Problem set 3

Exercise 1
Game 1
Game 1 shows that a strategy can be strictly dominated by a mixed strategy. It
is clear that if player 1 employs the mixed strategy 0.5T + 0.5M , the expected
payoff is strictly higher than when player 1 plays B, irrespective of the strategy
player 2 chooses. Another approach to demonstrate that B is strictly dominated
is to prove that no belief held by player 1 about player 2’s strategies could make
B the best response. Let p denote the probability that player 2 plays L, and
1 − p the probability of playing R. The expected payoffs are:
Playing T : 3p,
Playing M : 3(1 − p),
Playing B : 1.
Comparing these, for p > 13 , player 1 prefers playing T over B, and for p < 23 ,
prefers playing M over B. Thus, there is no probability p for which B is the best
response, meaning B is strictly dominated. Subsequently, with B eliminated,
L is strictly dominated by R, and then T by M . Therefore set of rationalizible
actions is (M,R).

Game 2
Game 2 demonstrates a case where a pure strategy can be the best response to
a mixed strategy, even though it might not be the best response to any pure
strategies. In this game, B is not the best response to either L or R individually,
but it can be shown to be the best response to mixed strategies. Define p as
the probability of player 2 playing L, and 1 − p as the probability of playing R.
The expected payoffs are:
Playing T : 3p − 2,
Playing M : 1 − 3p,
Playing B : 0.
Player 1 prefers to play T over B if p > 23 and M over B if p < 31 . For any value
of p between 31 and 23 , B is the best response, hence B is not strictly dominated.
In this scenario, all actions are rationalizable.

1
Exercise 2
Game 1: This game demonstrates that applying the solution concept of iterated
elimination of weakly dominated strategies can eliminate a Nash equilibrium.
In this game, strategy L is weakly dominated by strategy R (since 1 > 0 and
2 = 2), and therefore it can be eliminated. After this elimination, player 1 will
prefer to play T over B, resulting in the outcome (T, R). However, this process
eliminates another Nash equilibrium, (B, L), which can be verified as a Nash
equilibrium since B is the best response to L and L is the best response to B.
Game 2: This game shows that the outcome of iterated elimination of
weakly dominated strategies is order-dependent. Both B and T are weakly
dominated by M . If we start by eliminating B, then for player 2, R becomes
weakly dominated by L, leading to outcomes (T, L) and (M, L). However, if we
begin by eliminating T , then for player 2, L would be weakly dominated by R,
resulting in outcomes (M, R) and (B, R). Thus, the final outcome significantly
depends on the order of elimination.

Exercise 3
Part 1
Denote by U1 (a1 , q) the expected utility of player 1 from playing pure action a1
when player 2 uses a mixed strategy that puts probability q on B and probability
1 − q on S. Simple calculations show:
U1 (B, q) = 2q, U1 (S, q) = 1 − q
Player 1 best responds by playing B if U1 (B, q) > U1 (S, q), by playing S if
U1 (B, q) < U1 (S, q) and is indifferent between her actions when U1 (B, q) =
U1 (S, q). Since we regard player 1 as choosing a distribution on {B, S}, her
best response is probability p she puts on B and probability 1 − p she puts on S.
Hence her best response is p that maximizes her expected utility and is given
by 
1
 if q > 31
br1 (q) = 0 if q < 13
[0, 1] if q = 31



Part 2
Similar argument as in the previous part using:
U2 (B, p) = p, U2 (S, p) = 2(1 − p)

1
 if p > 32
br2 (p) = 0 if p < 23
[0, 1] if p = 23



The two best response correspondences are drawn below.

2
Part 3
From the figures above, there exist three Nash Equilibria (NE). Denote a strat-
egy profile by (p, q). Then the three NE are:
 
2 1
(0, 0), (1, 1), , .
3 3

Exercise 4
Part 1
Similar arguments as in the previous question give:

U1 (H, q) = q − (1 − q),

U1 (T, q) = −q + (1 − q).
Thus, the best response function is:

1
 if q > 21 ,
br1 (q) = 0 if q < 12 ,
[0, 1] if q = 12 .



Part 2
A similar argument as in the previous part gives:

U2 (H, p) = −p + (1 − p),

U2 (T, p) = p − (1 − p).

3
Thus, the best response function is:

1
 if p < 21 ,
br2 (p) = 0 if p > 12 ,
[0, 1] if p = 12 .



Figure 1: Best response correspondences

Part 3
From the figures above, there exists one Nash Equilibria. Denote a strategy
profile by (p, q). Then the NE is: (0.5,0.5)

Exercise 5
Part (a)
To find the pure strategy NE of the game use usual method discussed in class
and you will see that there is no pure strategy NE

Part (b)
Consider that every player i’s strategy is given by

Si = (Ri , Pi , 1 − Ri − Pi )

where Ri denotes the probability that he plays Rock, Pi represents the probabil-
ity he plays Paper, and the remaining probability 1 − Ri − Pi is the probability
he plays Scissors. Thus, player 1’s strategy is denoted by S1 and player 2’s
strategy is denoted by S2 .

4
 We now start by making player 1 indifferent between using Rock, Paper,
and Scissors. Thus, we need

u1 (Rock, S2 ) = u1 (Paper, S2 ) = u1 (Scissors, S2 ).

Let us separately evaluate the expected payoff that player 1 obtains from
each of his pure strategies (Rock, Paper, and Scissors) before setting these three
expected payoffs equal to each other to make him indifferent.

First, if Player 1 chooses Rock (top row of the matrix), he obtains an ex-
pected payoff of

u1 (Rock, S2 ) = 0R2 + (−1)P2 + 1(1 − R2 − P2 )

= 1 − R2 − 2P2 .

Second, if Player 1 chooses Paper (middle row of the matrix), he obtains an

expected payoff of

u1 (Paper, S2 ) = 1R2 + 0P2 + (−1)(1 − R2 − P2 )

= 2R2 − 1 + P2 .

Third, if Player 1 chooses Scissors (bottom row), he obtains the following

expected payoff:

u1 (Scissors, S2 ) = (−1)R2 + 1P2 + 0(1 − R2 − P2 )

= −R2 + P2 .

We can now equate our three expected payoffs found above, u1 (Rock, S2 ),
u1 (Paper, S2 ), and u1 (Scissors, S2 ).
First, equating u1 (Paper, S2 ) and u1 (Scissors, S2 ), we get

2R2 − 1 + P2 = −R2 + P2 ,

which yields
1
R2 = .
3

Similarly, equating u1 (Rock, S2 ) and u1 (Scissors, S2 ), we get

1 − R2 − 2P2 = −R2 + P2 ,

which yields
1
P2 = .
3

5
 Finally, since 1 − R2 − P2 = 1 − 13 − 13 = 31 , the strategy that player 2 chooses
in the mixed-strategy Nash equilibrium (msNE) of the game is
 
1 1 1
S2∗ = , , ,
3 3 3

which indicates that he puts the same probability weight on each column.

By symmetry, the analysis of player 1 yields the same result so that

 
∗ 1 1 1
S1 = , , .
3 3 3

Intuitively, player 1 assigns the same probability weight to each of his rows.

Exercise 6
Notice that s0 is strictly dominated by s5 for player 2. Without s0 , w1 is strictly
dominated by w0 for player 1. Without {s0 , w1 }, s2 is strictly dominated by
s1 for player 2. Without {s0 , w1 , s2 }, w3 is strictly dominated by w2 for player
1. Without {s0 , w1 , s2 , w3 }, s4 is strictly dominated by s3 for player 2. Hence
{s0 , w1 , s2 , w3 , s4 } do not survive the iterated deletion of strictly dominated
actions. Omitting the deleted actions gives the game below.

s1 s3 s5
w0 4, 14 4, 12 4, 10
w2 12, 4 2, 12 2, 10
w4 10, 4 10, 2 0, 10

It is easy to check that this game has no pure strategy NE. For each a ∈
{w0 , w2 , w4 , s1 , s3 , s5 } denote the probability of that action being played by pa .
We have

U1 (w0 , ps1 , ps3 ) = 4ps1 + 4ps3 + 4(1 − ps1 − ps3 ) = 4

U1 (w2 , ps1 , ps3 ) = 12ps1 + 2ps3 + 2(1 − ps1 − ps3 ) = 2 + 10ps1

U1 (w4 , ps1 , ps3 ) = 10ps1 + 10ps3 + 0(1 − ps1 − ps3 ) = 10ps1 + 10ps3

U2 (s1 , pw0 , pw2 ) = 14pw0 + 4pw2 + 4(1 − pw0 − pw2 ) = 4 + 10pw0

6
 U2 (s3 , pw0 , pw2 ) = 12pw0 + 12pw2 + 2(1 − pw0 − pw2 ) = 2 + 10pw0 + 10pw2

U2 (s5 , pw0 , pw2 ) = 10

This leaves the fully mixed NE as the only possibility. We need to solve the
following system:

4 = 2 + 10ps1 = 10ps1 + 10ps3

4 + 10pw0 = 2 + 10pw0 + 10pw2 = 10.

Clearly, p∗s1 = 15 , p∗s3 = 1
5 and hence p∗s5 = 35 . Similarly, p∗w0 = 35 , p∗w2 = 1
5
and hence p∗w4 = 15 .