabilities.
Therefore, a strategy profile where the goalie uses all of his pure strategies cannot be a
mixed-strategy Nash equilibrium.

(b) In light of your findings in (a), what should the goalie’s probability of using Center,
qC , be in the equilibrium mixture? Would the kicker be willing to use Center in this case?

The best the goalie can do is to set qC as low as possible – namely, to zero. But then
the kicker would not use Center, since it will never be a best response to any mixture of
the goalie. See Figure 4 below. Therefore, if the goalie does not use Center, the kicker
will not do so as well.

EU2 EU2

95
90 ght
Lef
t Ri

Center
70 70

60

45
p = Pr(goalie chooses Left)

Figure 4: Expected payoffs of the kicker from choosing either Left, Center, or Right as
a function of his belief about the goalie’s actions.

(c) Given your answer in (b), what is the mixed-strategy Nash equilibrium in this game?

9
ECON3308.01: Game Theory in Economics Summer 2021

Now we must check if both players mixing only over Left and Right is a mixed-strategy
equilibrium.

Again, start with the kicker’s payoffs. Substitute for qR = 1 − qL , since we set qC = 0.

EUK (Left, q) = 45qL + 90qR = 45qL + 90(1 − qL ) = 90 − 45qL

EUK (Right, q) = 95qL + 60qR = 95qL + 60(1 − qL ) = 60 + 35qL

By the opponent’s indifference property, 90 − 45qL = 60 + 35qL ⇒ qL∗ = 38 .

Similarly for the goalie

EUG (Left, q) = 55pL + 5pR = 55pL + 5(1 − pL ) = 5 + 50pL

EUG (Right, q) = 10pL + 40pR = 10pL + 40(1 − pL ) = 40 − 30pL
7
5 + 50pL = 40 − 30pL ⇒ p∗L =
16

So the candidate mixed-strategy Nash equilibrium is ( 38 Left + 58 Right, 16

7 9
Left + 16 Right).

Are we done? Not yet. We need to verify that none of the player’s wants to bring Center
back into the mix, keeping the mixture of the opponent fixed.

Start with the kicker

3 5
EUK (Left, q ∗ ) = · 45 + · 90 = 73.125
8 8
3 5
EUK (Right, q ∗ ) = · 95 + · 60 = 73.125
8 8
7 9
EUK (p∗ , q ∗ ) = · EUK (Left, q ∗ ) + · EUK (Right, q ∗ ) = 73.125
16 16
3 5
EUK (Center, q ∗ ) = · 70 + · 70 = 70
8 8
The payoff from deviating to pure Center is less then the payoff from using pure Left
or Right or any other mixture over these three strategies.1 Therefore, the kicker has no
incentive to deviate from his mixed strategy p∗ .

Similarly for the goalie

7 9
EUG (Left, p∗ ) = · 55 + · 5 = 26.875
16 16
1
Since the expected payoff of a mixed strategy is just a weighted average between the expected
payoffs of pure strategies, using the mixture’s probabilities as weights.

10
ECON3308.01: Game Theory in Economics Summer 2021

7 9
EUG (Right, p∗ ) = · 10 + · 40 = 26.875
16 16
3 5
EUG (q ∗ , p∗ ) = · EUG (Left, p∗ ) + · EUG (Right, p∗ ) = 73.125
8 8
7 9
EUG (Center, p∗ ) = · 10 + · 5 = 7.1875
16 16
Similarly, the payoff from deviating to pure Center is less then the payoff from using
pure Left or Right or any other mixture over these three strategies. Therefore, the goalie
has no incentive to deviate from his mixed strategy q ∗ as well.

7 9
Therefore, ( 16 Left + 16 Right, 38 Left + 58 Right) is indeed a mixed-strategy Nash equilib-
rium of the original 3 × 3 game.

It is not easy to find mixed-strategy Nash equilibria in general cases. An entire subfield
of computer science, called algorithmic game theory, studies the problems of finding
various Nash equilibria, including mixed-strategy ones.

11