Operation Research

Chapters completed:
1. Introduction
2. Linear programming
3. Transportation problem
4. Assignment problem
5. Network analysis: project scheduling [Half]
6. Model formulation
7. Game theory
Chapters not yet completed:
8. Network analysis (Partial)
9. Queueing model
10. Decision theory
11. Dynamic programming
12. Simulation
GAME THEORY
SADDLE POINT:
A game for which

Max min a ij=Min max aij; i= row; j= column

is called a game with saddle point. This implies that a game with saddle point is that in which the
players use pure strategies i.e. they choose some course of action through out the game. Saddle
point is the intersection point of their pure strategies. The saddle point gives the value of the
game.
Value of the game, v= value of saddle point.
The corresponding strategy of the saddle point is the pure strategy for both the players.
Max min= Maximum of minimum
Min max= Minimum of maximum
RULES FOR DETECTING A SADDLE POINT:
Step-1: calculate the value of the row minimum for each of the row and write down the m values
in a column under the heading “Row minima”. Find the maximum value of the elements in that
column headed by “Row minima”. Let is be α and let it occur at k-th row.
Step-2: calculate the value of the column maxima for each of the column and write down the n
values in a row under the heading “Column maxima”. Find the minimum value of the elements in
that row headed by “Column Maxima”. Let is be β and let it occur at l-th column.

Step-3: in general α<β and if α=β, the value of the game is α(=β) and the optimal strategies are A k
and Bl for A and B respectively and the saddle point of the pay-off matrix is (k,l)-th position of
the matrix. If α<β then the game value lies between α and β i.e. if gave value is v then α ≤ v ≤ β
and the game has no saddle point.
 PAY-OFF FUNCTION:

Let [ a ij ] be any pay-off matrix of order mxn. Then the pay-off function or mathematical
expectation of a game which is denoted by E (p,q) and is defined by
m n
E (p,q)= ∑ ∑ aij pi q j
i=1 j=1

m
Where p and q are the mixed strategies for A and B player respectively and ∑ pi =1, and
i=1
n

∑ q j=1 and pi ≥0 and q j ≥ 0

j=1

In particular if B takes his pure j-th move only then the expected game of A is given by
m
E j ( p)=∑ a ij p i, j=1,2,3………,n
i=1

Similarly, for particular i-th pure move of A only the expected loss of B is given by
n
Ei (q)=∑ a ij q j, i=1,2,3………,m
j=1

SOLUTION OF TWO-PERSON ZERO-SUM GAME

1. Maximin-minimax method
2. Arithmetic method for 2x2 game of mixed strategies
3. Arithmetic (matrix) method for 3x3 game
4. Algebraic method for 2x2 game of mixed strategies
5. Graphical method for 2xn or mx2 games
6. Method of sub-games for 2xn or mx2 games
7. Algebraic (system of equations) of mxn games
8. Dominance method
9. Iterative method for approximate solution (Brown’s Algorithm)
10. Linear programming method
Method-1: Maximin-minimax method
It can be applied for any order of game.
Algorithm to detect a saddle point:
Step-1: Calculate the value of row minimum for each of the row and write down these values in a
column right side of the pay-off matrix under the heading “row minima”. Find the largest (maximum)
of them and enclosed by () parentheses. Let it be α and it occurs at k-th row.
Step-2: Calculate the value of column maximum for each of the column and write down these values
in a row below the pay-off matrix under the heading “column maxima”. Find the lowest (minimum)
of them and enclosed by () parentheses. Let it be β and it occurs at l-th column.
Step-3: If these two elements are same i.e. α= β, the cell where the corresponding row and column
meet (k,l) is a saddle point and the element in that cell i.e. α(= β) is the value of the game.
Gaem value, v= α (= β)
Saddle point is (k,l)
A’s pure strategy is A k
B’s pure strategy is Bl
If α= β=0, the game is a fair game and the game value, v=0
If α =β ≠ 0, the game is strictly determined game.
If α= β >0, the game is in favour of A
If α= β < 0, the game is in favour of B
If α ≠ β there is no saddle point, and the value of the game lies between these two values i.e.
α ≤ v ≤ β ,if α < β∨β ≤ v ≤ α ,if β< α
In this case two players cannot use pure strategies as their optimum strategies, the game is of mixed
strategies.
If there are more than one saddle point then there will be more than one solution, each solution
corresponding to each saddle point.
Example-1: Solve the following game by using maximin-minimax method.
Player A Player B
B1 B2 B3
A1 4 -2 1
A2 3 4 2
A3 -3 4 0
SOLUTION:
Step-1: Calculate the value of row minimum for each of the row and write down these values in a
column right side of the pay-off matrix under the heading “row minima”. Find the largest (maximum)
of them and enclosed by () parentheses. Let it be α and it occurs at k-th row.
Step-2: Calculate the value of column maximum for each of the column and write down these values
in a row below the pay-off matrix under the heading “column maxima”. Find the lowest (minimum)
of them and enclosed by () parentheses. Let it be β and it occurs at l-th column.
Player A Player B Row minima
B1 B2 B3
A1 4 -2 1 -2
A2 3 4 2 (2) = α
2 Minimax
First minimum then maximum

A3 -3 4 0 -3
Column maxima 4 4 (2) = β
Maximin
[First maximum then minimum]

Step-3: Here minimax α=2 and maximin β=2, since α= β=2

The game value is v= α (= β) =2
Saddle point (2,3)
A’s is pure strategy A2
B’s is pure strategy B3
Game value is 2
V is positive, the game is in favour of A.
EXAMPLE-2: Solve the following game by using maximin-minimax principle.
Player A Player B
B1 B2 B3
A1 4 -2 -1
A2 1 5 -2
A3 5 6 2
A4 1 -2 -3
SOLUTION:
Step-1: Calculate the value of row minimum for each of the row and write down these values in a
column right side of the pay-off matrix under the heading “row minima”. Find the largest (maximum)
of them and enclosed by () parentheses. Let it be α and it occurs at k-th row.
Step-2: Calculate the value of column maximum for each of the column and write down these values
in a row below the pay-off matrix under the heading “column maxima”. Find the lowest (minimum)
of them and enclosed by () parentheses. Let it be β and it occurs at l-th column.

Player A Player B Row minima

B1 B2 B3
A1 4 -2 -1 -2
A2 1 5 -2 -2
A3 5 6 22 (2)
Minimax
A4 1 -2 -3 -3
Column maxima 5 6 (2) 2/2
Maximin
Step-3: Here minimax α=2 and maximin β=2, since α= β=2
The game value is v= α (= β) =2
Saddle point (3,3)
A’s is pure strategy A3
B’s is pure strategy B3
Game value is 2
V is positive, the game is in favour of A.
Method-2: Arithmetic method for 2 x 2 game of mixed strategies
Steps are:
Step-1: Subtract the two elements in first row and write down it at the right side of second row,
ignoring sign.
Step-2: Subtract the two elements in second row and write down it at the right side of first row,
ignoring sign.
Step-3: Subtract the two elements in first column and write down it under the second column,
ignoring sign.
Step-4: Subtract the two elements in second column and write down it under the first column,
ignoring sign.
These values are called oddments.
Player A Player B A’s oddment A’s strategy
B1 B2
A1 a 11 a 12 |a21−a22|= p 1 p1
=x
p 1+ p 2 1
A2 a 21 a 22 |a11−a12|= p2 p2
=x
p 1+ p 2 2
B’s oddment |a12−a22|=q1 |a11−a21|=q 2
B’s strategy q1 q2
= y1 = y2
q1 +q 2 q1 +q 2
Value of game= V
Using A’s oddments:
a11 p1 +a21 p 2
B plays B1, V= =a11 x 1+ a21 x 2
p1 + p2
B plays B2 , V= a 12 x 1+ a22 x 2
Using B’s oddments:
A plays A1, V= a 11 y1 + a12 y 2
A plays A2, V= a 21 y 1+ a22 y 2
In all the cases game value are equal.
EXAMPLE-11: Solve the following game by arithmetic method.
Player A Player B
B1 B2
A1 2 -1
A2 -1 0
SOLUTION:
Finding the saddle point.
Player A Player B Row minima
B1 B2
A1 2 -1 -1
Minimax
A2 -1 0 -1
Minimax
Column maxima 2 0
Maximin
There is no saddle point. There is no pure strategy.
We have to solve it by using arithmetic method.
Step-1: Subtract the two elements in first row and write down it at the right side of second row,
ignoring sign.
Step-2: Subtract the two elements in second row and write down it at the right side of first row,
ignoring sign.
Step-3: Subtract the two elements in first column and write down it under the second column,
ignoring sign.
Step-4: Subtract the two elements in second column and write down it under the first column,
ignoring sign.
Player A Player B A’s A’s strategy
B1 B2 oddment
A1 2 -1 |−1−0|=1= p1 p 1 1 1
= = =x1
p 1+ p 2 1+3 4
A2 -1 0 |2−(−1)|=3= p2 p 2 3 3
= = =x2
p 1+ p 2 1+3 4
B’s oddment |−1−0|=1=q 1 |2−(−1)|=3=q 2
B’s strategy q1 1 1 q2 3 3
= = = y1 = = = y2
q1 +q 2 1+3 4 q1 +q 2 1+3 4
So, A’s optimum strategies are (A1, A2)= (1/4, ¾)=(0.25, 0.75)
B’s optimum strategies are (B1, B2)= (1/4, ¾)= (0.25, 0.75)
Game value= V
Using A’s oddments:
2∗1 (−1 )∗3 −1
B plays B1, value of the game, V= a 11 x 1 +a 21 x 2= + =
4 4 4
(−1)∗1 ( 0 )∗3 −1
B plays B2, value of the game, V= a 12 x 1+ a22 x 2= + =
4 4 4
Using B’s oddments:
 −1
A plays A1, value of the game, V= a 11 y1 + a12 y 2=¿ 2 * ¼ + (-1)* ¾=
4
−1
A plays A2, value of the game, V= a 21 y 1+ a22 y 2=(−1) * ¼ + (0)* ¾=
4

Thus the full solution of the game is

A’s optimum strategies are (A1, A2)= (1/4, ¾)=(0.25, 0.75)
B’s optimum strategies are (B1, B2)= (1/4, ¾)= (0.25, 0.75)
Value of the game, V= -1/4
The value of the game is negative, the game is in favour of B.
Method-3: Arithmetic Method (Matrix method) for 3x3 or higher order games:

In the first step, we have to try to find out any saddle point in the game. Later we can try to reduce the
game to a matrix of 2x2 or 2xn or mx2 matrix by applying the rules of dominance or any other
method. If we can reduce the game to a 2x2 size then we can easily solve it.
Steps are:
Step-1: Looking for saddle point.
Step-2: Subtract each row from the row above it (i.e. subtract 2 nd row from first row and third row
from second row) and write down the values below the matrix.
Step-3: Subtract each column from the column to its left (i.e. subtract 2 nd column from first column
and third column from second column) and write down the values to the right side of the matrix.
Step-4: Calculate the oddments for all the options (strategies) given.
Step-5: Write down these oddments calculated in step-4 neglecting their sign. Both the cases the sum
of oddments should be equal.
EXAMPLE-14:
Solve the following 3x3 game by using arithmetic (matrix) method:
Player A Player B
B1 B2 B3
A1 7 1 7
A2 9 -1 1
A3 5 7 6
SOLUTION:
Step-1: Let us look for a saddle point
Player A Player B Row minima
B1 B2 B3
A1 7 1 7 1
A2 9 -1 1 -1
A3 5 7 6 5
Maximum
Column maxima 9 7 7
Minimum Minimum
Since maximin is not equal to minimax. Hence, there is no saddle point.
Step-2: Subtract each row from the row above it (i.e. subtract 2 nd row from first row and third row
from second row) and write down the values below the matrix.
Step-3: Subtract each column from the column to its left (i.e. subtract 2 nd column from first column
and third column from second column) and write down the values to the right side of the matrix.
Player A Player B
B1 B2 B3
A1 7 1 7 7-1= 6 1-7= -6
A2 9 -1 1 9-(-1)= 10 -1-1= -2
A3 5 7 6 5-7=-2 7-6=1
 7-9= -2 1-(-1)= 2 7-1= 6
9-5= 4 -1-7= -8 1-6= -5
Step-4: Calculate the oddments for all the options A1 , A 2 , A3 , B 1 , B2 , B3(strategies) given.

Oddment for A1= |−210 −21 | = 10 x 1 – (-2) x (-2)= 10-4= 6

A =|
−2 1 |
6 −6
Oddment for 2 = 6-12= 6 [Ignoring sign]

A =|
10 −2|
6 −6
Oddment for 3 =-12+60= 48

B =|
−8 −5|
2 6
Oddment for 1 =-10+48= 38

B =|
4 −5|
−2 6
Oddment for 2 = 10-24= 14 [Ignoring sign]

B =|
4 −8|
−2 2
Oddment for 3 =16-8= 8

Step-5: Write down these oddments calculated in step-4 neglecting their sign.
Player A Player B A’s A’s strategies
Oddments
B1 B2 B3
A1 7 1 7 6= p1 x 1=6 /60
A2 9 -1 1 6= p2 x 2=6 /60
A3 5 7 6 48= p3 x 3=48/60
B’s Oddments 38= q 1 14= q 2 8= q 3 60
B’s strategies y 1=38 /60 y 2=14 /60 y 3=8/60
' p1
A s strategies , x 1= similarly, we can find x 2 , x 3 , y 1 , y 2 , y 3
p 1+ p2 + p3
Game value:
38 14 8 336 28
When A plays A1 strategy, the game value V= 7 × +1× +7 × = =
60 60 60 60 5
Method-4: Algebraic Method for 2x2 games:
Player A Player B
B1 B2
A1 a 11 a 12 x1
A2 a 21 a 22 x2
y1 y2
We assume, A uses its i-th course with probability x i and B uses its j-th course of action with
probability y j. Thus
A’s best mixed strategies are, X= [ x 1 , x 2 ¿
B’s best mixed strategies are, Y= [ y 1 , y 2 ]
Where, x 1+ x2 =1∧ y 1 + y 2=1
And x 1 , x 2 , y 1 , y 2 ≥ 0
Now the expected gain to A are:
E1 ( X )=a11 x 1+ a21 x 2; when B uses B1
And E2 ( X ) =a12 x 1 +a22 x2 ; when B uses B2
Similarly, the expected losses for B are
E1 ( Y )=a11 y 1+ a12 y 2; when A uses A1
And E2 ( Y )=a21 y 1 +a 22 y 2; when A uses A2
The value of the game= V
Since A expects to get at least V. we have
E1 ( X ) ≥V and E2 ( X ) ≥V
a 11 x 1 +a 21 x 2 ≥ V and a 12 x 1+ a22 x 2 ≥ V
Again since value of game is V. B expects to loose at most V.
E1 ( Y ) ≤V ∧E2 (Y ) ≤ V
a 11 y1 + a12 y 2 ≤V ∧a21 y1 + a22 y 2 ≤V
The solutions are:
a 22−a21
x 1=
(a ¿ ¿ 11+a 22)−( a12+ a21)¿
a11−a12
x 2=
(a ¿ ¿ 11+a 22)−(a12 +a21 )¿
a22−a 12
y 1=
( a ¿ ¿ 11+ a22)−(a12 +a 21)¿
a11 −a21
y 2=
( a ¿ ¿ 11+ a22)−(a12 +a 21)¿
Value of the game= V
a11 a 22−a12 a21
V=
(a ¿ ¿ 11+a 22)−(a12 +a21 )¿
EXAMPLE-17:
Solve the following game by Algebraic method
Player A Player B
B1 B2
A1 2 3
A2 4 -1
SOLUTION:
Let us try to find saddle point.
Player A Player B Row minima
B1 B2
A1 2 3 2
Maximum
A2 4 -1 -1
Column maxima 4 3
Minimum
Minimax is not equal to maximin. Hence, there is no saddle point. We can solve the game by using
Algebraic method.
Let us assume,
A’s best mixed strategies are, X= [ x 1 , x 2 ¿
B’s best mixed strategies are, Y= [ y 1 , y 2 ]
Where, x 1+ x2 =1∧ y 1 + y 2=1
And x 1 , x 2 , y 1 , y 2 ≥ 0
Here,
 a22−a 21
x 1=
−1−4 −5 5
(a ¿ ¿ 11+a 22)−( a12+ a21)= = = ¿
( 2−1 )−(3+4 ) −6 6
5 1
x 2=1−x 1=1− =
6 6
a22−a 12
y 1=
−1−3 −4 4
( a ¿ ¿ 11+ a22)−(a12 +a 21)= = = ¿
( 2−1 )−(3+ 4) −6 6

4 2
y 2=1− y 1=1− =
6 6

Value of the game= V

a11 a22−a 12 a21
V= (2 )(−1 )−(4)(3) −2−12 −14 14
(a ¿ ¿ 11+a 22)−(a12 +a21 )= = = = ¿
( 2−1 )−(3+4 ) −6 −6 6

( )
5 1
Mixed strategies for A: ( A ¿ ¿ 1 , A 2 )= , ¿
6 6

( )
4 2
Mixed strategies for B: (B ¿ ¿ 1 , B2 )= , ¿
6 6
Game value, V= 14/6

SOLUTION:
Let H and T represents head and tail respectively. The pay-off can be developed as follows:
Player A Player B
H T
H 2 -1
T -1 3
Two methods:
i. Arithmetic method
ii. Algebraic method
Homework.
Method-5: Graphical Method
By using graphical method, it is possible to reduce a game of order 2xn or mx2 to a game of order 2x2
and then we can solve it by using either arithmetic or algebraic method.
Let us develop a game of order 2xn
Player A Player B
B1 B2 B3 ………….. Bn
A1 a 11 a 12 a 13 …………… a 1n x1
A2 a 21 a 22 a 23 …………… a 2n x2
y1 y2 y3 …………….. yn
Without any saddle point.
Let us assume the mixed strategies used by player A; X= ( x 1 , x 2 ¿ and by player B; Y= (
y1 , y2 … … … . yn ¿
Where x 1+ x2 =1 and y 1 + y 2 +… … … .+ y n=1
x1 , x2 ≥ 0 , y1 , y2 , … … … . , y n ≥ 0
The net expected gain of A when player B plays his pure strategies B j is given by
E j ( X )=a 1 j x 1 +a 2 j x 2 ; j=1 , 2 ,3 … … … … ,n
x 1+ x2 =1; x 2=1−x 1
E j ( X )=a 1 j x 1 +a 2 j (1−x ¿¿ 1)¿
When x 1=0
E j ( X )=a 1 j x 1 +a 2 j (1−x ¿¿ 1)¿= a 1 j × 0+a 2 j (1−0)=a 2 j
j=1; E1 ( X )=a21
j=2; E2 ( X ) =a22

When x 1=1
E j ( X )=a 1 j x 1 +a 2 j (1−x ¿¿ 1)¿= a 1 j × 1+ a2 j (1−1)=a 1 j
We have to draw two vertical lines where x 1=0∧x1 =1
Then we have to draw n line segments joining the points (0, a 2 j ¿∧( 1 , a1 j ) , j=1 ,2 , 3 , … … … … , n
After drawing all the lines in the graph we have to find out the lower envelope of these segments
which will give the minimum expected gain of A as a function of
x 1 . The highest point of thelower envelope will givethe maximum of minimum gainof A .
The line segments passing through the point corresponding to B’s pure moves say Bi∧B j are the
critical moves for B which will maximize the minimum expected gain of A.
EXAMPLE-20:
Solve the following 2x4 game using graphical method.
Player A Player B Mixed
B1 B2 B3 B4 strategies
A1 3 2 -1 4 x1
A2 2 5 6 -2 x2
Mixed y1 y2 y3 y4
strategies
SOLUTION:
Finding a saddle point
Player A Player B Row
B1 B2 B3 B4 minima
A1 3 2 -1 4 -1
Maximum
value
A2 2 5 6 -2 -2
Column 3 5 6 4
maxima Minimum
value
So, the minimax value is not equal to maximin value, as a result there is no saddle point. We have
apply graphical method to reduce the game into 2x2 game.
Let us assume the mixed strategies used by player A; X= ( x 1 , x 2 ¿ Where x 1+ x2 =1
And both x 1 , x 2lie in the open interval (0,1)
The net expected gain of A when player B plays his pure strategies B j is given by
E j ( X )=a 1 j x 1 +a 2 j x 2 ; j=1 , 2 ,3 … … … … ,n
x 1+ x2 =1; x 2=1−x 1
E j ( X )=a 1 j x 1 +a 2 j (1−x ¿¿ 1)¿
When x 1=0
E j ( X )=a 1 j x 1 +a 2 j (1−x ¿¿ 1)¿= a 1 j × 0+a 2 j (1−0)=a 2 j
When x 1=1
E j ( X )=a 1 j x 1 +a 2 j (1−x ¿¿ 1)¿= a 1 j × 1+ a2 j (1−1)=a 1 j
We have to draw two vertical lines where x 1=0∧x1 =1
The points to draw the different lines based on B’s pure strategies are
When B plays B1; The point are (0, 2) and (1, 3)
When B plays B2; The point are (0, 5) and (1, 2)
When B plays B3; The point are (0, 6) and (1, -1)
When B plays B4; The point are (0, -2) and (1, 4)

From the graph, the maximum point of the lower envelope lies on the point of intersection of the line
segments B3 and B4
Therefore, the game reduces to 2x2 pay-off matrix

Player A Player B
B3 B4
A1 -1 4
A2 6 -2
The game now can be solved by using either arithmetic method or algebraic method.
HOMEWORK:
EXAMPLE-21
Solve the following game using graphical method
Player A Player B
B1 B2
A1 2 -3
A2 -2 5
A3 0 -1
SOLUTION:
Finding saddle point
 Player A Player B Row
B1 B2 minima
A1 2 -3 -3
A2 -2 5 -2
A3 0 -1 -1
Maximum
Column 2 5
maxima Minimum
The minimax is not equal to the maximin. Hence, there is no saddle point. We can reduce the game by
using graphical method to a 2x2 game.
Let us assume the mixed strategies used by player B; Y= ( y 1 , y 2 ¿ Where y 1 + y 2=1
And both y 1 , y 2 lie in the open interval (0,1)
The net expected loss of B when player A plays his pure strategies Ai is given by
Ei ( Y )=ai 1 y 1 +a i2 y 2 ; i=1 , 2 ,3 … … … … , m
y 1 + y 2=1 ; y 2=1− y 1
Ei ( Y )=ai 1 y 1 +a i2 (1− y ¿¿ 1)¿
When y 1=0
Ei ( Y )=ai 1 y 1 +a i2 (1− y ¿¿ 1)¿= a i1 ×0+ ai 2 (1−0)=a i2
When y 1=1
Ei ( Y )=ai 1 y 1 +a i2 (1− y ¿¿ 1)¿= a i1 ×1+a i 2(1−1)=a i1
We have to draw two vertical lines where y 1=0∧ y 1=1
The points to draw the different lines based on A’s pure strategies are
When A plays A1; The point are (0, -3) and (1, 2)
When A plays A2; The point are (0, 5) and (1, -2)
When A plays A3; The point are (0, -1) and (1, 0)
Let us draw the graphs for the line segments for above 3 pairs of points.
From the graph, we can see from the upper envelope the minimum point is the intersection point of
line segment corresponds to A1 and A2 strategies of player A. hence the reduced 2x2 game is as
follows:
Player A Player B
B1 B2
A1 2 -3
A2 -2 5
Now we can solve this 2x2 game by using either arithmetic or algebraic method.
Method-6: Sub-Game
Method-7: Dominance method
The size of the pay-off matrix can be reduced by using the principle of dominance which states:
It can be reduced to a 2x2 [arithmetic or algebraic method] or it can be reduced either to 2xn or mx2
[Graphical method].
Rules of dominance:
i. If all the elements of i-th row are less than or equal to the corresponding elements of j-th
row then we shall say that i-th row is dominated by j-th row and we reject the i-th row.
We can take corresponding strategy of the player, x i=0 .
Player B
B1 B2 B3 B4
A1 -5 3 1 20
Player A A2 5 5 4 6
A3 -4 -2 0 -5
All the elements in third row are less than or equal to corresponding elements in second
row. So, third row is dominated by second row. We eliminate third row and x 3=0 The
reduced pay-off matrix is
Player B
B1 B2 B3 B4
A1 -5 3 1 20
Player A A2 5 5 4 6

ii. If all the elements of i-th column are greater than or equal to the corresponding elements
of j-th column then we shall say that i-th column is dominated by j-th column and we
reject the i-th column. We can take corresponding strategy of the player, y i=0 .
Player B
B1 B2 B3 B4
A1 -5 3 1 20
Player A A2 5 5 4 6
For pure move B2 and B4 of B, all the elements of the second and fourth columns are
greater than of equal to the corresponding elements of the third column. Thus the second
 and fourth columns are dominated by third column. Hence, we reject second and fourth
columns and y 2=0 , y 4 =0, the reduced pay-off matrix is as follows
Player B
B1 B3
A1 -5 1
Player A A2 5 4
iii. If all the elements of i-th row are less than or equal to the corresponding elements of
convex combination of other two or more rows, then we shall say that i-th row is
dominated by that two or more other rows and we shall reject i-th row and we can take
the strategy of row player x i=0 .
Player B
B1 B2 B3 B4
A1 5 4 3 20
Player A A2 5 5 4 6
A3 4 2 0 5
4.5 3.5 2 5.5

iv. If all the elements of i-th column are greater than or equal to the corresponding elements
of convex combination of other two or more columns, then we shall say that i-th column
is dominated by that two or more other columns and we shall reject i-th column and we
can take the strategy of column player y i=0 .
v. If convex combination of two or more rows (or columns) is dominated by any single row
(or column) then we shall reject any one row (or column) from the combinated rows (or
columns).
Player B
B1 B2 B3 B4
A1 5 4 3 20
Player A A2 5 5 4 6
A3 4 2 0 5
(A2+A3)/2 4.5 3.5 2 5.5
A1 5 4 3 20
The rules (iii), (iv) and (v) also called the modified dominance property or rows and
columns or pay-off matrix.
EXAMPLE-30: Solve the following game
Player B
B1 B2 B3 B4
A1 -5 3 1 20
Player A A2 5 5 4 6
A3 -4 -2 0 -5
SOLUTION:
Let us find the saddle point
Player B Row
B1 B2 B3 B4 minima
A1 -5 3 1 20 -5
Player A A2 5 5 4 6 4
Maximum
A3 -4 -2 0 -5 -5
Column maxima 5 5 4 20
Minimum
Here, the maximin= minimax, so the game has a saddle point at A2 and B3
Let us apply the dominance rules to reduce the game.
Let us assume the optimal strategies of the player A and B are X=( x 1 , x 2 , x 3 ¿ and Y= (
y 1 , y 2 , y 3 , y 4 ¿ respectively.

Player B
B1 B2 B3 B4
A1 -5 3 1 20
Player A A2 5 5 4 6
A3 -4 -2 0 -5
Applying dominance rules:
Step-1: All the elements in third row are less than or equal to corresponding elements in second row.
So, third row is dominated by second row. We eliminate third row and x 3=0 The reduced pay-off
matrix is:
Player B
B1 B2 B3 B4
A1 -5 3 1 20
Player A A2 5 5 4 6
Step-2: For pure move B2 and B4 of B, all the elements of the second and fourth columns are greater
than of equal to the corresponding elements of the third column. Thus, the second and fourth columns
are dominated by third column. Hence, we reject second and fourth columns and y 2=0 , y 4 =0, the
reduced pay-off matrix is as follows:
Player B
B1 B3
A1 -5 1
Player A A2 5 4
Step-3: All the elements in first row are less than or equal to the corresponding elements in second
row. So, the first row is dominated by second row and we reject first row and x 1=0, The reduced
pay-off as follows:
Player B
B1 B3
Player A A2 5 4
Step-4: all the elements in first column are greater than or equal to the corresponding elements in
second column. So, the first column is dominated by second column and we reject first column and
y 1=0 . The reduced pay-off is as follows:
Player B
B3
Player A A2 4
The value of the game is, v= 4, since this is positive, the game is in favour of A.
Optimum strategies for A= (0,1,0)
Optimum strategies for B= (0,0,1,0)
EXAMPLE-31, 32, 33, 34, 35
Method-8: Iterative method for approximate solution (Brow’s algorithm)

DECISION ANALYSIS
Decision analysis can be used to determine an optimal strategy when a decision maker is
faced with several decision alternatives and an uncertain or risk-filled pattern of future
events. For example, a company would like to find the best location for its plant. Say, it has
five options in five countries.
Factors that influence the decision are: world economy, demand in various regions of the
world, labor availability, raw materials costs, transportation costs, and so on.
Problem Formulation:
We have to start with a verbal statement of the problem.
Then we have to identify the decision alternatives, the uncertain future events, referred to
as chance events, and the consequences associated with each decision alternative and each
chance event outcome.
PDC (Pitsburgh development Corporation) has three different sized projects:
Three decision alternatives
d 1=a small complex
d 2=a mediumcomplex
d 3=a large complex
There are two uncertain future events which we call state of nature
s1=strong demand for the complex
s2=weak demand for thecomplex
Pay-off table for consequences
Pay-off for the PDC project (in $ million profit)
Decision alternatives State of Nature
Strong demand s1 Weak demand s2
Smallcomplex , d 1 8 7
Medium complex , d 2 14 5
Large complex , d 3 20 -9
Influence Diagram:
An influence diagram is a graphical device that shows the relationships among the decisions,
the chance events, and the consequences for a decision problem. The nodes in an in fluence
diagram represent the decisions, chance events, and consequences. Rectangles or squares
depict decision nodes, circles or ovals depict chance nodes, and diamonds depict
consequence nodes. The lines connecting the nodes, referred to as arcs, show the direction
of influence that the nodes have on one another.

State of Nature
Strong (s1)
Demand Weak (s2)

Complex size Profit

Consequence
Profit
Decision alternatives
Small complex (d1)
Medium complex (d2)
Large complex (d3)
Pay-off Tables:
A table showing payoffs for all combinations of decision alternatives and state of nature if payoff
table.
Pay-off for the PDC project (in $ million profit)
Decision alternatives State of Nature
Strong demand s1 Weak demand s2
Smallcomplex , d 1 8 7
Medium complex , d 2 14 5
Large complex , d 3 20 -9

Decision Tree:
A decision tree provides a graphical representation of the decision-making process. It shows the
natural or logical progression that will occur over time.

Decision making without probabilities:

Optimistic Approach:
The optimistic approach evaluates each decision alternative in terms of the best payoff that can occur.
The decision alternative that is recommended is the one that provides the best possible payoff. For a
problem in which maximum profit is desired, as in the PDC problem, the optimistic approach would
lead the decision maker to choose the alternative corresponding to the largest profit. For problems
involving minimization, this approach leads to choosing the alternative with the smallest payoff.
 Maximum payoff for each PDC decision alternative

Decision alternative Maximum payoff

Smallcomplex , d 1 8
Medium complex , d 2 14
Large complex , d 3 20
Maximum of the maximum payoff values
The maximum of the maximum payoff values is 20 which corresponds to Large complex , d 3. The
decision to construct the large complex is recommended.
Conservative Approach:
The conservative approach evaluates each decision alternative in terms of the worst pay off that can
occur. The decision alternative recommended is the one that provides the best of the worst possible
payoffs. For a problem in which the output measure is profit, as in the PDC problem, the conservative
approach would lead the decision maker to choose the alternative that maximizes the minimum
possible profit that could be obtained. For problems involving minimization, this approach identifies
the alternative that will minimize the maximum payoff.
Minimum payoff for each PDC decision alternative

Decision alternative Maximum payoff

Smallcomplex , d 1 7
Maximum of the minimum payoff values
Medium complex , d 2 5
Large complex , d 3 -9
The maximum of the minimum payoff values is 7 which corresponds to Smallcomplex , d 1, the
decision alternative small complex is recommended.
Minimax Regret Approach:

The minimax regret approach to decision making is neither purely optimistic nor purely
conservative. Let us illustrate the minimax regret approach by showing how it can be used to
select a decision alternative for the PDC problem.
Opportunity loss or regret:
Rij =|V j¿ −V ij|

Where,
Rij =the regret associated with decision alternative d i∧state of nature s j
¿
V j =the payoff value corresponding¿ the best decision for the state of nature s j
V ij =the payoff corresponding ¿ decision alternative d i∧state of nature s j
To calculate the regret, we simply subtract each entry in a column from the largest entry in the
column. Thus, the regret table is as follows:
Opportunity loss or regret table

Decision alternatives State of Nature

s
Strong demand 1 Weak demand s2
Smallcomplex , d 1 |20−8|=12 |7−7|=0
 Medium complex , d 2 |20−14|=6 |7−5|=2
Large complex , d 3 |20−20|=0 |7−(−9)|=16
In the next step, we list the maximum regret for each decision alternative and then we select the
alternative corresponding to the minimum regret from the maximum regret.

Maximum regret for each decision alternative

Decision alternative Maximum regret

Smallcomplex , d 1 12
Medium complex , d 2 6
Minimum of the maximum regret
Large complex , d 3 16
The minimum of the maximum regret is 6 corresponding to the decision alternative
m edium complex , d 2, hence the decision alternative for medium complex is recommended.
Decision making with probabilities:
When probabilities are available, we can use the expected value approach to identify the best decision
alternative.
Let,
N= the number of states of nature

P ( s j ) =the probability of state of nature s j

The probabilities must satisfy two conditions:

P ( s j ) ≥ 0 for all states of nature

N

∑ P ( s j )=P ( s1 ) + P ( s2 ) + … … … … … …+ P ( s N )=1
j=1

The expected value (EV) of decision alternative d i is defined as follows

N
EV ( d i )=∑ P ( s j ) V ij
j=1

Let us assume,

P ( s 1 )=0.8

P ( s 2 )=0.2
Expected value for each decision alternative is:

EV(d 1 ¿=¿ 8 x 0.8+ 7 x 0.2= 7.8

EV(d 2 ¿=14 x 0.8+ 5 x 0.2= 12.2

EV(d 3 ¿=20 x 0.8+ (-9) x 0.2= 14.2

The maximum expected value is 14.2 corresponding to decision alternative d 3 large complex,
hence the large complex is recommended based on expected maximum profit.
Expected Value of Perfect Information:
If PDC could determine with certainty, prior to making a decision, which state of nature is
going to occur.
Expected value of perfect information
EVPI= |EVwPI −EVwoPI |
EVPI= expected value of perfect information
EVwPI= expected value with perfect information about the state of nature
EVwoPI= expected value without perfect information about the state of nature
If PDC sure about s1 then the best strategy is d 3 and receive a payoff of $20 million and if PDC
sure about s2 then the best strategy is d 1 and receive a payoff of $7 million and we are given
P ( s 1 )=0.8 and P ( s 2 )=0.2

Thus the expected value with perfect information is

EVwPI= 20 x 0.8+ 7 x 0.2= 17.4
Expected value without perfect information is based on expected value for each decision alternative

EV(d 1 ¿=¿ 8 x 0.8+ 7 x 0.2= 7.8

EV(d 2 ¿=14 x 0.8+ 5 x 0.2= 12.2

EV(d 3 ¿=20 x 0.8+ (-9) x 0.2= 14.2

Expected value without perfect information is the value which is the highest among expected values
for decision alternatives. In this problem,
EVwoPI= 14.2

Expected value of perfect information

EVPI= |EVwPI −EVwoPI |=|17.4−14.2|=3.2
Additional expected value is $3.2 million that can be obtained if perfect information were available
about the state of nature.
Sensitivity analysis:
Sensitivity analysis can be used to determine how changes in the probabilities for the state of nature
and/or changes in the payoffs affect the recommended decision alternative.

Case-1: how changes in the probabilities for the state of nature affect the recommended decision
alternative
Pay-off for the PDC project (in $ million profit)
Decision alternatives State of Nature
Strong demand s1 Weak demand s2
Smallcomplex , d 1 8 7
Medium complex , d 2 14 5
Large complex , d 3 20 -9
Let, p denote the probability of state of nature s1 ; that is P( s¿¿ 1)= p ¿

Since, there are two state of nature in the PDC problem, the probability of state of nature s2

P(s¿¿ 2)=1−P(s ¿¿ 1)=1− p ¿ ¿

The expected value for each decision alternative can be calculated as follows:

EV(d 1 ¿=¿ 8 x p+ 7 x (1-p)= 8p+7-7p= p+7

EV(d 2 ¿=14 x p+ 5 x (1-p)= 14p+5-5p= 9p+5

EV(d 3 ¿=20 x p+ (-9) x (1-p)= 20p-9+9p= 29p-9

We have to draw two vertical lines where p=0 and p=1

In the both lines, we have to show the expected value vertically for both p=0 and p=1 for each
decision alternative.

EV(d 1 ¿=¿ p+7

When p=0, EV(d 1 ¿=7 ; p=1, EV(d 1 ¿=8

EV(d 2 ¿=14 x p+ 5 x (1-p)= 14p+5-5p= 9p+5

When p=0, EV(d 2 ¿=5 ; p=1, EV(d 2 ¿=14

EV(d 3 ¿=20 x p+ (-9) x (1-p)= 20p-9+9p= 29p-9
When p=0, EV(d 3 ¿=−9 ; p=1, EV(d 3 ¿=20

From the graph, at point A for the probability A, EV(d1)= EV(d2)

p+7=9p+5
or, 9p+5= p+7
or, 9p-p= 7-5
or, 8p= 2
or, p= 2/8= ¼= 0.25
The probability at point A is 0.25
From the graph, at point B for the probability B, EV(d2)= EV(d3)
9p+5=29p-9
Or, 29p-9= 9p+5
Or, 29p-9p= 5+9
Or, 20p= 14
Or, p= 14/20= 0.70
At point B, p= 0.70
For the probability of state of nature s1
p ≤ 0.25 d 1 is the best
0.25 ≤ p ≤ 0. 70 d 2 is the best
p ≥ 0.70 d 3 is the best
Due to changes in the probability of state of nature the recommended decision alternative will
change.
Case-2: how changes in the payoffs affect the recommended decision alternative
Pay-off for the PDC project (in $ million profit)
Decision alternatives State of Nature
Strong demand 1s Weak demand s2
Smallcomplex , d 1 8 7
Medium complex , d 2 14 5
Large complex , d 3 20 -9
Let the probability for state of nature

P(s¿¿ 1)=0.80 ¿
P(s¿¿ 2)=1−0.80=0.20 ¿
EV(d 1 ¿=¿ 8 x 0.8+ 7 x 0.2= 7.8

EV(d 2 ¿=14 x 0.8+ 5 x 0.2= 12.2

EV(d 3 ¿=20 x 0.8+ (-9) x 0.2= 14.2

Based on expected value, decision alternative d3 is recommended as because it has the highest
expected value.
Sensitivity analysis for decision alternative d3
The expected value for d3 (the highest one) is 14.2 and the second highest expected value is for d 2
which is 12.2

Decision alternative d3 will remain the optimal decision alternative as long as EV( d 3 ¿ is greater than
or equal to the expected value of the second best decision alternative. Thus, decision
alternative d 3 will remain the optimal decision alternative as long as
EV (d 3 )≥ 12.2

Let, S= the payoff of decision alternative d 3 when demand is strong

W= the payoff of decision alternative d 3 when demand is weak.

Using the probability, P(s¿¿ 1)=0.8 0 ; P( s¿ ¿ 2)=1−0.80=0.20 ¿ ¿

EV(d 3 ¿=S x 0.8+ Wx 0.2= 0.8S+0.2W

Assuming that the payoff for d 3 stays at its original value of -9 million when demand is weak,
d 3 will remain optimal as long as

EV(d 3 ¿=¿ 0.8S+0.2(-9)≥ 12.2

Or, 0.8S-1.8≥ 12.2

Or, 0.8S-1.8+1.8≥ 12.2+ 1.8
Or, 0.8S≥ 14
Or, 0.8S/0.8≥ 14/0.8
Or, S≥ 17.5
The decision alternative d 3 will remain optimal as long as the payoff for d 3 when demand is
strong is at least $17.5 million.
Assuming that the payoff for d 3 stays at its original value of 20 million when demand is
strong, d 3 will remain optimal as long as
EV(d 3 ¿=¿ 0.8 x 20+0.2W≥ 12.2

Or, 16+0.2W≥ 12.2

Or, 16+0.2W-16≥ 12.2−16
Or, 0.2W≥−3.8
Or, 0.2W/0.2≥−3.8 /0.2
Or, W≥−19
The decision alternative d 3 will remain optimal as long as the payoff for d 3 when demand is
weak is at least -$19 million.
Sensitivity analysis for decision alternative d2

Decision alternative d3 will remain the optimal decision alternative as long as EV( d 2 ¿ is less than or
equal to the expected value of the best decision alternative. Thus, decision alternative d 3 will
remain the optimal decision alternative as long as
EV (d 2 )≤ 1 4 .2

Let, S= the payoff of decision alternative d 2 when demand is strong

W= the payoff of decision alternative d 2 when demand is weak.

Using the probability, P(s¿¿ 1)=0.8 0 ; P( s¿ ¿ 2)=1−0.80=0.20 ¿ ¿

EV(d 2 ¿=S x 0.8+ Wx 0.2= 0.8S+0.2W

Assuming that the payoff for d 2 stays at its original value of $5 million when demand is
weak, d 3 will remain optimal as long as
EV(d 2 ¿=¿ 0.8S+0.2(5)≤ 1 4 .2

Or, 0.8S+1≤ 1 4 .2
Or, 0.8S+1-1≤ 1 4 .2−1
Or, 0.8S≤ 13.2
Or, 0.8S/0.8≤ 13.2/0.8
Or, S≤ 16.5
The decision alternative d 3 will remain optimal as long as the payoff for d 2 when demand is
strong is at best $16.5 million.
Assuming that the payoff for d 2 stays at its original value of $14 million when demand is
strong, d 3 will remain optimal as long as
EV(d 2 ¿=¿ 0.8 x 14+0.2W≤ 14 .2

Or, 11.2+0.2W≤ 1 4 .2
Or, 11.2+0.2W-11.2≤ 1 4 .2−11.2
Or, 0.2W≤ 3
Or, 0.2W/0.2≤ 3/0.2
Or, W≤ 15
The decision alternative d 3 will remain optimal as long as the payoff for d 2 when demand is
weak is at best $15 million.
Sensitivity analysis for decision alternative d1

Decision alternative d3 will remain the optimal decision alternative as long as EV( d 1 ¿ is less than or
equal to the expected value of the best decision alternative. Thus, decision alternative d 3 will
remain the optimal decision alternative as long as
EV (d 1 )≤ 14.2

Let, S= the payoff of decision alternative d 1 when demand is strong

W= the payoff of decision alternative d 1 when demand is weak.

WAITING LINE MODELS