Lecture_22_Game Theory :Simplex Method

22.1 Simplex Method
Let us consider the 3 x 3 matrix
As per the assumptions, A always attempts to choose the set of strategies with the non-zero
probabilities say p1, p2, p3 where p1 + p2 + p3 = 1 that maximizes his minimum expected gain.
Similarly B would choose the set of strategies with the non-zero probabilities say q1, q2, q3 where
q1 + q2 + q3 = 1 that minimizes his maximum expected loss.
Step 1
Find the minimax and maximin value from the given matrix
Step 2
The objective of A is to maximize the value, which is equivalent to minimizing the value 1/V.
The LPP is written as
Min 1/V = p1/V + p2/V + p3/V
and constraints ? 1
It is written as
Min 1/V = x1 + x2 + x3
and constraints ? 1
Similarly for B, we get the LPP as the dual of the above LPP
Max 1/V = Y1 + Y2 + Y3
and constraints ? 1
Where Y1 = q1/V, Y2 = q2/V, Y3 = q3/V
Lecture 22
Game Theory :
Simplex Method
Step 3
Solve the LPP by using simplex table and obtain the best strategy for the players
1
Example 1
Solve by Simplex method
Solution
We can infer that 2 ? V ? 3. Hence it can be concluded that the value of the game lies between 2
and 3 and the V > 0.
LPP
Max 1/V = Y1 + Y2 + Y3
Subject to
3Y1 – 2Y2 + 4Y3 ? 1
-1Y1 + 4Y2 + 2Y3 ? 1
2Y1 + 2Y2 + 6Y3 ? 1
Y1, Y2, Y3 ? 0
SLPP
Max 1/V = Y1 + Y2 + Y3 + 0s1 + 0s2 + 0s3
Subject to
3Y1 – 2Y2 + 4Y3 + s1 = 1
-1Y1 + 4Y2 + 2Y3 + s2 =1
2Y1 + 2Y2 + 6Y3 + s3 = 1
Y1, Y2, Y3, s1, s2, s3 ? 0
2
Cj? 1 1 1 0 0 0
Basic
Variables CB YB Y1 Y2 Y3 S1 S2 S3 Min Ratio
YB / YK
S1 0 1 3 -2 4 1 0 0 1/3?
S2 0 1 -1 4 2 0 1 0 -
S3 0 1 2 2 6 0 0 1 1/2
1/V = 0
?
-1
-1
-1
0
0
0
Y1 1 1/3 1 -2/3 4/3 1/3 0 0 -
S2 0 4/3 0 10/3 10/3 1/3 1 0 2/5
S3 0 1/3 0 10/3 10/3 -2/3 0 1 1/10?
1/V =1/3
0
?
-5/3
1/3
1/3
0
0
Y1 1 2/5 1 0 2 1/5 0 1/5
S2 0 1 0 0 0 1 1 -1
Y2 1 1/10 0 1 1 -1/5 0 3/10
1/V = 1/2
0
0
2
0
0
1/2
1/V =1/2
V = 2
y1 = 2/5 * 2 = 4/5
y2 = 1/10 * 2 = 1/5
y3 = 0 * 2 = 0
x1 = 0*2 = 0
x2 = 0*2 = 0
x3 = 1/2*2 = 1
SA = (0, 0, 1)
SB = (4/5, 1/5, 0)
Value = 2
Example 2
3
Solution
Maximin = -1
Minimax = 1
We can infer that -1 ? V ? 1
Since maximin value is -1, it is possible that value of the game may be negative or zero, thus the
constant ‘C’ is added to all the elements of matrix which is at least equal to the negative of
maximin.
Let C = 1, add this value to all the elements of the matrix. The resultant matrix is
LPP
Max 1/V = Y1 + Y2 + Y3
Subject to
2Y1 + 0Y2 + 0Y3 ? 1
0Y1 + 0Y2 + 4Y3 ? 1
0Y1 + 3Y2 + 0Y3 ? 1
Y1, Y2, Y3 ? 0
SLPP
Max 1/V = Y1 + Y2 + Y3 + 0s1 + 0s2 + 0s3
Subject to
2Y1 + 0Y2 + 0Y3 + s1 = 1
0Y1 + 0Y2 + 4Y3 + s2 = 1
0Y1 + 3Y2 + 0Y3 + s3 = 1
Y1, Y2, Y3, s1, s2, s3 ? 0