Decision theory

Steps in Decision theory approach

Decision making Environments
Making under conditions of
Certainty,
Uncertainty,
Conditions of Risk
 what is a decision?
• A decision is the conclusion of a process
designed to weigh the relative utilities or
merits of a set of available alternatives so that
the most preferred course of action can be
selected for implementation.
• Decision-making involves all that is necessary
to identify the most preferred choice to satisfy
the desired goal or objective.

2
 what is a decision?...
• Hence decision-making process must involve
– a set of goals or objectives,
– a system of priorities,
– methods of enumerating the alternative courses
of feasible and viable courses and
– a system of identifying the most favorable
alternative.

3
 why must decisions be made?
• In management theory we study that the essence of
management is to make decisions that commit
resources in the pursuit of organizational objectives.
• Resources are limited and wants and needs of human
beings are unlimited and diversified and each wants to
satisfy his needs in an atmosphere, where resources
are limited.
• Here the decision theory helps to take a certain
decision to have most satisfactory way of satisfying
their needs.
• Decisions are made to achieve these goals and
objectives.

4
 Decision and conflict
• When a group of people is working together in an
organization, due to individual behaviour and mentality,
there exists a conflict between two individuals.
• Not only that in an organization, each department has its
own objective, which is
– subordinate to organizational goal, and
– in fulfilling departmental goals, there exists a conflict between
the departments.
• Hence, any decision maker has to take all these factors into
consideration, while dealing with a decision process, so
that
– the effect of conflicts between departments or between
subordinate goals is kept at minimum in the interest of
achieving the overall objective of the organization.

5
 Phases of
decision-making
• In decision-making process we recognize two
phases:
– (1) How to formulate goals and objectives, enumerate
environmental constraints, identify alternative
strategies and project relevant payoffs.
– (2) Concentration on the question of how to choose
the optimal strategy when we are given a set of
objectives, strategies, payoffs.
• We concentrate more on the second aspect in
our discussion

6
 Classifications of decisions
• In general, decisions are classified as
– Strategic decision, which is related to the
organization's outside environment,
– Administrative decisions dealing with structuring
resources and
– Operational decisions dealing with day-to-day
problems

7
 Classifications of decisions …
• Depending on the nature of the problem there are
Programmed decisions, to solve repetitive and well-
structured problems, and Non-programmed decisions,
designed to solve non-routine, novel, illstructured
problems.
• Depending on the scope, complexity and the number of
people employed decision can be divided as individual and
managerial decisions.
• Depending on the sphere of interest, as political,
economic, or scientific etc.
• Decision can be divided as static decision requiring only
one decision for the planning horizon and dynamic decision
requiring a series of decisions for the planning horizon.

8
 Steps in
decision theory approach
• 1. List the viable alternatives (strategies) that can
be considered in the decision.
• 2. List all future events that can occur. These
future events (not in the control of decision
maker) are called as states of nature.
• 3. Construct a payoff table for each possible
combination of alternative course of action and
state of nature.
• 4. Choose the criterion that results in the largest
payoff.
9
 Decision making
under certainty (DMUC)
• Decision making under certainty assumes that all
relevant information required to make decision is
certain in nature and is well known.
• It uses a deterministic model, with complete
knowledge, stability and no ambiguity.
• To make decision, the manager will have to be
quite aware of the strategies available and their
payoffs and each strategy will have unique payoff
resulting in certainty.
• The decision-making may be of single objective or
of multiple objectives.
10
 Example 1
ABC Corporation wants to launch one of its mega campaigns to promote
a special product. The promotion budgets not yet finalized, but they
know that some Rs. 55,00,000 is available for advertising and
promotion. Management wants to know how much they should spend
for television spots, which is the most appropriate medium for their
product. They have created five ‘T.V. campaign strategies’ with their
projected outcome in terms of increase in sales. Find which one they
have to select to yield maximum utility. The data required is given
below.

11
 Example 1 …
• Solution
• The criteria for selecting the strategy (for maximum utility)
is to select the strategy that yields for maximum utility i.e.
highest ratio of outcome i.e. increase in sales to cost.

The company will select the third strategy, C, which yields

highest utility.

12
 Decision making
under risk (DMUR)
• Decision-making under risk (DMUR) describes a situation in which
each strategy results in more than one outcomes or payoffs and the
manager attaches a probability measure to these payoffs.
• This model covers the case when the manager projects two or
more outcomes for each strategy and he or she knows, or is
willing to assume, the relevant probability distribution of the
outcomes.
• The following assumptions are to be made:
– (1) Availability of more than one strategies,
– (2) The existence of more than one states of nature,
– (3) The relevant outcomes and
– (4) The probability distribution of outcomes associated with each
strategy.
• The optimal strategy in decision making under risk is identified by
the strategy with highest expected utility (or highest expected
value).

13
 Example 2
• A marketing manager of an insurance company has kept complete records
of the sales effort of the sales personnel. These records contain data
regarding the number of insurance policies sold and net revenues received
by the company as a function of four different sales strategies. The
manager has constructed the conditional payoff matrix given below, based
on his records. (The state of nature refers to the number of policies sold).
The number within the table represents utilities. Suppose you are a new
salesperson and that you have access to the original records as well as the
payoff matrix. Which strategy would you follow?

14
 Example 2 …
• Solution
• As the decision is to be made under risk, multiplying
the probability and utility and summing them up give
the expected utility for the strategy

As the third strategy gives highest expected utility 1 call

and 2 follow up yield highest utility.
15
 Decision making
under uncertainty
• In decision making under uncertainty, remember that
no probabilities are attached to set of the states of
nature.
• Sometimes we may have only positive elements in the
given matrix, indicating that the company under any
circumstances will have profit only.
• Sometimes, we may have negative elements, indicating
potential loss.
• While solving the problem of decision making under
uncertainty, we have two approaches,
– the first one is pessimistic approach and
– the second one is optimistic approach.

16
 Example 3
The management of XYZ company is considering the use of a newly discovered
chemical which, when added to detergents, will make the washing stet, thus
eliminating the necessity of adding softeners. The management is considering at
present time, these three alternative strategies. S1 = Add the new chemical to the
currently marketed detergent DETER and sell it under label ‘NEW IMPROVED DETER’.
S2 = Introduce a brand new detergent under the name of ‘SUPER SOFT’ S3 = Develop a
new product and enter the softener market under the name ‘EXTRA WASH’. The
management has decided for the time being that only one of the three strategies is
economically feasible (under given market condition). The marketing research
department is requested to develop a conditional payoff matrix for this problem. After
conducting sufficient research, based on personal interviews and anticipating the
possible reaction of the competitors, the marketing research department submits the
payoff matrix given below. Select the optimal strategy.

17
 Example 3 …
• Solution
• When no probability is given, depending upon
risk, subjective values, experience etc., each
individual may choose different strategies.
• These are selected depending on the choice
criterion. That is why sometimes the decision
making under uncertainty problems are labeled
as choice creation models.
• Two criterions may be considered here. One is
Criterion of Optimism and the other is Criterion
of Pessimism.
18
 References
• Premkumar Gupta and Hira, “Operation
Research”, Third Edition S Chand Company
Ltd., New Delhi 2003.
• Rama Murthy P “Operations Research” second
edition, New Age International (P) Limited,
New Delhi, 2007.
• Sundaresan. V, GanapathySubramanian. K.S,
“Resource Management Techniques:
Operations Research” A.R Publications, 2003

19
 Game theory

Theory of Games ,
Characteristics
Game models -Definition - Rules -
Pure Strategy
Optimal solution of two person zero sum games,
mixed strategies
 What is Game?
• A game is defined as a collection of rules, known to all players,
which determine what players may do, & the outcomes &
payoffs resulting from their choices.

• A competitive situation is called a game if it has the following

characteristics:
– a) There are a finite number of participants called players.
– b) Each player has a finite number of possible courses of action.
– c) A play occurs when each player chooses one of his courses of
action. The choices are assumed to be made simultaneously, i.e., no
player knows the opponent's choice until he has decided on his
own.
– d) Every combination of courses of action finds an outcome, which
results in a gain to each player. (A loss is considered as a negative
gain)
2
 Zero sum two-person game
• The game which has only two players, with
the gain of one player is the loss of the other
player is called zero sum two-person game. In
other words, no capital is created or destroyed
during the courses of action in zero sum
games.
• The gains resulting from a zero sum two
persons game is mostly represented in the
form of a matrix, known as the pay off matrix.

3
 Saddle Point and Value of the Game
• Definition :A Saddle point of a payoff matrix is that position in
the payoff matrix where maximum of row minima coincides
with the minimum of the column maxima. The payoff at the
saddle point is called the value of the game denoted by v. The
saddle point need not be unique.

• Mixed Strategy : When maximin # minimax, then pure

strategy fails. Therefore each player with certain probabilistic
fixation. This type of strategy is called mixed strategy.
4
 Example 1
• Solve the game whose pay-off matrix is given
by

5
Example 1…

6
 Example 2
• Solve the following game whose pay-off
matrix is

7
Example 2…

8
 Games without Saddle Points, Mixed
Strategies
Example 1

9
10
11
12
Example 2

13
14
15
 References
• Premkumar Gupta and Hira, “Operation
Research”, Third Edition S Chand Company
Ltd., New Delhi 2003.
• Rama Murthy P “Operations Research” second
edition, New Age International (P) Limited,
New Delhi, 2007.
• Sundaresan. V, GanapathySubramanian. K.S,
“Resource Management Techniques:
Operations Research” A.R Publications, 2003

16
 Game theory

Dominance property
Graphical method_2 x n or m x 2
games
 Dominance Property
• Dominance Property player is inferior to at
least one of the remaining ones. The superior
strategies are said to dominate the inferior
ones. In such cases of dominance, we can
reduce the size of the payoff matrix by
deleting those strategies which are dominated
by others.

2
Dominance Property…

3
• Solve the following game using dominance
Property

4
• Row II is dominated by Row II and column III is
dominated by column I.

5
6
7
 Example 2
Using Dominance property solve

8
9
10
 Graphical Method
• Example 1: Use the notion of dominance to
simplify the rectangular game with the
following payoff, and solve it graphically.

11
12
13
14
 Example 2
• Solve the following 2 x 4 game graphically

15
• Consider two axes say axis 1, axis II vertically
at unit distance apart.

16
17
18
 References
• Premkumar Gupta and Hira, “Operation
Research”, Third Edition S Chand Company
Ltd., New Delhi 2003.
• Rama Murthy P “Operations Research” second
edition, New Age International (P) Limited,
New Delhi, 2007.
• Sundaresan. V, GanapathySubramanian. K.S,
“Resource Management Techniques:
Operations Research” A.R Publications, 2003

19
 12
Chapter
Game Theory

12.1 Introduction

In many practical problems, it is required to take a decision in a

competitive situation in which the outcome depends not on that
decision alone but rather on the interaction between the decision -
maker and that of a competitor. Such problem occurs frequently in
economic military, social, political, advertising & marketing by
competing business forms. Game theory is the study of such
situations & in general both conflict and co - operation are involved.

12.2 Definitions of some important terms

A game is defined as a collection of rules, known to all players, which

determine what players may do, & the outcomes & payoffs resulting
from their choices.

A competitive situation is called a game if it has the following

characteristics:

a) There are a finite number of participants called players.

b) Each player has a finite number of possible courses of action.
c) A play occurs when each player chooses one of his courses of
action. The choices are assumed to be made simultaneously,
i.e., no player knows the opponent's choice until he has
decided on his own.
d) Every combination of courses of action finds an outcome,
which results in a gain to each player. (A loss is considered as
a negative gain)
300 Operations Research

Zero sum two-person game

The game which has only two players, with the gain of one player is
the loss of the other player is called zero sum two-person game. In
other words, no capital is created or destroyed during the courses of
action in zero sum games.

The gains resulting from a zero sum two persons game is mostly
represented in the form of a matrix, known as the pay off matrix. An
example of such a matrix is given below.

Player B
1 2 3 4
1 5 0 -2 1
Player A 2 3 -1 0 3
3 1 4 5 3

The number of rows of the matrix corresponds to the number of

courses of action of player A & the number of columns corresponds to
the number of courses of action of player B. The elements within the
matrix represent the gain to player A for each outcome of the game.
Thus, a negative entry indicates the negative gain i.e., loss to player A.
i.e., player A has to pay to player B. For example, if player A chooses
first course of a action & the player B chooses fourth course of action,
then player B has to pay 1 unit to player A. Suppose, if player B
chooses third course of action instead of fourth, then player A has to
pay 2 units to player B.

The strategy of player is the decision rule he uses to decide which

course of action he should employ. This strategy may be of two kinds:
A pure strategy is a decision always to select the same course of
action. Mixed strategy is a decision to choose at least two of his
courses of action with fixed probabilities.

This value of a game is the expected gain of player A if both players

use their best strategies. The best strategy is based on minimax
criterion of optimality. This states that if a player lists the worst
possible outcome of all his strategies, he will choose the strategy,
 Game Theory 301

which corresponds to the best of the worst outcomes. This is

illustrated in the following example.

Example 1:

Using minimax criterion, find the best strategy for each player.

Player B Row Maximin

1 2 3 Minima (Best of
(worst for worst)
player A)
Player A 1 -1 5 -2 -2
2 1 2 4 1 1
3 -4 -6 12 -6
Column maxima 1 5 12
(worst for player B)
Minimax 1
(Best of worst)

Solution:

If player B chooses first strategy, the worst that can happen is that he
loses 1 unit. If he chooses second strategy, the worst that can happen
to him is that he loses 5 units. If he chooses third strategy, the worst
outcome to him is 12 units. So, the best out of the three outcomes is 1-
unit loss, i.e., the best strategy for him would be the first strategy that
would yield 1-unit loss ton him. To minimize the maximum loss
(minimax criteria), the player B has to choose first strategy always.

Similarly, if player A chooses first strategy, the maximum he would

loose is 2 units. If he chooses, second strategy, the worst that can
happen is that he wins 1 unit. If he chooses third strategy, he would
loose 6 units. So, the player A will always choose the second strategy
in order to minimize the maximum losses or maximizing the
minimum gain (i.e.,) 1 unit.

In the above table, the Minimax (minimum of maximum) for the

columns represents the minimum of maximum loss to player B.
302 Operations Research

Similarly, the Maximin (maximum of minimum) for the rows

represents the maximum of minimum gain to player A.

A saddle point of a payoff matrix is that position in the payoff matrix

where maximum of row minimum (Maximin) coincides with the
minimum of column maximum (Minimax). The element at the saddle
point is called the value of the game denoted by V.

The saddle point for the above problem is row 2 & column 1 i.e., (A2,
B1). The value of the game is 1. However, the saddle point need not
be unique. This is explained in the following example.

Example 2:

Determine the solution of the following game

-2 10 -2
-5 -6 -4
-5 25 -8

Solution:
Player B Row Maxi
1 2 3 Minima Min
1 -2 10 -2 -2 -2
2 -5 -6 -4 -6
3 -5 25 -8 -8
Player A
Column Maxima -2 25 -2
Mini Max -2 -2

In order to get maximum profit or minimize the loss, at any worst

situation the player A will always choose I strategy & the player B will
choose either I or III strategy or a mixed strategy which uses any
combination of these two strategies.

In this problem, the minimax is - 2 & the Maximin is -2. Hence, the
value of the game is -2. However, there are two saddle points;
1. (A1, B1) 2. (A1, B3)
 Game Theory 303

Let us denote maximin value by V& the minimax value by  & the
value of the game by V.

If V = 0 = , then the value of the game is 0 & the game is said to be

fair.

If V = v = then the game is said to be strictly determinable. For the

game which is strictly determinable, the value (a) of unknown
element must be such that V  a This is illustrated in the following
examples.

Example 3:

For the game with payoff matrix as below determine the best
strategies form player A & B & also the value of the game. Is this
game
i) fair ii) strictly determinable

3 0 1
4 -1 5
-3 -4 1
Solution:
Player B Row Maxi
1 2 3 Minima Min
1 3 0 1 0 0

Player A 2 4 -1 5 -1
3 -3 -4 1 -4
Column Maxima 4 0 5
Mini Max 0

Since Minimax () = Maximin (V)= 0, the value of the game (V)= 0.
The saddle point is (A1, B2).

Since V = 0, the game is said to be fair.

Since = V, (pure strategy), the game is said to be strictly
determinable.
304 Operations Research

Example 4:

For what value of , the game with the following matrix in strictly
determinable.
 5 3
-1  -6
-2 4 0
Solution:
Ignoring the value of ,
Player B Row Maxi
1 2 3 Minima Min
1  5 3 3 3
Player A 2 -1  -6 -6
3 -2 4 0 -2
Column Maxima -1 5 3
Mini Max -1
We know that minimax () = maximin (V) = value of the game(V),
when the game is strictly determinable.

Here, = -1, V = 3. So  must be such that     V

i.e., -1≤   3

Example 5 :

For the game with the following payoff

1 5
-3 a

i. show that the game is strictly determinable whatever 'a' may be

ii. find the value of the game

Solution:

Ignoring the value of ‘a’

 Game Theory 305

Player B Row Max

1 2 Minima Min
1 1 5 1 1
Player A
2 -3 a -3
Column Max 1 5
Min Max 1

Saddle point = (A1, B1)

Value of the game = 1

For this problem, even if the value of ‘a’ is more than 5 or less than -3,
the minimax & maximin will be 1 only. So, the value of will not affect
the value of the game. So, the game is strictly determinable for any
value of a.

Example 6:

Find the range of value of a & b that will make the payoff element
(A2, B2) a saddle point for the game whose pay off matrix is given
below
Player B
1 3 4
Player A 9 6 b
5 a 7

Solution

Ignoring a, b, we will determine the maximin & the Minimax values

of the payoff matrix.
306 Operations Research

Row Maxi
Player B Minima min
1 2 3
1 1 3 4 1
Player A 2 9 6 b 6 6
3 5 a 7 5
Column Maxima 9 6 7
Minimax 6

Since the saddle point is (A2, B2)

The value of the game is 6.

Therefore, maximin value must be 6 & minimax value must be 6.

This implies the condition on 'a' as a  6 & on 'b' as b ≥ 6

Therefore, required range is a  6  b

12.3 Dominance

We can reduce the size of a payoff matrix by eliminating a course of

action (strategy) which is so inferior to another as never to be used.
Such a course of action is said to be determined by the other. The
concept of dominance is especially useful for the evaluation of two
person zero - sum games where a saddle point does not exist. This is
illustrated in the following example.

Example 7:

Reduce the following game whose pay off matrix is given below

Player B
I II III IV V VI
1 3 2 -1 2 1 -4
2 3 3 1 3 2 2
Player A 3 3 4 2 4 -1 2
4 4 -2 1 0 -3 4
5 0 4 0 8 5 8
 Game Theory 307

Solution

Always you first look for a saddle point, if it exists in the given
problem. If the saddle point exists in the given matrix, then the given
matrix size need not be reduced to solve it. In our problem, there is
no saddle point, so, to solve the problem easily, the size of the matrix
is to be reduced.

In the given game, all the elements in the row 1 are equal to or less
than corresponding column elements in the row 2. So, the player A
will not choose strategy 1 at all. Because for any strategy of player B,
he will get less gain in strategy 1 compared to the gains in strategy 2.
Hence, row 2 dominates row 1. So, row 1 can be neglected. The
reduced matrix thus becomes as,
Player B
I II III IV V VI
2 3 3 1 3 2 2
Player A 3 3 4 2 4 -1 2
4 4 -2 1 0 -3 4
5 0 4 0 8 5 8

For the player B, the strategy III yields less loss, compared to strategy
I irrespective of the strategies followed by player A. So, player B will
not choose strategy I at all. So, column I can be neglected. Similarly,
column II dominates column IV. So, column IV also can be neglected.
Thus, the matrix is reduced as follows:

Player B
II III V VI
2 3 1 2 2
Player A 3 4 2 -1 2
4 -2 1 -3 4
5 4 0 5 8

In the above matrix, no single row/column dominates any other

row/column respectively.

Dominance need not be based on the superiority of pure strategies

only. A given strategy can be dominated if it is inferior to an average
308 Operations Research

of two more other pure strategies. Consider the above-reduced

matrix.

Average of A's third and fifth pure strategies gives as

 4 + 4 2 + 0 −1 + 5 2 + 8 
 2 , 2 , 2 , 2  = 4,1,2,5
This is obviously superior to A's second & fourth strategy. Therefore,
the second & fourth strategies can be neglected from the matrix. The
resulting matrix becomes
Player B
II III V VI
Player A 3 4 2 -1 2
5 4 0 5 8

Now the column III dominates column II & column VI dominates

column V so, column II & column V can be neglected. The matrix
becomes as,
Player B
III V
Player A 3 2 -1
5 0 5

Thus, at the maximum, the given payoff matrix is reduced to (2X2)

matrix form. Again, no pure strategy can be applied. There is no
saddle point still. So, any one of the methods described in section 12.4
and 12.5 is to be adopted to solve this mixed strategy-based game.

12.4 Mixed Strategy – Arithmetic Method

It provides an easy method for finding the optimum strategies for

each player in a 2X2 game. It consists of the following steps.
Subtract the two elements in column 1 & write it under column 2,
ignoring sign.
Subtract the two elements in column 2 & write it under column 1,
ignoring sign.
Subtract the two elements in row 1 & write it against row 2, ignoring
sign.
 Game Theory 309

Subtract the two elements in row 2 & write it against row 1, ignoring
sign.

These values are called oddments. They are the frequencies with
which the player must use their courses of action in their optimum
strategies. The arithmetic method can be illustrated with the reduced
matrix of example 7.

Example 8 :

Solve the example problem 7 by Arithmetic method and method of

dominance.

Solution

The reduced matrix using the method of dominance is

Player B
III V
Player A 3 2 -1
5 0 5

Using the steps for the arithmetic method as discussed above, the
oddments are written as

Player B
III V
Player A 2 -1 5
0 5 3
6 2
Write down the oddments in percentages as below

For column oddments

6 → {6/[6+2]} X 100 = 75% = 0.75
1 → {2/[6+2]} X 100 = 25% = 0.25

For row oddments,

5 → {5/[5+3]} X 100 = 62.5% = 0.625
3 → {3/[5+3]} X 100 = 37.5% = 0.375
310 Operations Research

Thus, the matrix can be written as

Player B
III V
3 2 -1 5 5/8 0.625
Player A 5 0 5 3 3/8 0.375
6 2
6/8 2/8
0.75 0.25

Thus for optimum gains, player A should use strategy 3 for 62.5% of
the time & strategy 5 for 37.5% of the time, while player B should use
strategy III for 75% of the time & strategy V for 25% of the time.

To obtain the value of the game, any one of the following expressions
may be used.

Using A's oddments,

If Player B plays strategy III then, the value of the game,
V =2(5/8) + 0(3/8) =10/8 = 1.25
or V = (2 x 0.625) + (0 x 0.375) = 1.25

If player B plays strategy V, then the value of the game,

V = -1(5/8) + 5(3/8) =10/8 = 1.25

Using B's oddments

If player A strategy 3, then the value of the game,
V = 2(6/8) + -1 (2/8) = 10/8 = 1.25

If player A plays strategy 5, then the value of the game,

V = 0(6/8) + 5(2/8) =10/8 = 1.25

Thus, the value of the game is found to be same computing for any
player and for combination of any strategies of the player. Thus, the
full solution of the game is

Optimal strategy for player A (0, 0, 5/8, 0, 3/8)

Optimal strategy for player B (0, 0, 6/8, 0, 2/8)
Value of the game, V = 10/8 or 1.25
 Game Theory 311

12.5 Mixed Strategy – Algebraic Method

Example 9 :

Find the optimal mixed strategies & value of the game for the
following 2x2 two-person zero - sum game without any saddle point.

Player B
B1 B2
Player A A1 a11 a12
A2 a21 a22

Solution

Let X and (1 - X) be the probabilities of selecting strategies A1 & A2 by

player A and

Let y and (1 - Y) be the probabilities of selecting strategies B1 and B2 by

player B

The expected value of the game to player is given as

E (x, y) = x y a11 + (1 - x) y a21 + x (1-y) a12 + (1 - x) (1 - y) a22

To find optimum values of X & Y, differentiate E(X, Y) w.r.t x & y

E(x, y) = x y a11 + y a21 - x y a21 + x a12 - x y a12 + a22 +x y a22

- x a22 - y a22

E(x, y ) = xa11 + a21 − xa21 − xa12 + xa22 − a22
y


E(x, y ) = ya11 − ya21 + a12 − ya12 + ya22 − a22
x


E( x, y) = 0 and  E( x, y) = 0 , we get
x y
312 Operations Research

ya11 − ya 21 + a12 − ya12 + ya 22 − a 22 = 0

y (a11 − a 21 − a12 + a 22 ) = a 22 − a12

a 22 − a12
y=
a11 − a 21 − a12 + a 22
a 22 − a12
y=
a11 + a 22 − (a12 + a 21 )

xa11 + a21 − xa21 − xa12 + xa22 − a22 = 0

x(a11 − a21 − a12 + a22 ) = a22 − a21

a22 − a21
x=
a11 − a21 − a12 + a22
a22 − a21
x=
a11 + a22 − (a12 + a21)

Substituting the value of x & y in E(x, y), the value of the game is
a11a22 − a12a21
V = E ( x, y) =
(a11 + a22 ) − (a12 + a21 )

Example 10 :

Solve the example problem 7 (reduced matrix) by Algebraic method

Solution

Player B
III V
Player A 3 2 -1
5 0 5

a11 = 2; a12 = -1; a21 = 0; a22 = 5

 Game Theory 313

The optimum mixed strategies:

for player A: for player B:

1 2 3 4 5 I II III IV V

[0 0 x 0 (1-x)] [0 0 y 0 (1-y)]

it is known that
a 22 − a 21
X =
a11 + a 22 − (a12 + a 21 )

substituting the values of the matrix elements, we get

5−0 5
X = =
2 + 5 − (− 1 + 0) 8
a 22 − a 12
and y =
a11 + a 22 − (a12 + a 21 )
5 − (−1) 6
= =
2 + 5 − (− 1 + 0) 8
5
and (1 − x) = 1 − = 3 / 8
8
6 2
(1 − y ) = 1 − =
8 8
the value of the game is given by
a11a 22 − a12 a 21
v = E ( x, y ) =
(a11 + a 22 ) − (a12 + a 21 )
2  5 − (− 1) 0 10
= =
(2  5) − (− 1 + 0) 8
thus, the full solution of the game is

Optimal strategy for player A is given by

1 2 3 4 5
(0, 0, x, 0, (1-x) )
(0, 0, 5/8, 0, 3/8 )
314 Operations Research

Optimal strategy for player B

I II III IV V
(0, 0, y, 0, (1-y))
(0, 0, 6/8, 0, 2/8 )
10
value of the game, v=
8
Note: The above answers are similar to the answers obtained through
arithmetic method in example 8.

12.6 Mixed strategy – Matrix oddment method for n x n games

This method is also known as Method of matrices for n x n games. It

consists of following steps:

1. Obtain a new matrix C, whose first column is obtained from the

given payoff matrix by subtracting 2nd column from 1st column;
the second column is obtained by subtracting 3rd column from 2nd
column and so on. Thus C is a matrix of nXn-1

2. Obtain a new matrix R, whose first row is obtained from the given
payoff matrix by subtracting 2nd row from 1st row; the second
row is obtained by subtracting 3rd row from 2nd row; and so on.
Thus R is a matrix of (m - 1) X m.

3. Determine the magnitude of oddments corresponding to each row

and each column of given payoff matrix. The oddments
corresponding to i row of given payoff matrix is obtained from C
th

as the determinant of C i.e., l Cj l by deleting ith row.

4. Determine the oddments corresponding to jth column of given

payoff matrix. It is obtained from R as l Rj l by deleting jth column.

5. Write the magnitude of oddments against their respective rows &

columns, ignoring the sign.

6. Check whether the sum of row oddments is equal to the sum of

column oddments. If so, the oddments expressed ass fractions of
 Game Theory 315

the grand total yields the optimum strategies. If not, this method
fails.

7. Calculate the value of the game, taking row elements for any
strategy of the player B or taking row elements for any strategy of
the player B or taking column elements for any strategy of the
player A.

This method is illustrated in the following example.

Example 11:

Solve the game whose payoff matrix is given by matrix oddment

method
Player B
I II III
1 1 -1 -1
Player A
2 -1 -1 3
3 -1 2 -1

Note : First you look for a saddle point in the problem. If the saddle
point does not exist, then you try to apply the method of dominance
to reduce the size of the matrix. Then you apply the method
prescribed in the question.

In the above matrix, there is no saddle point. The method of

dominance also is not able to reduce the matrix. So, we will directly
apply the method of oddment to solve the game.

Find out the matrix C & matrix R by the steps described earlier.

C= 1-(-1) -1-(-1)
-1-(-1) -1-3
-1-2 2-(-1)

= 2 0
0 -4
-3 3
316 Operations Research

R= 1-(-1) -1-(-1) -1-3

-1-(-10 -1-2 3-(-1)

= 2 0 -4
0 -3 4

C1 = 0 -4
-3 3

= 0 – (12) = -12

C2 = 2 0
-3 3

=6–0=6

C3 = 2 0
0 -4

= -8 – 0 = -8

R1 = 0 -4
-3 4

= 0 –12 = -12

R2 = 2 -4
0 4

=8–0=8

R3 = 2 0
0 -3

= -6 – 0 = -6
 Game Theory 317

the augmented payoff matrix, therefore is written as

Column
oddments
1 -1 -1 12
-1 -1 3 6
-1 2 -1 8

Row oddments 12 8 6 26

Optimum strategies are:

1 2 3
 12 6 8 
for playerA:  , , 
 26 26 26 
6 3 4
 , , 
 13 13 13 

I II III
 12 8 6 
for playerB :  , , 
 26 26 26 
6 4 3
 , , 
 13 13 13 

Value of the game: taking column 1 (strategy I of player B)

 12   6   8 
V = 1   +  − 1   +  − 1  
 26   26   26 
12 6 8
= − −
26 26 26
− 2 −1
= =
26 13
318 Operations Research

12. 7 Graphical method for 2 x n or m x 2 games

Graphic method is applicable to only those games in which one of the

players has only two strategies.

Consider the following 2 x n game

Player y
Y1 Y2 Yn
Player X X1 a11 a12 ….. a1n
X2 a21 a22 ….. a2n

It is assumed that the game has no saddle point.

Player X has two strategies x1 and x2 (= 1- x1) and player y has n

strategies.

Expected payoffs for player X corresponding to pure strategies of

player Y are given below.
Player Y’s strategy Player X’s expected payoffs
Y1 a11 X1 + a21X2 = a11 X1 + a21(1- X1)
= (a11 – a21) X1 + a21
Y2 a12 X1 + a22 X2 = a12 X1 + a22(1- X1)
= (a12 – a22) X1 + a22
. .
. .
yn (a1n– a2n) X1 + a2n

Player X’s expected payoff varies linearly with x 1. So, player X should
select that the value of x1 which maximizes the minimum expected
payoff. This can be done by plotting the above lines as a function of
x1. The lower of these lines (shaded) gives the minimum expected
payoff to players X as a function of) X1. The highest point on this
lower boundary (X1*) is the maximum of minimum expected payoff to
player X. The point is called as maximin point.
 Game Theory 319

This point indicates the level at which player X can hold player Y to
minimize his loss, otherwise the level at which player Y can hold
player x’s winnings.
Similarly, we can solve (m x 2) games. For the (m x 2) games, we
should get minimax point, which will be the lowest point on the
upper boundary.

This is illustrated in the following example.

Example 12 :

Solve the game by graphical method.

Player B
I II III IV V
1 -5 2 -4 -2 1
Player A 2 1 -5 -3 -4 -2

Solution

The saddle if any is always checked first, before applying the method
prescribed in the question.

Since minimax is not equal to maximin in this problem, there is no

saddle point existing for this problem. Hence we will apply graphical
method now, taking A1 and A2 in vertical axes with appropriate scale
covering all the elements given in the matrix.

Now show all the strategies of B by means of lines connecting the

corresponding elements. It is time to find out the Maximin value for
player A as player A has got only two strategies. Find out the
common minimum area covered by all lines first. Then find out the
maximum point of the common minimum (shaded) area, which is
considered as Maximin.
320 Operations Research

In this problem, the maximin is found at the intersection of B 3 and B4.

Hence the matrix is reduced into (taking only the strategy lines
passing through the intersection point),

B3 B4
A1 -4 -2
A2 -3 -4

Applying arithmetic method now, we get

B3 B4
A1 -4 -2 1 1/3
A2 -3 -4 2 2/3
2 1
2/3 1/3
 Game Theory 321

Value of the game is calculated as : (taking for B3)

= -4(1/3) + -3(2/3)= -10/3
1 2
Optimum strategies for player A :  , 
3 3
 2 1 
Optimum strategies for player B :  0,0, , ,0 
 3 3 

Example 13 :

Solve the game by graphical method.

Player B
I II
1 6 5
Player A 2 3 6
3 4 4
4 8 1
5 1 4

Solution

The saddle if any is always checked first, before applying the method
prescribed in the question.

Since minimax is not equal to maximin in this problem, there is no

saddle point existing for this problem. Hence we will apply graphical
method now, taking B1 and B2 in vertical axes with appropriate scale
covering all the elements given in the matrix.

Now show all the strategies of A by means of lines connecting the

corresponding elements. It is time to find out the Minimax value for
player B as player B has got only two strategies. Find out the common
maximum area covered by all lines first. Then find out the minimum
point of the common maximum (shaded) area, which is considered as
Minimax.
322 Operations Research

In this problem, the minimax is found at the intersection of A 1 and A2.

Hence the matrix is reduced into
B1 B2
A1 6 5
A2 3 6

Applying arithmetic method now, we get

B1 B2
A1 6 5 3 3/4
A2 3 6 1 1/4
1 3
1/4 3/4
 Game Theory 323

Value of the game is calculated as : (taking for B1)

= 6(3/4) + 3(1/4)= 21/4
3 1 
Optimum strategies for player A :  , ,0,0,0 
4 4 
1 3
Optimum strategies for player B :  , 
4 4

Example 14 :

Solve the game by graphical method.

B1 B2 B3
A1 3 4 0
A2 4 3 7
Solution:

The saddle if any is checked first always, before applying the method
asked in the question.

Since minimax is not equal to maximin in this problem, there is no

saddle point existing for this problem. Hence we will apply graphical
method now, taking A1 and A2 in vertical axes with appropriate scale
covering all the elements given in the matrix.

Now show all the strategies of B by means of lines connecting the

corresponding elements. It is time to find out the Maximin value for
player A as player A has got only two strategies. Find out the
common minimum area covered by all lines first. Then find out the
maximum point of the common minimum (shaded) area, which is
considered as Maximin.
324 Operations Research

All the three lines are passing through the maximum point, so, we
cannot straight away neglect any one of the B’s strategies to make it
into 2 x 2 matrix for getting it solved easily through
arithmetic/algebraic method.

A’s expected payoffs corresponding to B’s pure strategies are

B’s pure A’s expected payoffs Slopes
strategies (from y = mx +c)
B1 3x1 + 4(1-x1) = -x1 + 4 m = -1
B2 4x1 + 3(1-x1) = x1 + 3 m=1
B3 0x1 + 7(1-x1) = -7x1 + 7 m = -7

As B1 &B3 have the same sign for their slopes, the combination of B 1 &
B3 should not be used. Ignoring one of them, we can get the game
reduced to two (2x2) game which can be easily solved, by any method
(say arithmetic method).
 Game Theory 325

Applying arithmetic method keeping B1 and B3, we get

B1 B2
A1 3 4 1 ½
A2 4 3 1 ½
1 1
½ ½
Solution:
1 1
A’s optimum strategy :  , 
2 2
1 1 
B’s optimum strategy :  , ,0 
2 2 
Value of the game = 3(1/2) + 4(1/2) = 7/2

Similarly applying arithmetic method keeping B2 and B3, we get

B2 B3
A1 4 0 4 ½

A2 3 7 4 ½
7 7/8
7/8 1/8

Solution (alternative) :
1 1
A’s optimum strategy:  , 
2 2
 7 1
B’s optimum strategy:  0, , 
 8 8
Value of the game = 4(1/2) + 3 (1/2) = 7/2

Choosing any one of the strategies among B1 and B3, we get same
value of the game, but with different probability for the chosen B
strategies.