OPTIMIZATION TECHNIQUES

BCA IV SEMESTER

NOTES FOR UNIT 5

Syllabus
Game Theory: Characteristics of games, maxima, minimax criteria of optimality, dominance

property, algebraic and graphical method of solution of solving 2 x 2 games.

GAME THEORY

A game with two players where a gain of one player is equal to the loss of other player, is known
as Two person Zero sum game. It is represented by rectangular Pay- off matrix. If the number
of players are more than 2 then it is known as n person game

Maximin rule As per this rule the player choose a strategy that results in the best out of worst
ie maximum out of minimum. Adopted by a player who is maximizer

Minimax rule According to this rule a player chooses minimum out of maximum. As
adopted by Minimizer

Value of the game: It is expected pay off of the game when all players use their best strategies.
If the sum of gains and losses is zero it also known as Fair game otherwise as an Unfair game.

Pure Strategy game where Saddle point exists

Determining saddle point

a. Select the minimum element of each row of matrix and mark it with *

b. Select the maximum element of each column and mark it with ○ or +

Algebraic method for 2by2 matrix

For a matrix like as under

Player
B
Player
A I II
I a11 a12
II a21 a22

If for player A probability for strategy I is p then it will be 1-p for II strategy. Same way if
player B has probability q for I strategy then it will be 1-qfor II strategy

So we have formula for calculations of p and q

As under

P= a22- a21/(a11+a22)-(a12+a21).

. q= a22- a12/(a11+a22)-(a12+a21).

Value of the game = a11*a22 –a12*a21/(a11+a22)-(a12+a21).

Now 1-p and 1-q can be calculated easily

ALTERNATIVE METHOD beside above quoted algebraic method we have another method
to calculate the value . for example:

Player
B
Player
A I II
I a11 a12
II a21 a22

If the probability for I strategy of A is p then the probability for II strategy of A will be (1-p).
similarly if probability for I strategy of player B is q then probability for II strategy of player B
will be (1-q)

Thus for p
 p * a11 + (1-p) * a21 = p * a12 + (1-p) * a22

and for q *a11 + (1-q) * a12 = q* a21 + (1-q) * a22

value of the game = p * a11 + (1-p) * a21 or p * a12 + (1-p) * a22

Principle of dominance this rule is useful in reducing the matrix in 2 by 2 form. There are two
rules of dominance

1. if all the elements of any row (say x row) ,of any pay off matrix are less than or equal to
the corresponding elements of other row (say y row) then player A will never choose x row
and thus can be deleted safely.

2. If all the elements of a column (say a) are greater or equal to the corresponding elements of
other column (say b) then Player B will never choose column a and thus column a can be
deleted safely from the matrix

3. If neither rows nor columns have any chance to remove any row or column then we take
average of two rows and then look for dominance over other rows and if it is there then the
dominated row is removed. Same is applied for column also

4. After application of Law of Dominance if saddle point is available then it is pseudo saddle
point and it should be ignored and the problem should be solved by usual method.

Graphical method Useful for 2 by n or m by 2 matrix where n is number of options for B

Two vertical line apart from each others are used to represent the two strategies of any player A
or B . One horizontal line is drawn across these two vertical lines almost in the middle and this
line acts as a zero mark. Bothe vertical lines are marked with same scale. Below the horizontal
line it is negative and the above it, it is positive value. Now for values pairs for various options
available to other player are found on two vertical lines and are met with a line for each pair.
After all line are drawn then, if it is player A who has two options and shown on two vertical
lines then the area under the lower envelop is considered and highest point of intersection is
marked. The graph lines making the intersection are identified and their value make the 2x2
matrix that gives us the result by using usual method. maximum values is picked (maximin) and
if it is for Player B then the area above the graph i.e. upper envelop, is considered and minimum
of the value is chosen and the corresponding values of graph is taken for 2x2 matrix. (Minimax)
Questions for the practice

B
B1 B2 B3 B4 B4
A1 1 3 2 7 4
A2 3 4 1 5 6
A A3 6 5 7 6 5
A4 2 0 6 3 1

B
B1 B2 B3
A1 -5 10 20
A A2 5 -10 -10
A3 5 -20 -20

B
B1 B2 B3 B4 B4
A1 -5 5 0 1 8
A A2 8 -4 -1 6 -5

B
B1 B2 B3 B4
A1 3 3 4 0
A A2 5 4 3 7

5
 B
1 2 3
a -6 10 11
A b -1 -2 -3
c -1 -2 -4

B
1 2
a 6 5
A b 3 6
c 8 4
d 7 1
7

B
1 2 3
a 1 7 2
A b 6 2 7
c 6 1 6

B
B1 B2 B3 B4
a -3 4 2 9
A b 7 8 6 10
c 6 2 4 -1

B
B1 B2 B3 B4
a 1 7 3 4
A b 5 6 4 5
c 7 2 0 3

10

B
 B1 B2 B3
a 60 50 40
A b 70 70 50
c 80 60 75

11

B
B1 B2 B3 B4
a 1 2 -1 2
A b 3 1 2 3
c -1 3 2 1
d -2 2 0 -3

12

B
B1 B2 B3 B4
a 1 7 2 4
A b 0 3 7 8
c 5 2 6 10

13

B
B1 B2 B3 B4 B4
a 3 0 6 -1 7
A b -1 5 -2 2 1

14

B
1 2 3 4
a 6 4 8 0
A b 6 8 4 8
c 8 4 8 0
d 0 8 0 16

15 16
 B B
1 2 3 b1 b2 b3
a 4 6 4 A a1 2 -1 3
A b 2 10 0 a2 4 5 6

17
B
1 2 3
A A1 1 3 11
A2 8 5 2