MATH 4321 – Game Theory

Solution to Homework One

Course Instructor: Prof. Y.K. Kwok

1. (a) With 4 pennies in single pile, the game tree is depicted as follows:

(b) Strategies for Player I (2 choices of strategy in later move when Player I plays 3 and
no choice of strategy in later move when Player I plays 2)
(1) play 3 then if at 2 play 1
(2) play 3 then if at 2 play 0
(3) play 2
Strategies for Player II (2 choices of strategy when Player I plays 3 and 2 choices of
strategy when Player I plays 2)
(1) if at 3→2 if at 2 → 1
(2) if at 3→2 if at 2 → 0
(3) if at 3→1 if at 2 → 1
(4) if at 3→1 if at 2 → 0
HH
HH II 1 2 3 4
I HH
1 1 1 -1 -1 min = −1
2 -1 -1 -1 -1 min = −1
3 -1 1 -1 1 min = −1
max = 1 max = 1 max = −1 max = 1

(c) We obtain v + = −1 and v − = −1, so the value of the game is −1. Player II always wins
since Strategy 3 is a dominant strategy for player II. This is obvious since one penny is
always left behind for Player I under strategy 3 of Player II. Everyone would prefer to be
the player who makes the second move (Player 2).

1
2. (a) The game matrix is given by
H
HH II
HH 1 2 3 4 5
I H
1 0 2 -1 -1 -1 min = −1
2 -2 0 2 -1 -1 min = −2
3 1 -2 0 2 -1 min = −2
4 1 1 -2 0 2 min = −2
5 1 1 1 -2 0 min = −2
max = 1 max = 2 max = 2 max = 2 max = 2
(b) We obtain v + = 1 and v − = −1. Since v + ̸= v − , so there exists no saddle point in
pure strategies.

3. The game tree of the Russian roulette is depicted as follows:

Strategies for player II:

II1 If I spins, then pass; if I passes, then spin
II2 If I spins, then pass; if I passes, then pass
II3 If I spins, then spin; if I passes, then spin
II4 If I spins, then spin; if I passes, then pass

The game matrix of the Russian roulette is given by

HH II
HH II1 II2 II3 II4
I H 2 1 5 1
6 (−1) + 6 (1)[ 6 (1) − 36
1
I1 1
6 (−1) + 56 (1) = 2 3 min = − 36
1
3 (same as I1, II1) + 6 (0)] = − 36
5 1
(same as I1, II3)
−2 3
0
I2 5
6 (−2) + 16 (1) = − 32 0 min = − 32
(same as I2, II1) (same as I2, II2)
max = 2
3 max = 2
3 max = − 361
max = 0

Note that v + = v − = − 36 1
. The saddle point under pure strategies is (I1, II3); that is,
both players choose to spin. Note that even though the payoff under (I1, II4) is the same
as v + = v − = − 36
1
, it is not a maximum using the column strategy II4.

2
4. Consider the following 2 separate cases:

(i) If 0 < x ≤ 3, then

( )
3 6 min = 3
A= min = 0.
x 0
max = 3 max = 6

We have v + = v − = 3, so the game has (1, 1) as the saddle point in pure strategies.
(ii) If x > 3, then
( )
3 6 min = 3
A= min = 0.
x 0
max = x max = 6

We have v + = min(x, 6) > 3 = v − , so the game has no saddle point in pure strategies.

5. (a) Note that

( ) ( )
v − (A) = max 1, min(z, 0) = 1 and v + (A) = min 1, max(2, z) = 1,

so the saddle point in pure strategies exists and it is at row 2 and column 1. Also,
v(A) = 1.
On the other hand, we have
( ) ( )
v − (B) = max 1, min(z, 0) = 1 and v + (B) = min 1, max(2, z) = 1,

so the saddle point in pure strategies exists and it is at row 1 and column 2. Also,
v(B) = 1.
(b) (i) Suppose we pick z = 3 so that
( )
2 4
A+B = .
4 2

Note that E(X, 1) = 2x1 + 4x2 and E(X, 2) = 4x1 + 2x2 . Setting E(X, 1) =
E(X, 2) and imposing x1 + x2 = 1, we obtain
1 1
X∗ = ( ).
2 2
In a similar manner, we obtain Y ∗ = ( 12 12 ). Also, we observe
( ) (1)
1 1 2 4 2
v(A + B) = ( ) 1 = 3 > v(A) + v(B).
2 2 4 2 2

(ii) Next, we pick z = −1 so that

( )
2 0
A+B = .
0 2

We obtain X ∗ = Y ∗ = ( 12 12 ) and v(A + B) = 1 < v(A) + v(B).

3
 ( )
a11 a12
6. Analytic method: For game matrix A = , we have X ∗ = (x 1 − x), where
a21 a22
22 −a21 −a12
x∗ = a11 −aa12 −a21 +a22
, Y ∗ = (y 1 − y), where y = a11 −aa22
12 −a21 +a22
; v = 1
(1) =
(1 1)A−1
1
 detA   .
a22 −a12 1
(1 1)
−a21 a11 1
We obtain X ∗ = ( 15 , 7 ), Y ∗ = ( 22
22 22
, 22 ), v = − 22
9 13 3
.

7. There is a pure strategy saddle point at (1, 3). This is because (1, 3) entry corresponds to
a minima of the 1st row and maxima of the 3th column. This gives the value of the game,
1 1
v = 3. One can verify that v = −1
=  13   =
Jn A Jn T
18
1
2
− 19
18
1
(1 1 1)  1
0 −3 1 
1 
3
− 18 − 2
17 1 29
18
1
1
1 = 3. To search for saddle point in mixed strategies, we use the formulas:
3
 
13 1
− 19
18 2 18 1 2
X∗ = vJ3 A−1 = 3(1 1 1)  13 0 − 13  = ( 0 )
3 3
− 18
17
− 12 29
 13    1
18

18
1
2
− 19
18
1 2
Y ∗T = vA−1 J3T = 3  13 0 − 13  1 =  0  .
− 17
18
− 12 29
18
1 1
2

Remark
A smart student may doubt about the applicability of the computational formulas:

X ∗ = vJ3 A−1 and Y ∗ = vA−1 J3T ,

when the saddle point strategies X ∗ and Y ∗ are not completely mixed (not all components
of X ∗ and Y ∗ are strictly positive). In this numerical example, we observe
 1 
2
E(2, Y ∗ ) = (4 − 3 2)  0  = 3 = v;
1
 2

( ) 5
1 2  −3  = 5 + 4 = 3 = v.
E(X ∗ , 2) = 0
3 3 3 3
2

In general, we have E(2, Y ∗ ) ≤ v and E(X ∗ , 2) ≥ v. In this example, it happens to have

strict equality in the above two cases, though x∗2 = 0 and y2∗ = 0.
The computational formulas manage to give the saddle point in mixed strategies (though
not completely mixed). On the other hand, the saddle point in pure strategies cannot be
obtained using the computational formulas.
In summary, xi > 0 and yi > 0 for all i and j are the sufficient conditions for guaranteed
success of finding X ∗ and Y ∗ using the formulas. When xi = 0, it is still possible to have
E(i, Y ∗ ) = v; similarly for yj = 0, we have E(X ∗ , j) = v.

4
 Alternatively, suppose we apply the two systems of equations to solve for X ∗ , Y ∗ and v,
where 

 v = E(1, Y ) = 3y1 + 5y2 + 3y3

v = E(2, Y ) = 4y1 − 3y2 + 2y3
;

 v = E(3, Y ) = 3y1 + 2y2 + 3y3

y1 + y2 + y3 = 1
and 

 v = E(X, 1) = 3x1 + 4x2 + 3x3

v = E(X, 2) = 5x1 − 3x2 + 2x3
.

 v = E(X, 3) = 3x1 + 2x2 + 3x3

x1 + x2 + x3 = 1
( ) ( )
∗ 1 2 ∗ 1 1
We obtain X = 0 , Y = 0 and v = 3. The two systems of
3 3 2 2
equations give X ∗ , Y ∗ and v as well since this non-zero game matrix happens to have
E(2, Y ∗ ) = v and E(X ∗ , 2) = v.

8. First, we compute

E(1, Y ) = 5(1 − y), E(2, Y ) = 4 − 3y, E(3, Y ) = 3y, E(4, Y ) = 2.

By plotting the lines E(1, Y ), E(2, Y ), E(3, Y ) and E(4, Y ) for Y = (y, 1−y), 0 ≤ y ≤ 1,
we observe that the lowest point of the upper envelope is given by the intersection of
E(2, Y ) and E(3, Y ). We skip the line E(1, Y ) for simplicity of the plots since it is
irrelevant in the subsequent calculation of Y ∗ . Also, the line E(4, Y ) also passes through
the lowest point of the upper envelope.

( L <

  (  <
 

(  < 
   


(  <
\
   

We solve for Y ∗ = (y ∗ , 1 − y ∗ ) by setting E(2, Y ∗ ) = E(3, Y ∗ ). This leads to

y ∗ + 4(1 − y ∗ ) = 3y ∗ ,
( )
∗ 2 ∗ 2 1
giving y = . The mixed strategy of the column player is Y = , . The value of
3 3 3
the game is E(3, Y ∗ ) = 2.
( )
∗ 1
Also, note that E(4, Y ∗ ) = 2 while E(1, Y ) = 5 < 2. This indicates that row 1 is
3
never used in the mixed strategy of the row player.

5
 1 1
We observe that row 4 can be duplicated by the convex combination of × row 2 + ×
2 2
row 3. First, we consider X = (0, x, 1 − x, 0) by leaving row 4 for a while and set
E(X, 1) = E(X, 2). This leads to

x + 3(1 − x) = 4x
1
so that x = . Accordingly, we have E(X, 1) = E(X, 2) = 2 = value of the game. Since
2
1 1
row 4 = × row 2 + × row 3, the most general mixed strategies of the row player is
2 2
given by ( α α )
X ∗ = 0, , , 1 − α , where 0 ≤ α ≤ 1.
2 2
As a verification, we observe
 
0
( α α ) 1 
E(X ∗ , 1) = 0, , , 1 − α  
 3 =2
2 2
2

and  
5
( α α ) 4 
E(X ∗ , 2) = 0, , , 1 − α  
 0  = 2.
2 2
2
When α = 0, we have the pure strategy X ∗ = (0 0 0 1) for Player 1.

9. Let Aij denote the Attacker’s strategy of putting the ith division on the first road and the
j th division on the second road. Let Dij denote the Defender’s strategy of putting the ith
division on the first road and the j th division on the second road.

number of defending number of attacking

divisions divisions to win
0 1 or more
1 3 or more
2 –
3 –

It is inferior to put 2 or 4 attacking divisions in a road.

(a) The payoff matrix is given by

D30 D21 D12 D03

 
A40 −1 −1 1 1
A31  1 −1 1 1 
 
A22  1 −1 −1 1 .
 
A13  1 1 −1 1 
A04 1 1 −1 −1

6
 (b) The 1st and 3rd rows are dominated by the 2nd row; it is suboptimal to put 2 or 4
attacking divisions in a row.
In a similar manner, the 5th row is dominated by the 4th row.
For the column strategies, the 1st column is dominated by the 2nd column; the 4th
column is dominated by the 3rd column.
The reduced payoff matrix becomes

D21 D12
( )
A31 −1 1
.
A13 1 −1

This payoff matrix resembles the Odds and Evens game. The value of this symmetric
game is seen to be zero. The saddle point in mixed strategies is given by
( ) ( )
∗ 1 1 ∗ 1 1
X = 0, , 0, , 0 and Y = 0, , , 0 .
2 2 2 2

10. Consider the matrix game

 
a4 a5 a3
A =  a1 a6 a5  , a1 < a2 < a3 < a4 < a5 < a6 .
a2 a4 a3

We observe that:

column 2 is dominated by column 3

row 3 is dominated by row 1.

The reduced game matrix becomes

( )
a4 a3
.
a1 a5

There is no saddle point in pure strategies, so we seek saddle point in mixed strategies.
We compute

E(X, 1) = a4 x1 + a1 (1 − x1 ) = (a4 − a1 )x1 + a1

E(X, 2) = a3 x1 + a5 (1 − x1 ) = (a3 − a5 )x1 + a5 .

By solving E(X, 1) = E(X, 2), we obtain

a5 − a1
x1 = .
a4 + a5 − a1 − a3
The optimal mixed strategy of Player I is
( )
∗ a5 − a1 a4 − a3
X = , , 0 .
a4 + a5 − a1 − a3 a4 + a5 − a1 − a3
Similarly, we can find the optimal mixed strategy of Player II to be
( )
∗ a5 − a3 a4 − a1
Y = , 0, .
a4 + a5 − a1 − a3 a4 + a5 − a1 − a3

7
 The value of the game is given by
 a5 −a3 
a4 +a5 −a1 −a3 a4 a5 − a1 a3
v(A) = E(1, Y ∗ ) = (a4 a3 a3 )  0 = .
a4 −a1 a4 + a5 − a1 − a3
a4 +a5 −a1 −a3

11. Given the game matrix ( )

3 −2 4 7
A= ,
−2 8 4 0
it is easily seen that column 3 and column 4 are dominated by column 1, we delete these
two dominated strategies. The reduced game matrix becomes
( )
3 −2
.
−2 8
There is no saddle point in pure strategies. We seek for saddle point in mixed strategies.
(a) Consider
E(X, 1) = 3x1 − 2(1 − x1 ) = 5x1 − 2
E(X, 2) = −2x1 + 8(1 − x1 ) = −10x2 + 8.
2
By equating E(X, 1) = E(X, 2), we obtain x1 = . This gives
3
( ) ( )( )
∗ 2 1 ∗ 2 1 3 2 4
X = and v(A) = E(X , 1) = =2− = .
3 3 3 3 −2 3 3
In a similar manner, we consider
E(1, Y ) = 3y1 − 2(1 − y1 ) = 5y1 − 2
E(2, Y ) = −2y1 + 8(1 − y1 ) = −10y1 + 8.
( )
∗ 2 1
We obtain Y = , , 0, 0 .
3 3
( )
1 1 1 1
(b) Given the fixed mixed strategy of Player II Y = , , , , we have
4 2 8 8
[ ]
1 1 1 1
E(X, Y ) = 3 · + (−2) + 4 · + 7 · x1
4 2 8 8
[ ]
1 1 1 1
+ (−2) · + 8 · + 4 · + 0 · (1 − x1 )
4 2 8 8
23
= 4 − x1 .
8
{ }
23
Note that max 4 − x1 = 4 when x1 = 0. Hence, X = (0, 1) is the best
0≤x1 ≤1 8 ( )
1 1 1 1
response for player I to the strategy Y = , , , . Alternatively, we observe
4 2 8 8
 1   1 
4 4
( ) 1  4 ( ) 1 
E(1, Y ) = 3 −2 4 7 

2
1
=
 8 and E(2, Y ) = −2 8 4 0 

2
1
 = 4.

8 8
1 1
8 8

8
 Since E(2, Y ) > E(1, Y ), so the optimal strategy of Player I is to play the pure
strategy of row 2. This shows agreement with the earlier result.
(c) In a similar manner, it is easy to show that Y = (1, 0, 0, 0) is the best response
for Player II when X = (0, 1).

12. (a) The game matrix is seen to be

H T
( )
H 3 −2
.
T −2 1

There is no saddle point in pure strategies, so we seek for mixed strategy. Note that

E(X, 1) = 3x1 − 2(1 − x1 ) = 5x1 − 2

E(X, 2) = −2x1 + (1 − x1 ) = −3x1 + 1.
( )
∗ 3 5
By equating E(X, 1) = E(X, 2), we obtain X = . Also, the value of the
8 8
1
game is v(A) = E(X ∗ , 1) = − . To solve for Y ∗ , we consider
8
1
E(1, Y ) = v(A) ⇔ 3y1 − 2(1 − y1 ) = −
8
( )
1 15 3 ∗ 3 5
giving 5y1 = 2 − = or y1 = . Therefore, Y = .
8 8 8 8 8
(b) Note that
( ) [ ]
1 2 1 2 x1
E(X, Y ) = 3 · − 2 · x1 + (−2) + (1 − x1 ) = − .
3 3 3 3 3
{ x }
1
Also, max − = 0 when x1 = 0. The best response is the pure strategy
0≤x1 ≤1 3
X = (0 1) and E(X, Y ) = 0.

( )
∗ ∗ 3 1 3 9
13. Let X = (x1 x2 x3 x4 ). Given that Y = , 0, , and v = , we can check that
7 7 7 7
E(1, Y ∗ ) = E(2, Y ∗ ) = E(3, Y ∗ ) = E(4, Y ∗ ) = v are satisfied. Next, we determine X ∗ by
setting the following equations:
 
−1
 1
E(X ∗ , 1) = (x1 x2 x3 x4 )  2


2
9
= −x1 + x2 + 2x3 + 2x4 = v =
7
9
E(X ∗ , 3) = 3x1 + 3x4 =
7
9
E(X ∗ , 4) = 3x1 + 2x2 + x3 =
7
x1 + x2 + x3 + x4 = 1.

9
 Solving the 4 equations for the 4 unknowns, we obtain
1 13 3 29
X∗ = ( , , , ).
12 28 28 84
Interestingly, though y2 = 0, we still observe
 
( ) 0
1 13 3 29 1 9
E(X ∗ , 2) =   = = v.
12 28 28 84 −2 7
3

Remark
Though yj > 0 ⇒ E(X, j) = v or E(X, j) > v ⇒ yj = 0, it is still plausible to have
yj = 0 while E(X, j) = 0.

14. Consider

f (x1 , x2 , y1 , y2 ) = XAY T
= 4x1 y1 − 3x1 y2 − 2x1 (1 − y1 − y2 ) − 3x2 y1 + 4x2 y2
− 2x2 (1 − y1 − y2 ) + (1 − x1 − x2 )(1 − y1 − y2 ).

We derive the system of equations by equating the first order partial derivatives of
f (x1 , x2 , y1 , y2 ) to be zero:

∂f 3
= 7y1 − 3 = 0 ⇒ y1 =
∂x1 7
∂f 3
= 7y2 − 3 = 0 ⇒ y2 =
∂x2 7
∂f 1
= 7x1 − 1 = 0 ⇒ x1 =
∂y1 7
∂f 1
= 7x2 − 1 = 0 ⇒ x2 = .
∂y2 7
Hence, we obtain
1 1 5 3 3 1
X ∗ = ( , , ) and Y ∗ = ( , , ).
7 7 7 7 7 7
Also, value of the game = f ( 71 , 1 3 3
, , )
7 7 7
= 17 . As a check, we have

1
E(X ∗ , j) = , j = 1, 2, 3
7
1
E(i, Y ∗ ) = , i = 1, 2, 3.
7
Hence, the validity of these optimal strategies is verified.

15. The strategy can be represented by the ordered pair (a, b), where a is the number of
fingers to show, a = 1, 2; b is the number of fingers to guess. The game matrix can be
represented by

10
 HH
HH II (1,1) (1,2) (2,1) (2,2)
I HH
(1,1) 0 2 -3 0
(1,2) -2 0 0 3
(2,1) 3 0 0 -4
(2,2) 0 -3 4 0

The game matrix is skew symmetric, where A = −AT , and this is a symmetric game.
The value of the game is then equal to zero.
We solve for the mixed strategy saddle point of Player I by solving the following set of
inequalities:
2
E(X, 1) ≥ 0 ⇔ −2x2 + 3x3 ≥ 0 ⇔ x3 ≥ x2 (1)
3
3
E(X, 2) ≥ 0 ⇔ 2x1 − 3x4 ≥ 0 ⇔ x1 ≥ x4 (2)
2
3
E(X, 3) ≥ 0 ⇔ −3x1 + 4x4 ≥ 0 ⇔ x4 ≥ x1 (3)
4
4
E(X, 4) ≥ 0 ⇔ 3x2 − 4x3 ≥ 0 ⇔ x2 ≥ x3 . (4)
3
By inequalities (2) and (3), we obtain
3 3 3 9
x1 ≥ x4 ≥ · x1 = x1 ⇒ x1 = 0
2 2 4 8
3 3 3 9
x4 ≥ x1 ≥ · x4 = x4 ⇒ x4 = 0.
4 4 2 8
Given that x1 = 0 and x4 = 0, these dictate x2 + x3 = 1. Next, we consider E(X, 1) ≥ 0
2 4
and E(X, 4) ≥ 0, where x3 ≥ x2 and x2 ≥ x3 , we obtain
3 3
2 2
x3 ≥ (1 − x3 ) ⇔ x3 ≥ .
3 5
4 4 2
Similarly, we observe x2 ≥ (1 − x2 ) ⇔ x2 ≥ . On the other hand, since x3 ≥ so that
3 7 5
3 4
x2 ≤ , together with x2 ≥ , we have
5 7
3 4
≥ x2 ≥ .
5 7
Similarly, since x3 = 1 − x2 , so
3 2
≥ x3 ≥ .
7 5
We conclude that the optimal mixed strategy is given by
3 4
X ∗ = (0, α, 1 − α, 0), where ≥α≥ .
5 7
By symmetry, it is also optimal for Player II.
Suppose Player II plays columns 2 and 3, the ratio of loss to gain in row 1 is 3/2 while the
ratio of gain to loss in row 4 is 4/3. This may explain why Player I drops the combination
of row 1 and row 4.

11
16. (a) The submarine can hide in a pair of adjacent squares out of 9 squares. The bomber
can bomb any one of the nine squares. The payoff matrix can be represented as
follows:

```
``` Submarine
``` 12 23 36 69 98 87 74 41 25 65 85 45
Bomber ```
`` `
1 1 0 0 0 0 0 0 1 0 0 0 0
2 1 1 0 0 0 0 0 0 1 0 0 0
3 0 1 1 0 0 0 0 0 0 0 0 0
4 0 0 0 0 0 0 1 1 0 0 0 1
5 0 0 0 0 0 0 0 0 1 1 1 1
6 0 0 1 1 0 0 0 0 0 1 0 0
7 0 0 0 0 0 1 1 0 0 0 0 0
8 0 0 0 0 1 1 0 0 0 0 1 0
9 0 0 0 1 1 0 0 0 0 0 0 0
Using MATLAB to solve the corresponding linear programming problem, we can
obtain the following solution:
1 1 1 1
X ∗ = (0, , 0, , 0, , 0, , 0)
4 4 4 4
1 1 1 1 1 1 1 1
Y ∗ = ( , , , , , , , , 0, 0, 0, 0).
8 8 8 8 8 8 8 8

(b) Using symmetry, the submarine’s strategies can be reduced to

S1: hide in a pair of squares that include the center square
S2: hide in a pair of squares that does not include the center square
The bomber’s strategies can be reduced to
B1: fire at a corner
B2: fire at a square in the middle of each side
B3: fire at the center square
The payoff matrix is then reduced to
```
``` Submarine
``` S1 S2
Bomber ```
`` `
1
B1 0 4
1 1
B2 4 4
B3 1 0
1
As an illustration, we show the payoff equals under the profile (B1, S2). Suppose
4
one plays B1 against S2, the bomber will fire at one of the four corners while the
submarine will hide along the edges. This results in 25% chance of hitting the
submarine.
(i) Pure saddle point
The reduced game has a pure saddle point at (B2, S2) and value of the game
v(A) = 41 . The bomber fires at one of the four middle side squares with equal
probability. The submarine hides with equal probability in one of the eight
locations that does not include the center square.

12
 (ii) Mixed saddle point
Let X ∗ = (x1 , x2 , x3 ) and Y ∗ = (y1 , y2 ) be the probability vectors. Knowing
1
v(A) = , the set of algebraic inequalities and equation for finding X ∗ are given
4
by
1 1
E(X ∗ , 1) = x2 + x3 ≥
4 4
1 1 1
E(X ∗ , 2) = x1 + x2 ≥
4 4 4
x1 + x2 + x3 = 1.

The solution is seen to be X ∗ = (0, 1, 0). On the other hand, the set of algebraic
inequalities and equations for finding Y ∗ are given by
1 1
E(1, Y ∗ ) = y2 ≤
4 4
1 1 1
E(2, Y ∗ ) = y1 + y2 ≤
4 4 4
1
E(3, Y ∗ ) = y1 ≤
4
y1 + y2 = 1.
1
The solution is Y ∗ = (α, 1 − α), 0 ≤ α ≤ .
4
Under mixed saddle point strategies, the bomber chooses to deliver the torpedo
in pure strategy at a square that is in the middle of a side (position 2, 4, 6 or
8). The submarine chooses mixed strategies. It hides in a pair of squares that
include the center square with probability α and in a pair of squares that do not
1
include the center square with probability 1 − α, where 0 ≤ α ≤ .
4

17. We take $1, 000 as one unit. This is a constant sum game of $4, 000 (half of the worth of
the car). We subtract $4, 000 from the payoffs of the two brothers and end up with the
following game matrix of a zero sum game that shows the payoff to Curly.
XXX
XXX Shemp
XXX 1 2 3 4 5 6 7 8
Curly XXX
1 0 -2 -1 0 1 2 3 4
2 2 0 -1 0 1 2 3 4
3 1 1 0∗∗ 0∗ 1 2 3 4
4 0 0 0∗ 0∗ 1 2 3 4
5 -1 -1 -1 -1 0 2 3 4

It is obvious that the saddle points are (3, 3), (3, 4), (4, 3) and (4, 4), all give the same
expected payoff of $4, 000 to each brother. Also, the value of the game is 0. Note that
strategy “3” dominates strategy “4” for both players. This refinement can be used to
argue that (3, 3) is the Nash equilibrium to be more likely to be played by both brothers.

13