Game Theory, Fall 2022

Problem Set 2
Zhuoyuan He and Sihao Tang∗

1. ST Exercise 4.6.

(a) Soln: Given player j’s strategy tj , player i solves

max (10 − tj )ti − t2i

ti ≥0

When tj ≥ 10, it is obvious that i’s best response is ti = 0. When tj < 10, the first
10−tj
order condition is 10 − tj − 2ti = 0, implying that the best response is ti = 2
.
In sum, i’s best response is
 
10 − tj
ti = max ,0 .
2

(b) Soln: Clearly, any ti ∈ [0, 5] is not strictly dominated because it is a best response
to tj = 10 − 2ti . Moreover, consider any strategy ti of the form ti = 5 + k for
some k > 0. We have, for all tj ≥ 0,

vi (5 + k, tj ) = (10 − tj )(5 + k) − (5 + k)2 = 25 − 5tj − k 2 − tj k < 25 − 5tj = vi (5, tj ).

Therefore, any strategy ti > 5 is strictly dominated by ti = 5. Thus all strategies

in (5, +∞) should be deleted in the first round. We have S11 = S21 = [0, 5].
(c) Soln: In the second round of elimination, any strategy ti ∈ [ 52 , 5] should be kept
because it is a best response to tj = 10 − 2ti ∈ [0, 5]. Moreover, any strategy
5
ti < 2
is strictly dominated by ti = 52 . To see this, consider a strategy of the form
5
ti = 2
− k for some k > 0. Then

5 75 5
vi ( − k, tj ) = − tj − (5 − tj )k − k 2 .
2 4 2
∗
Special thanks go to Peixuan Fu and Shuang Wu, who wrote the last version of these solutions.

1
 When tj ∈ [0, 5], we have
5 75 5 5
vi ( − k, tj ) < − tj = vi ( , tj ).
2 4 2 2
5
Thus, ti = 2
− k is strictly dominated by ti = 25 . Therefore, S12 = S22 = [ 52 , 5].
We can continue the above arguments in a similar fashion. Define function ϕ(x) =
(10−x)/2. In fact, we can show after the kth round of deletion, S1k = S2k = [q k , q k ]
where
q k = ϕ(k−1) (0) and q k = ϕ(k) (0) if k is odd,

q k = ϕ(k) (0) and q (k) = ϕ(k−1) (0) if k is even.

and ϕ(k) is the kth composition of ϕ itself. Because limk→∞ ϕk (0) = 10 3

, we know
that Sj∗ = Si∗ = k≥1 Sik = { 10
T
3
}. Hence, the unique pair of strategies that survive
10
IESDS for this game are t1 = t2 = 3
.

2. ST Exercise 4.7.

(a) Soln: The set of players is {1, 2}. The strategy spaces of players are S1 =
{P, B, N } and S2 = {P, B, N }. The payoffs to player 1 are

v1 (P, P ) = 0.5, v1 (B, B) = 0.5, v1 (N, N ) = 0.5,

v1 (P, B) = 0, v1 (P, N ) = 0.3, v1 (N, B) = 0.6,
v1 (B, P ) = 1, v1 (N, P ) = 0.7, v1 (B, N ) = 0.4.

And, v2 (s) = 1 − v1 (s) for all s ∈ {P, B, N } × {P, B, N }.

(b) Soln: The matrix is in Figure 1.

Player 2
P B N
P 0.5, 0.5 0, 1 0.3, 0.7
Player 1 B 1, 0 0.5, 0.5 0.4, 0.6
N 0.7, 0.3 0.6, 0.4 0.5, 0.5

Figure 1: The normal form game for Question 2

(c) Soln: We delete all strictly dominated strategies in each round. In the first round,
P for both players is strictly dominated by B. In the second round, B for both
players is strictly dominated by N . Hence {N } × {N } survives iterated deletion
of strictly dominated strategies.

2
3. ST Exercise 4.8.

(a) Soln: Consider that player i chooses either 19 or 18. In both cases the average is
3
less than 20 regardless of the number of players, and 4
of the average is less than
15. This means that either 19 or 18 makes i the only winner. Therefore, both 19
and 18 are best responses.
(b) Soln: Obviously x = 20 is not one of the best responses because there will be
20(n−1)+x
many winners. Consider any x < 20. Because the average is n
, player i is
the only winner if and only if
3 20(n − 1) + x 3 20(n − 1) + x
20 − · > · − x,
4 n 4 n
or equivalently
30
x > 10 − .
2n − 3
Therefore the set of best-responses is
 
30
x ∈ Z 0 ≤ x < 20, x > 10 − ,
2n − 3
which clearly depends on the number of players n. For instance, if n = 2, this set
is {0, 1, . . . , 19}. If n is sufficiently large, this set is {10, 11, . . . , 19}.

4. Consider the following two-player game. Each player announces a nonnegative real
number. The payoffs are

2, if xi = 0, xj = 1,
vi (xi , xj ) =
arctan x , if otherwise.
i

(a) Argue that every positive announcement is strictly dominated.

Soln: Suppose xi > 0. Then

vi (xi , xj ) = arctan xi < arctan(xi + 1) = vi (xi + 1, xj ), ∀xj .

Therefore, xi is strictly dominated.

(b) Argue that announcement 0 is not strictly dominated.
Soln: Because

vi (0, 1) = 2 > arctan xi = vi (xi , 1), ∀xi > 0,

we know 0 for player i is a best response to xj = 1. Therefore, it is not strictly

dominated.

3
 (c) From the above two questions, we know only 0 survives IESDS for both players.
Are they mutual best responses?
Soln: Obviously not. This is because

vi (0, 0) = arctan 0 < arctan 1 = vi (1, 0).

5. Consider the n-firm Cournot competition. The demand curve is still

D(Q) = max{100 − Q, 0}.

Pn
If each firm i supplies qi , the total supply is i=1 qi . Suppose each firm’s marginal
cost is 10.

(a) Write down its normal form game.

Soln: There are n firms, N = {1, 2, . . . , n}. Firm i can choose any nonnegative
quantities, so Si = R+ . Given the strategy profile (q1 , . . . , qn ), the payoff to firm
i is
 X 
vi (q1 , . . . , qn ) = max{100 − qj − qi , 0} − 10 qi .
j̸=i

(b) For each firm, what are the strategies that survive IESDS?
Soln: We only consider n ≥ 3. As in the duopoly case, we can write firm i’s
payoff function as
X
vi (qi , q−i ) = max{(90 − qj − qi )qi , −10qi }.
j̸=i

Let Si0 = R+ for all i. It is easy to see that given the opponents’ strategy profile
q−i ∈ Rn−1
+ , firm i’s best response is
P
90 − j̸=i qj
 
qi = max ,0 .
2
P
Thus, as j̸=i qj increases over the range [0, +∞), firm i’s best response decreases
from 45 to 0. We immediately know that any qi ∈ [0, 45] is not strictly dominated.
Moreover, as in the duopoly case, it is also easy to see that vi ( · , q−i ) is strictly
decreasing over the interval [45, +∞) for any q−i ∈ Rn−1
+ . This in turn implies
that any quantity qi > 45 is strictly dominated by 45. Therefore, in the first
round, we delete all quantities above 45, and Si1 = [0, 45] for all i.
1
P
We move to the second round. Notice j̸=i Sj = [0, 45(n − 1)] ⊃ [0, 90] for
P
n ≥ 3. Moreover, as j̸=i qj increases over the range [0, 90], firm i’s best response

4
 decreases from 45 to 0. This implies that no quantity qi ∈ [0, 45] = Si1 is strictly
dominated. Therefore, the process of IESDS must stop here. For each i, Si∗ =
[0, 45].

6. ST Exercise 5.5.

(a) Soln: Let si be the value that player i contributes, which can only be 0 or 1.
P
Denote the value that the other two players contribute as s−i . If j̸=i sj = 0, the
streetlamp would not be erected regardless of whether player i contributes or not.
P
Hence it is optimal for player i not to contribute. If j̸=i sj = 1, player i could
obtain 3−1 = 2 from the streetlamp if he contributes and 0 if he does not. Hence,
P
it is optimal for player i to contribute. Finally, if j̸=i sj = 2, the streetlamp
would be erected regardless of whether player i contributes or not. Hence it is
optimal for player i not to contribute. As a result, the best response of player i is

P


 0, if j̸=i sj = 0,

BRi (s−i ) = 1, if
P
 j̸=i sj = 1,

 P
0, if
j̸=i sj = 2.

(b) Soln: When no player contributes, none of them could change the result by
deviating to contributing. Hence no one would deviate, and (0, 0, 0) is a Nash
equilibrium.
When there is only one contributing player, the streetlamp would not be erected.
Therefore, the player who contributes would deviate from contributing to not
contributing. This means that any strategy profile in which there is only one
contributor is not a Nash equilibrium.
When there are two contributing players, the streetlamp would be erected. For
the contributors, they would not deviate since the deviation will make them lose
the streetlamp. The player who does not contribute would not deviate either since
the streetlamp has been ensured. Hence any strategy profile in which there are
two contributors is a Nash equilibrium.
When all the players contribute, every player would like to deviate because the
deviation could save money without losing the streetlamp. Hence it is not a Nash
equilibrium.
In sum, (0, 0, 0), (0, 1, 1), (1, 0, 1), (1, 1, 0) are the Nash equilibria.