Professional Documents
Culture Documents
Notes Game Theory
Notes Game Theory
Notes Game Theory
➢ Dominance Properties:
1) In the payoff matrix of player A, if all the entries in a row X (column X) are greater than
or equal to (less than or equal to) the corresponding entries of another row Y (column Y)
then row Y (column Y) is dominated by row X (column X). Under such situation row Y
(column Y) of the payoff matrix can be deleted.
2) In the payoff matrix of player A, if sum of each of the entries of two rows X and Y
(column X and Y) is greater than or equal to (less than or equal to) to the corresponding
entries of a third row Z (column Z) then row Z (column Z) is dominated by row X and
row Y (column X and column Y). Under such situation row Z (column Z) of the payoff
matrix can be deleted.
Player B
a11 a12
Player A
a21 a22
Then the following formulae are used to find the optimal strategies and value of the game:
p1 =
( a22 − a21 ) , Where p2 = 1 − p1
( a11 + a22 ) − ( a12 + a21 )
Optimal strategies for Player B
=
( a11.a22 − a12 .a21 )
( a11 + a22 ) − ( a12 + a21 )
➢ Solution of m x n Games as an L.P.P.:
Game problem can be formulated and solved as a linear programming problem. Suppose 3x3 pay-
off matrix is as follows.
Let us consider the problem first from A’s point of view. A is interested in maximizing the
minimum gain that can be obtained by playing this game. The objective, then, is to maximize the
minimum value of the game, U, subject to the constraints that regardless of the alternative that B
chooses, the expected pay-off shall be at least equal to U. Now, let x1, x2 and x3 be the
probabilities with which A plays the strategies A1, A2 and A3; E1 be the expected pay – off of A
when he plays the three strategies with these probabilities and B decides B1. Similarly, let E2 and
E3 be the expected pay – offs associated with the strategies B2 and B3 respectively, adopted by B.
and x1 + x2 + x3 = 1
1
Minimize = X1 + X 2 + X 3
U
Subject to
Looking from B’s point of view, it is easy to visualize that his problem to minimize, V, the
maximum loss that he is to incur by playing this game. B’s problem can be interpreted as the dual
to the linear programming model obtained above in respect of A. Also, we can derive it directly
from the information given. It follows,
Let y1, y2 and y3 be the probabilities with which B plays strategies B1, B2 and B3, respectively and
E1’, E2’ and E3’ be the expected pay – offs associated with the events that A chooses,
respectively, A1, A2 and A3 while B plays the strategies B1, B2 and B3 with the given probabilities.
Accordingly,
and
Now, the problem is to minimize V, subject to
and y1 + y2 + y3 = 1
If we define (assuming as before that V is positive) Yi = yi/V, and divide all constraints by V, we
can express the L.P.P. as follows:
1
Maximize = Y1 + Y2 + Y3
V
Subject to
Player A Player B
B1 B2 B3
A1 5 4 7
A2 5 8 4
A3 8 5 6
Solution:
CB B Cj 1 1 1 0 0 0 Mini.
b (= YB) Y1 Y2 Y3 S1 S2 S3 Ratio
YBi / Yij
(Where
Yij > 0)
0 S1 1 5 4 7 1 0 0 1/5
0 S2 1 5 8 4 0 1 0 1/5
0 S3 1 8 5 6 0 0 1 1/8
Zj 0 0 0 0 0 0
Cj - Zj 1 1 1 0 0 0
CB B Cj 1 1 1 0 0 0 Mini.
b (= YB) Y1 Y2 Y3 S1 S2 S3 Ratio
YBi / Yij
(Where
Yij > 0)
0 S1 3/8 0 7/8 13 / 4 1 0 -5/8 3/7
0 S2 3/8 0 39 / 8 1/4 0 1 -5/8 1 / 13
1 Y1 1/8 1 5/8 3/4 0 0 1/8 1/5
Zj 1 5/8 3/4 0 0 1/8
Cj - Zj 0 3/8 1/4 0 0 -1/8
Second Iteration
CB B Cj 1 1 1 0 0 0 Mini.
b (= YB) Y1 Y2 Y3 S1 S2 S3 Ratio
YBi / Yij
(Where
Yij > 0)
0 S1 4 / 13 0 0 125 / 39 1 - 7 / 39 - 20 / 39 12 / 125
1 Y2 1 / 13 0 1 2 / 39 0 8 / 39 - 5 / 39 3 /2
1 Y1 1 / 13 1 0 28 / 39 0 - 5 / 39 8 / 39 3 / 28
Zj 1 1 10 / 13 0 1 / 13 1 / 13
Cj - Zj 0 0 3 / 13 0 - 1 / 13 - 1 /13
Third Iteration
CB B Cj 1 1 1 0 0 0
b (= YB) Y1 Y2 Y3 S1 S2 S3
1 Y3 12 / 125 0 0 1 39 / 125 - 7 / 125 - 20 / 125
1 Y2 9 / 125 0 1 0 - 2 /125 26 / 125 - 15 /125
1 Y1 1 / 125 1 0 0 -28 / 125 - 11 / 125 40 / 125
Zj 1 1 1 9 / 125 8 / 125 5 / 125
Cj - Zj 0 0 0 - 9 /125 - 8 /125 - 5 /125
Here all Cj – Zj = Δij ≤ 0. So, the current table gives an optimal solution.
Optimal Solution is as follows:
2) The matrix below show the pay – off in a 3 X 3 game between two players X and Y. X is
maximizing player and Y the minimizing player.
a) Formulate a suitable Linear Programming model of the game with respect to the
minimizing player Y and derive the optimal strategy for Y.
b) Write down the dual of the above Linear Program and derive the solution to the dual
from the optimal tableau of the primal and interpret the meaning of this solution:
Y
1 -1 -1
X -1 -1 3
-1 4 -1
Solution:
a) Since some of the entries are negative, we add a constant 2 to each of the values
in the matrix.
Y
3 1 1
X 1 1 5
1 6 1
Next, let y1, y2 and y3 be the respective probabilities of player Y playing the
three strategies and V be the value of the game. We define Yi = yi / V. The LP
model is
CB B Cj 1 1 1 0 0 0 Mini.
b (= YB) Y1 Y2 Y3 S1 S2 S3 Ratio
YBi / Yij
(Where
Yij > 0)
0 S1 1 3 1 1 1 0 0 1/3
0 S2 1 1 1 5 0 1 0 1
0 S3 1 1 6 1 0 0 1 1
Zj 0 0 0 0 0 0
Cj - Zj 1 1 1 0 0 0
First Iteration
CB B Cj 1 1 1 0 0 0 Mini.
b (= YB) Y1 Y2 Y3 S1 S2 S3 Ratio
YBi / Yij
(Where
Yij > 0)
1 Y1 1/3 1 1/3 1/3 1/3 0 0 1
0 S2 2/3 0 2/3 14 / 3 -1/3 1 0 1
0 S3 2/3 0 17 / 3 2/3 -1/3 0 1 2 / 17
Zj 1 1/3 1/3 1/3 0 0
Cj - Zj 0 2/3 2/3 -1/3 0 0
Second Iteration
CB B Cj 1 1 1 0 0 0 Mini.
b (= YB) Y1 Y2 Y3 S1 S2 S3 Ratio
YBi / Yij
(Where
Yij > 0)
1 Y1 5 / 17 1 0 5 / 17 6 / 17 0 - 1 /17 1
0 S2 10 / 17 0 0 78 / 17 - 5 /17 1 - 2 / 17 5 / 39
1 Y2 2 / 17 0 1 2 / 17 - 1 / 17 1 3 / 17 1
Zj 1 1 7 / 17 5 / 17 1 2 / 17
Cj - Zj 0 0 10 / 17 - 5 / 17 -1 - 2 / 17
CB B Cj 1 1 1 0 0 0
b (= YB) Y1 Y2 Y3 S1 S2 S3
1 Y1 10 / 39 1 0 0 29 / 78 - 5 / 78 - 2 / 39
1 Y3 5 / 39 0 0 1 - 5 / 78 17 / 78 - 1 / 39
1 Y2 4 / 39 0 1 0 - 2 / 39 - 1 / 39 7 / 39
Zj 1 1 1 10 / 39 5 / 39 4 / 39
Cj - Zj 0 0 0 - 10 / 39 - 5 / 39 - 4 / 39
Here all Cj – Zj = Δij ≤ 0. So, the current table gives an optimal solution.
Optimal Solution is as follows:
b) Dual Problem:
Minimum 1 / U = X1 + X2 + X3
Subject to
3Y1 + Y2 + Y3 ≥ 1
Y1 + Y2 + 6Y3 ≥ 1
Y1 + 5Y2 + Y3 ≥ 1
Y1, Y2, Y3 ≥ 0