CONTENTS

INTRODUCTION ..................................................................... 1
PRELIMINARY ....................................................................... 3

1 TWO PERSON ZERO SUM GAMES 5

1.1 Zero Sum Game ............................................................... 5
1.2 Two Person Zero Sum Games ......................................... 6
1.3 Payoff Matrix .................................................................... 7
1.4 Pure Strategy.................................................................... 9
1.5 Maximin principle and Minimax principle ................... 10
1.6 Games with Saddle point ............................................... 11

2 MIXED STRATEGIES: GAME WITHOUT SADDLE POINT 14

2.1 Mixed Strategy .............................................................. 14
2.2 Games without Saddle Point ......................................... 17
2.2.1 Solution of 2 × 2 Game Without a Saddle point 18
 2.3 Methods to solve games without saddle point .............. 21
2.3.1 Graphical Method ............................................. 21
2.3.2 Linear Programming Method ........................... 26

3 THE RULES OF DOMINANCE 30

3.1 The Rules of Dominance ............................................... 30
CONCLUSION .............................................................................. 35
BIBLIOGRAPHY.............................................................................36
 INTRODUCTION
A theory of games provides a series of mathematical models that

may be useful in explaining interactive decision making concepts.

The competitors are referred to as players. A player may be an indi-
vidual group of individuals or an organization.
This project deals with:
1. Understanding the principles of zero-sum ,two person games
and applying various methods to select and execute various op-
timal strategies to win the game.

2. Use dominance rules to reduce the size of a game payoff ma- trix
and compute value of the game with mixed strategies.Using
linear programming approach to compute the value of the game
when dominance rules does not apply.

3. Applying minimax and maximin principle to compute the value

of the game,when there is a saddle point and using graphical

1
method and L.P method for games without saddle point.

2
3
 Chapter 3

THE RULES OF DOMINANCE

3.1 The Rules of Dominance

The rules of dominance are used to reduce the size of the payoff ma-
trix. These rules help in deleting certain rows and/or columns of the
payoff matrix that are inferior (less attractive) to at least one of the
remaining rows and/or columns (strategies), in terms of payoffs to
both the players. Rows and/or columns once deleted can never be used
for determining the optimum strategy for both the players.

The rules of dominance are especially used for the evaluation of

two-person zero-sum games without a saddle (equilibrium) point.
Certain dominance principles are stated as follows:

33
 CONCLUSION

The theory of games provide mathematical models that are useful in

explaining interactive decision-making process, where two or more
competitors (players) are involved under condition of conflict and
competition. The optimal strategies in game theory helps in formu-
lating a conflicting and competitive environment. As winning of the
game is the goal, applying various methods to select and execute
various optimal strategies is the key to win the game. The player must
be able to make distinction between pure and mixed strategies. Using
dominance rules can help to reduce the size of the payoff ma- trix and
compute value of the game with mixed strategies. Linear
programming can be used to determine the value of the game when
dominance rules do not apply. Thus game theory provides technique
to determine the best strategies given a competitive environment.

34
 Bibliography

[1] J K Sharma ,Operations Research:Theory and applications

4th edition,Laxmi Publications(Delhi),2009

[2] J K Sharma ,Operations Research:Theory and applications

6th edition,Laxmi Publications(Delhi),2016

[3] Rama P. Rama Murthy, Operations Research, second edition,

New Age International Publishers (New Delhi), 2007

[4] Shaun P. Hargreaves - Heap, Yanis Varoufakis, Game Theory:

A Critical Introduction, Published by Taylor and Francis
(2004),(New fetter lane, London), 1995

35
 PRELIMINARY
Now let us see some basic definitions:
Dot Product: Let 𝑥, 𝑦 ∈ ℝ𝑛 be two vectors.If 𝑥 = (𝑥1 , 𝑥2 , ..., 𝑥𝑛 )
and 𝑦 = (𝑦1 , 𝑦2 , ..., 𝑦𝑛 ). Then the dot product of these vectors is
𝑥 ⋅ 𝑦 = 𝑥1 𝑦1 + 𝑥2 𝑦2 + ... + 𝑥𝑛 𝑦𝑛 .Do the case when x and y are column
or row vectors in the obvious way.
Matrix Addition: If 𝐴 and 𝐵 are both in ℝ𝑚∗𝑛 ,then 𝐶 = 𝐴 + 𝐵 is
the matrix sum of 𝐴 and 𝐵 and 𝐶𝑖𝑗 = 𝐴𝑖𝑗 + 𝐵𝑖𝑗 for 𝑖 = 1, 2, ...𝑚 and
𝑗 = 1, 2, ...𝑛.
Row/Column Vector: A 1 × 𝑛 matrix is called a row vector, and a
𝑚 × 1 matrix is called a column vector.It should be clear that any
row of matrix 𝐴 could be considered as a row vector in ℝ𝑛 and any
column of 𝐴 could be considered as a column vector in ℝ𝑚 .
Matrix Multiplication: If 𝐴 ∈ ℝ𝑚∗𝑛 and 𝐵 ∈ ℝ𝑛∗𝑝1 then 𝐶 = 𝐴𝐵
is the matrix product of 𝐴 and 𝐵 and 𝐶𝑖𝑗 = 𝐴𝑖 ⋅ 𝐵𝑗
Matrix Transpose: If 𝐴 ∈ ℝ𝑚∗𝑛 is a 𝑚 × 𝑛 matrix then the transpose
of 𝐴 denoted by 𝐴𝑇 is an 𝑚 × 𝑛 matrix defined as:𝐴𝑇𝑖𝑗 = 𝐴𝑗𝑖

3
 ⎡1 0 ... 0⎤
⎢ ⎥
⎢0 1 ... 0⎥
⎢ ⎥
⎢. . ... . ⎥
Identity Matrix: The 𝑛 × 𝑛 identity matrix is 𝐼𝑛 =⎢ ⎥
⎢. . ... . ⎥
⎢ ⎥
⎢. . ... . ⎥⎥
⎢
⎢ ⎥
⎣0 0 ... 1⎦
Standard Basic Vector: The standard basic vector 𝑒𝑖 ∈ ℝ𝑛 is :
𝑒𝑖 = (0, 0, ...1, 0, ...0). This definition is only valid for n≥i. Fur-
ther the standard basis vector 𝑒𝑖 is also the 𝑖𝑡ℎ row or column of 𝐼𝑛 .
Unit and Zero Vector: The vector 𝑒 ∈ ℝ𝑛 is the one vector 𝑒 =
(0, 0, ...0) ∈ ℝ𝑛 . Similarly the zero vector 0 = (0, 0, ...0) ∈ ℝ𝑛 .
Saddle Point: A saddle point of a matrix is the position of such an
element in the playoff matrix which is minimum in its row and the
maximum in its column.

4
 Chapter 1

TWO PERSON ZERO SUM GAMES

1.1 Zero Sum Game

A zero sum is a situation where if one partly loses, the other partly
wins and the net change in wealth is zero. These games can include
just two players to millions of participants. The game of matching
pennies is often with as an example of a zero sum game according
to game according to game theory. The game involves two players
A and B simultaneously placing a penny on the table. The payoffs
depends on whether the pennies match or not. If both pennies are
heads or tails, player A wins and keeps player B’s penny. If they do
not match than player B wins and keeps player A’s penny. Matching

5
pennies is a zero sum game because one players gain is the others
loss.

1.2 Two Person Zero Sum Games

A game with only two players, say player A and player B is called
a Two person zero sum game, only if one player’s gain is equal to
the loss of other player in a way that the total sum is zero. A saddle
point is an element of the matrix that is both the smallest element in
its row and largest element in its column. If the losses of one player
are equivalent to the gains of the other so that the sum of their net
gains is zero. This game also known as rectangular game.

The characteristics of such games are

• Only two players participate

• Each player has finite number of strategies to use

• Each specific strategy results in payoffs

• Total payoff to the two player at the end of each play is zero

Example 1. Assume we are playing ROCK-SCISSORS-PAPER for

6
one dollar. Then the payoff matrix is

Rock Scissors Paper

Rock 0 1 -1
Scissors -1 0 1
Paper 1 -1 0

The first cell says “0”, which stands for “0,0” a payoff of 0 for
both players. The second cell entry of "1" should be read as "1,-1",
a payoff of 1 for Ann which has to be paid by Beth, therefore a payoff
of -1 for Beth.

1.3 Payoff Matrix

The payoffs (a quantitative measure of satisfaction that a player gets

at the end of the play) in terms of gain or losses, when players select
their particular strategies can be represented in the form of a matri-
ces, called the payoff matrix. since the game is zero sum, the gain of
one player is equal to the loss of other and vice versa. In other words
one players payoff table would contain the same amounts in payoff
table of other player with the sign changed. Thus, it is sufficient to
construct a playoff table only for one of the players.

7
 If player A has m strategies represented by the letters 𝐴1 , 𝐴2 , ⋯ 𝐴𝑚
and a player B has n strategies represented by the letters 𝐵1 , 𝐵2 , ⋯ 𝐵𝑛 .
The numbers m and n need not be equal. The total number of possi-
ble outcomes is therefore 𝑚 × 𝑛. Here it is assumed that each player
not only knows his own list of possible outcomes of action but also
of his opponent. For convenience, it is assumed that player A is al-
ways a gains whereas B a loser. Let 𝑎𝑖𝑗 be the payoff that the player
A gains from player B, if player A chooses strategy i and player B
chooses strategy j. Then the playoff matrix is shown in table
Player B strategies
Player A strategy 𝐵1 𝐵2 ⋯ 𝐵𝑛
𝐴1 𝑎11 𝑎12 ⋯ 𝑎1𝑛
𝐴2 𝑎21 𝑎22 ⋯ 𝑎2𝑛
⋮ ⋮
𝐴𝑚 𝑎𝑚1 𝑎𝑚2 ⋯ 𝑎𝑚𝑛

By convention the rows of the payoff matrix denote the player

A’s strategies and the column denote the player B’s strategies. Since
player A is assumed to be always be the gainer be therefore wishes
to gain as large as payoff 𝑎𝑖𝑗 as possible player B on the other herd

8
would do his best to reach as small a value of 𝑎𝑖𝑗 as possible. Of
course, the gain to player B and loss to A must be −𝑎𝑖𝑗 . The entries
in the payoff matrix can be in any units as long as they represent the
utility of the player. Also the matrix for player two is the negative of
the matrix for player one in as zero sum game.
Payoff; The outcome of playing a game is called Payoff.

1.4 Pure Strategy

Definition: It is a decision, in advance of all plays always to choose

a particular course of action.
For example, Let’s how can list, count, and encode the pure strat-
egy’s a player has in a game. Since in principle in an information
set every choice of move is possible, the product of the number of
choices taken overall information sets of that player is the number
of a player’s pure strategies. We encode a player’s strategies by first
numbering all his or her information sets arbitrarily. We abbreviate
each choice by a single letter. Then a pure strategy can be encoded as
an n-letter word, where n is the number of the player’s information
sets and the 𝑘𝑡 ℎ letter of the word tells what alternative the player
chooses at the 𝑘𝑡 ℎ information set. If for Ann has three information

9
sets, the first and the last with two possible moves, say K,L and L,M
and the second with three possible moves, say M,O and P, then Ann
has the 2.3.2 = 12 pure strategies encoded as KML, KMM, KOL,
KOM, KPL, KPM, LML, LMM, LOL, LOM, LPL, LPM.

1.5 Maximin principle and Minimax principle

Maximin principle: For player A the minimum value in each row

represents the least gain (payoff) to him, if he chooses his particular
strategy. These are written in the matrix by row minima. He will then
select the strategy that gives the largest gain among the row minimum
values. This choice of player A is called the maximin principle, and
the corresponding gain is called the maximin value of the game.
Minimax principle: For player B, who is assumed to be the looser,
the maximum value in each column represents the maximum loss to
him, if he chooses his particular strategy. There are written in the
payoff matrix by column maxima. He will then select the strategy
that gives the minimum loss among the column maximum values.
This choice of player B is called the minimax principle, and the cor-
responding loss is the minimax value of the game.
Optimal strategy: The course of action which maximizes the profit

10
of a player or minimizes his loss is called optimal strategy.
Value of the game: It refers to the expected outcome per play when
players follows their optimal strategy.

1.6 Games with Saddle point

Saddle point: The outcome is a saddle point if it is the minimum in

its row and maximum in its column.
Rules to determine Saddle point

1. Select the minimum (lowest) element in each row of the payoff

matrix and write them under ’row minima’ heading. Then select
the largest element among these elements and enclose it in a
rectangle ▭.

2. Select the maximum (largest) element in each column of the

payoff matrix and write them under ’column maxima’ heading.
Then select the lowest element among these elements and en-
close it in a circle ○.

3. Find out the elements which is same in the circle as well as

rectangle and mark the position of such elements in the matrix.

11
 This element represents the value of the game and is called the
Saddle point.

Example 2. For the game with payoff matrix

Player B
Player A -1 2 -2
6 4 -6
Determine the best strategies for players A and B and also the value
of the game. Is this game (i) fair (ii) strictly determinable?

player B
BI BII BIII Row minimum
Player A AI -1 2 −2 −2

AII 6 4 -6 -6

−2
Column maximum 6 4

In this example, gains to player A or losses to player B are repre-

sented by the positive quantities, whereas, losses to A and gain to B
are negative quantities.
Select minimum from the column maximum values.

12
that is minimax = 1, (marked as circle)
select maximum from the row minimum values,
that is maximin = 1, (marked as rectangle) Player A will choose
strategy I, which yields the maximum payoff of −2
Player B will choose strategy III
The value of game is −2, which indicates that player A will gain −2
unit and player B will loss −2 unit.
Since the maximum value = the minimax value = −2
∴ the game has saddle point and game is not fair game (since values
of the game is non zero)
Also maximin = minimax = value of game
∴ the game is strictly determinable.
It is a pure strategy game and the saddle point is (A-I, b-II)
The optimal strategies for both players given by pure strategy player
A must select strategy II and player B must select strategy III.

13
 Chapter 2

MIXED STRATEGIES: GAME

WITHOUT SADDLE POINT

2.1 Mixed Strategy

The course of action that are to be selected on a particular occasion

with some fixed probability are called Mixed Strategies. That is,
there is a probabilistic situation and objective of the players is to max-
imize expected gain or minimize expected losses by making choice
among pure strategy with fixed probabilities.
The value of game obtained by the use of mixed strategies rep-
resents the least payoff which player A can expect to win and the
least payoff which player B can expect to lose. The expected payoff
to a player in a game with payoff matrix [𝑎𝑖𝑗 ]𝑚×𝑛 can be defined as

14
𝐸(𝑝, 𝑞) =
∑ ∑
𝑚 𝑛
𝑝𝑖 𝑎𝑖𝑗 𝑞𝑗 = 𝑃 𝑇 𝐴𝑄
𝑖=1 𝑗=1

where 𝑝 = (𝑝1 , 𝑝2 , ⋯ 𝑝𝑚 ) and 𝑞 = (𝑞1 , 𝑞2 , ⋯ 𝑞𝑛 ) denote probabilities

or relative frequency with which a strategy is chosen from the list of
strategies associated with m strategies of player A and n strategies of
player B respectively. Obviously, 𝑝𝑖 ≥ 0(𝑖 = 1, 2, ⋯ 𝑚),𝑞𝑗 ≥ 0(𝑗 =
1, 2, ⋯ 𝑛) and 𝑝1 + 𝑃2 + ⋯ + 𝑝𝑚 = 1, 𝑞1 + 𝑞2 + ⋯ + 𝑞𝑛 = 1.

An example for a mixed strategy in Rock-Scissors-Paper is to play

rock, scissors, or paper with probabilities 50%, 25%, or 25%. Be-
fore the game is played, the player decides randomly, based on these
probabilities, which strategy to play. We can view pure strategies as
mixed strategies where we play one of the options with probability
100% and the others with probability 0%.
If we add the mixed strategy (50% rock, 25% scissors, 25% paper)
as an option for Ann in Rock-scissors-paper then the expected pay-
off’s for Ann against Beth’s pure strategies rock, scissors, paper, are
0, 0.25, −0.25 respectively see the table

15
 Rock scissors Paper
Rock 0 1 -1
Scissors -1 0 1
Paper 1 -1 0
50-25-25 mix 0 0.25 -0.25

If Ann plays the mixed strategy and Beth plays rock then with 50%
probability there will be a tie (rock versus rock), with 25% probability
Beth will win (Beth’s rock against Ann’s Scissors) and 25% proba-
bility Ann will win (paper against Beth’s rock). Thus the expected
payoff for Ann when playing the mixed strategy against Beth’s rock
is 50%.0 + 25%.(−1) + 25%.1 = 0. For instance, the second value
in the fourth row is 50%.1 + 25%.0 + 25%.(−1) = 0.25 and the third
50%.(−1) + 25%.1 + 25%.0 = −0.25
Though the mixed strategy does not dominate any of the pure strate-
gies, it may be attractive to a player aiming at the maximum strategy
since it guarantees a payoff of −0.25 compared to −1 in the other
cases. Beth is also entitled to mixed strategies. We assume that she
choose a mix of 25% rock, 50% scissors, 25% paper. Adding this mix
as another one of Beth’s options we obtain the bimatrix.

16
 Rock scissors Paper 25-50-25 mix
Rock 0 1 -1 0.25
Scissors -1 0 1 0
Paper 1 -1 0 -0.25
50-25-25 mix 0 0.25 -0.25 0.0625

The new values are computed as before, as expected values, us-

ing the payoffs of the same row, weighted by the probability of the
mix. For instance, the last entry in the first row is computed as
25%.0 + 50%.1 + 25%.(−1) = 0.25
Even the payoff for 50-25-25 mix against 25-50-25 mix is computed
this way, using the fourth row values, as 25%.0+50%.0.25+25%.(−0.25) =
0.0625. Not too surprisingly, Ann’s mix with its Emphasis on rock,
beats Beth’s mix, which is heavy on scissors.

2.2 Games without Saddle Point

Suppose if there is no pure strategy solution for a game, then there

is no saddle point exist. In these situations, to solve the game both
the player must determine the optimal mixtures of strategies to find

17
saddle point.

2.2.1 Solution of 2 × 2 Game Without a Saddle point

A rectangular game without a saddle point cannot be solved using

pure strategies or with minimax and maximin criterion of optimality.
So, in this strategy the probability with which each action should be
selected is calculated.

Theorem 1. For any 2 × 2 two person zero sum game without any
saddle point having the payoff matrix for player A given by

𝑃 𝑙𝑎𝑦𝑒𝑟𝐵
⎡ 𝐵 𝐵 ⎤
𝐴1 ⎢
1 2 ⎥

𝑃 𝑙𝑎𝑦𝑒𝑟𝐴 ⎢𝑎 𝑎 ⎥
𝐴2 ⎢
11 12 ⎥
⎢𝑎 𝑎 ⎥
⎣ 21 22 ⎦
⎡𝐴 𝐴 ⎤ ⎡𝐵 𝐵 ⎤
the optimal mixed strategies 𝑆𝐴 = ⎢ ⎥ and 𝑆𝐵 = ⎢ 1 2 ⎥
1 2
⎢𝑝 𝑝 ⎥ ⎢𝑞 𝑞 ⎥
⎣ 1 2⎦ ⎣ 1 2⎦
𝑝1 𝑎22 −𝑎21 𝑞1 𝑎22 −𝑎12
are determined by 𝑝 = 𝑎 −𝑎 , 𝑞 = 𝑎 −𝑎 where 𝑝1 + 𝑝2 = 1 and
2 11 12 2 11 21

𝑞1 + 𝑞2 = 1. The value ’V’ of the game to A is given by

18
 𝑎11 𝑎22 −𝑎21 𝑎22
𝑉 = 𝑎11 +𝑎22 −(𝑎12 +𝑎21 )

Proof. Let 𝑝1 and 𝑝2 be the probability for player A. Let 𝑞1 and 𝑞2 be

the probability for player B.
Let the optimal strategy be 𝑆𝐴 for player A and 𝑆𝐵 for player B. Let
⎡𝐴 𝐴 ⎤
a mixed strategy for player A be given by 𝑆𝐴 = ⎢
1 2⎥
where
⎢𝑝 𝑝 ⎥
⎣ 1 2⎦
𝑝1 + 𝑝2 = 1. Thus if player B moves 𝐵1 then the net excepted gain of
A will be, 𝐸1 (𝑝) = 𝑎11 𝑝1 + 𝑎21 𝑝2 and if B moves 𝐵2 , the net excepted
gain of A will be 𝐸2 (𝑝) = 𝑎12 𝑝1 + 𝑎22 𝑝2 .
⎡𝐵 𝐵 ⎤
Similarly, if B plays his mixed strategy 𝑆𝐵 = ⎢
1 2⎥
where 𝑞1 +
⎢𝑞 𝑞 ⎥
⎣ 1 2⎦
𝑞2 = 1. Then B’s net excepted loss will be, 𝐸1 (𝑞) = 𝑎11 𝑞1 + 𝑎12 𝑞2
and if A plays 𝐴1 , and 𝐸2 (𝑞) = 𝑎21 𝑞1 + 𝑎22 𝑞2 if A plays 𝐴2
The excepted gain of player A, when B chooses his moves with prob-
abilities 𝑞1 and 𝑞2 , is given by
𝐸(𝑝, 𝑞) = 𝑞1 [𝑎11 𝑝1 + 𝑎21 𝑝2 ] + 𝑞2 [𝑎12 𝑝1 + 𝑎22 𝑝2 ]
Player A would always try to mix his moves with such probabilities
so as to maximize his excepted gain.

19
Now

𝐸(𝑝, 𝑞) = 𝑞1 [𝑎11 𝑝1 + 𝑎21 (1 − 𝑝1 )] + (1 − 𝑞1 )[𝑎12 𝑝1 + 𝑎22 (1 − 𝑝1 )]

= [𝑎11 + 𝑎22 − (𝑎12 + 𝑎21 )]𝑝1 𝑞1 + (𝑎12 − 𝑎22 )𝑝1 + (𝑎21 − 𝑎22 𝑞1 + 𝑎22 )
𝑎 − 𝑎21 𝑎 − 𝑎12 𝑎 𝑎 − 𝑎12 𝑎21
= 𝜆(𝑝1 − 22 )(𝑞1 − 22 ) + 11 22
𝜆 𝜆 𝜆
where 𝜆 = 𝑎11 + 𝑎22 − (𝑎12 + 𝑎21 )
𝑎22 −𝑎21
We see that if A chooses 𝑝1 = 𝜆
he ensures an excepted gain of
(𝑎11 𝑎22 −𝑎12 𝑎21 )
at least 𝜆
𝑎22 −𝑎12
Similarly if B chooses 𝑞1 = 𝜆
, then he can limit his excepted
(𝑎11 𝑎22 −𝑎12 𝑎21 )
loss to atmost 𝜆

These choices of 𝑝1 and 𝑞1 will thus be optimal to the two players.

Thus,
𝑎22 −𝑎21 𝑎22 −𝑎21
𝑝1 = 𝜆
= 𝑎11 +𝑎22 −(𝑎12 +𝑎21 )

𝑝2 = 1 − 𝑝1
then
𝑎11 −𝑎12
𝑝2 = 𝑎11 +𝑎22 −(𝑎12 +𝑎21 )

𝑎22 −𝑎12 𝑎22 −𝑎12

𝑞1 = 𝜆
= 𝑎11 +𝑎22 −(𝑎12 +𝑎21 )

𝑞2 = 1 − 𝑞1
then
𝑎11 −𝑎12
𝑞2 = 𝑎11 +𝑎22 −(𝑎12 +𝑎21 )

20
 𝑎11 𝑎22 −𝑎12 𝑎21
∴ value of the game, V = 𝑎11 +𝑎22 −(𝑎12 +𝑎21 )

Hence
𝑝1 𝑎22 −𝑎21 𝑞1 𝑎22 −𝑎12
𝑝2
= ,
𝑎11 −𝑎12 𝑞2
= 𝑎11 −𝑎21 )

and
𝑎11 𝑎22 −𝑎12 𝑎21
V= 𝑎11 +𝑎22 −(𝑎12 +𝑎21 )
.

2.3 Methods to solve games without saddle point

2.3.1 Graphical Method

The graphical method is used to solve the games whose payoff matrix
has

1. Two rows and n columns (2𝑥𝑛)

2. m rows and two columns (𝑚𝑥2)

Algorithm for solving 2 x n matrix games

1. Draw two vertical axes 1 unit apart. The two lines are 𝑥1 =
0, 𝑥1 = 1

21
 2. Take the points of the first row in the payoff matrix on the ver-
tical line 𝑥1 = 1 and the points of the second row in the payoff
matrix on the vertical line 𝑥1 = 0.

3. The point 𝑎1𝑗 on the axis 𝑥1 = 1 is then joined to the point 𝑎2𝑗
on the axis 𝑥1 = 0 to give a straight line. Draw ’n’ straight lines
lines for 𝑗 = 1, 2, ...𝑛 and determine the highest point of the
lower envelope obtained. This will be the maximin point.

4. The two or more lines passing through the maximin point de-
termines the required 2𝑥2 pay off matrix. This in turn gives the
optimum solution by making use of analytical method.

⎡𝐵 𝐵 𝐵 ⎤
⎢ 1 2 3⎥
Example 3. Solve by graphical method A1 ⎢ 1 3 12 ⎥
⎢ ⎥
A2 ⎣ 8 6 2 ⎥⎦
⎢
Solution

22
 𝑎11 𝑎22 −𝑎21 𝑎12 6−72
𝑉 = (𝑎11 +𝑎22 )−(𝑎12 +𝑎21 )
= 5−18

𝑉 = 66∕13
𝑆𝐴 = (4∕13, 9∕13)
𝑆𝐵 = (0, 10∕13, 3∕13)
Algorithm for solving m x 2 matrix games

1. Draw two vertical axes 1 unit apart. The two lines are 𝑥1 =
0, 𝑥1 = 1

2. Take the points of the first row in the payoff matrix on the ver-
tical line 𝑥1 = 1 and the points of the second row in the payoff
matrix on the vertical line 𝑥1 = 0.

23
 3. The point 𝑎1𝑗 on the axis 𝑥1 = 1 is then joined to the point 𝑎2𝑗
on the axis 𝑥1 = 0 to give a straight line. Draw ’n’ straight lines
for 𝑗 = 1, 2, ...𝑛 and determine the highest point of the lower
envelope obtained. This will be the minimax point.

4. The two or more lines passing through the minimax point de-
termines the required 2𝑥2 pay off matrix. This in turn gives the
optimum solution by making use of analytical method.

⎡ 𝐵1 𝐵2 ⎤
⎢ ⎥
A1 ⎢ −2 0 ⎥
⎢ ⎥
Example 4. Solve by graphical method A2 ⎢ 3 −1 ⎥
⎢ ⎥
A3 ⎢ −3 2 ⎥⎥
⎢
⎢ 5 −4 ⎥⎦
A4 ⎣
Solution

24
𝑉 = 𝑎2 11 𝑎22 −𝑎21 𝑎1 = 6−3
(𝑎11 +𝑎22 )−(𝑎12 +𝑎21 ) 5+4

25
𝑉 = 3∕9 = 1∕3
𝑆𝐴 = (0, 5∕9, 4∕9, 0)
𝑆𝐵 = (3∕9, 6∕9)

2.3.2 Linear Programming Method

The two person zero-sum games can also be solved by linear pro-
gramming. The major advantage of using linear programming tech-
nique is that it helps to solve the mixed strategy games of larger di-
mension pay off matrix.
Consider a payoff matrix of size m x n.Let 𝑎𝑖𝑗 be the element in the
ith row and jth column of game payoff matrix, and letting 𝑝𝑖 be the
probabilities of m strategies (𝑖 = 1, 2, ..., 𝑚) for player A. Then, the
expected gains for player A, for each of player B’s strategies will be:

∑
𝑚
V= 𝑃𝑖 𝑎𝑖𝑗 , 𝑗 = 1, 2, ..., 𝑛
𝑖=1

The aim of player A is to select a set of strategies with probability

𝑝𝑖 (𝑖 = 1, 2, ..., 𝑚 on any play of game such that he can maximize his
minimum expected gains.

26
Now to obtain values of probability 𝑝𝑖𝑗 , the value of the game to
player A for all strategies by player B must be at least equal to V.
Thus to maximize the minimum expected gains, it is necessary that:

𝑎11 𝑝1 + 𝑎21 𝑝2 + ... + 𝑎𝑚1 𝑝𝑚 ≥ 𝑉

𝑎12 𝑝1 + 𝑎22 𝑝2 + ... + 𝑎𝑚2 𝑝𝑚 ≥ 𝑉

𝑎1𝑛 𝑝1 + 𝑎2𝑛 𝑝2 + ... + 𝑎𝑚𝑛 𝑝𝑚 ≥ 𝑉

where
𝑝1 + 𝑝2 + ... + 𝑝𝑚 = 1𝑎𝑛𝑑𝑝𝑖 ≥ 0

Dividing both sides of the m inequalities and equation by V the divi-

sion is valid as long as V > 0. In case V < 0, the direction of inequal-
ity constraints must be reverses. But if 𝑉 = 0, the division would
be meaningless. In this case a constant can be added to all entries of
the matrix, ensuring the value of the game (V) for the revised matrix
becomes more than zero. After optimal solution is obtained, the true
value of the game is obtained by subtracting the same constant value.
𝑝𝑖
Let 𝑉
= 𝑥𝑖 (≥ 0) we then have
𝑃1 𝑃 𝑃
𝑎11 + 𝑎21 2 + ... + 𝑎𝑚1 𝑚 ≥ 1
𝑉 𝑉 𝑉
27
 𝑃2 𝑃 𝑃
𝑎12 + 𝑎22 2 + ... + 𝑎𝑚2 𝑚 ≥ 1
𝑉 𝑉 𝑉
⋮
𝑃1 𝑃 𝑃
𝑎1𝑛 + 𝑎2𝑛 2 + ... + 𝑎𝑚𝑛 𝑚 ≥ 1
𝑉 𝑉 𝑉
𝑃1 𝑃2 𝑃
+ + ... + 𝑚 = 1
𝑉 𝑉 𝑉
Since the objective of player A is to maximize the value of the game,
V, which is equivalent to minimizing 1/V, the resulting linear pro-
gramming problem can be stated as:
Minimize 𝑍𝑝 (= 1∕𝑉 ) = 𝑥1 + 𝑥2 + ... + 𝑥𝑚
subject to the constraints

𝑎11 𝑥1 + 𝑎21 𝑥2 + ... + 𝑎𝑚1 𝑥𝑚 ≥ 1

𝑎12 𝑥1 + 𝑎22 𝑥2 + ... + 𝑎𝑚2 𝑥𝑚 ≥ 1

𝑎1𝑛 𝑥1 + 𝑎2𝑛 𝑥2 + ... + 𝑎𝑚𝑛 𝑥𝑚 ≥ 1

and
𝑥1 , 𝑥2 , ..., 𝑥𝑚 ≥ 0

where
𝑥𝑖 = 𝑃𝑖 ∕𝑉 ≥ 0; 𝑖 = 1, 2, ..., 𝑚

28
Similarly, player B has a similar problem with the inequalities of the
constraints reversed, i.e, the minimize the expected loss. Since min-
imizing V is equivalent to maximizing 1/V, therefore, the resulting
linear programming problem can be stated as:
Maximize: 𝑍𝑝 (1∕𝑉 ) = 𝑦1 + 𝑦2 + ... + 𝑦𝑛
subject to the constraints

𝑎11 𝑦1 + 𝑎12 𝑦2 + ... + 𝑎1𝑛 𝑦𝑛 ≤ 1

𝑎21 𝑦1 + 𝑎22 𝑦2 + ... + 𝑎2𝑛 𝑦𝑛 ≤ 1

𝑎𝑚1 𝑦1 + 𝑎𝑚2 𝑦2 + ... + 𝑎𝑚𝑛 𝑦𝑛 ≤ 1

𝑦1 , 𝑦2 , ..., 𝑦𝑛 ≥ 0

where,
𝑦𝑗 = 𝑞𝑗 ∕𝑉 ≥ 0𝑚; 𝑗 = 1, 2, ..., 𝑛

It may be noted that the L.P problem for player B is the dual of L.P
problem for a player A and vice versa. Therefore, the solution of the
dual problem can be obtained from the primal simplex table. Since
for both the player 𝑍𝑝 = 𝑍𝑞 , the expected gain to player A in the
game will be exactly equal to expected loss to player B.

29
1. For player B, who is assumed to be the loser, if each element in
a column, say 𝐶𝑟 is greater than or equal to the corresponding
element in another column, say 𝐶𝑠 in the payoff matrix, then
the column 𝐶𝑟 is said to be dominated by column 𝐶𝑠 and there-
fore, column 𝐶𝑟 can be deleted from the payoff matrix. In other
words, player B will never use the strategy that corresponds to
column 𝐶𝑟 because he will loose more by choosing such strat-
egy.

2. For player A, who is assumed to be the gainer, if each element

in a row, say 𝑅𝑟 , is less than or equal to the corresponding ele-
ment in another row, say 𝑅𝑠 , in the payoff matrix, then the row
𝑅𝑟 is said to be dominated by row 𝑅𝑠 and therefore, row 𝑅𝑟 can
be deleted from the payoff matrix. In other words, player A will
never use the strategy corresponding to row 𝑅𝑟 , because he will
gain less by choosing such a strategy.

3. A strategy say, k can also be dominated if it is inferior (less

attractive) to an average of two or more other pure strategies.

31
 In this case, if the domination is strict, then strategy k can be
deleted. If strategy k dominates the convex linear combination
of some other pure strategies, then one of the pure strategies
involved in the combination may be deleted. The domination
would be decided as per rules 1 and 2 above.

Remark 1. Rules (principles) of dominance discussed are used when

the payoff matrix is a profit matrix for the player A and a loss matrix
for player B. Otherwise the principle gets reversed.

Example 5. Players A and B play a game in which each has three

coins, a 5p, 10p and a 20p. Each selects a coin without the knowl-
edge of the other’s choice. If the sum of the coins is an odd amount,
then A wins B’s coin. But, if the sum is even, then B wins A’s coin.
Find the best strategy for each player and the values of the game.

Solution
The payoff matrix for player A is

32
 Player B
Player A 5𝑝 ∶ 𝐵1 10𝑝 ∶ 𝐵2 20𝑝 ∶ 𝐵3
5𝑝 ∶ 𝐴1 -5 10 20
10𝑝 ∶ 𝐴2 5 -10 -10
20𝑝 ∶ 𝐴3 5 -20 -20

It is clear that this game has no saddle point. Therefore, further

we must try to reduce the size of the given payoff matrix as further
as possible. Note that every element of column 𝐵3 (strategy 𝐵3 for
player B) is more than or equal to every corresponding element of
row 𝐵2 (strategy 𝐵2 for player B). Evidently, the choice of strategy
𝐵3 , by the player B, will always result in more losses as compared to
that of selecting the strategy 𝐵2 . Thus, strategy 𝐵3 is inferior to 𝐵2 .
Hence, delete the B3 strategy from the payoff matrix. The reduced
payoff matrix is shown below:

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

33
 After column 𝐵3 is deleted, it may be noted that strategy 𝐴2 of
player A is dominated by his 𝐴3 strategy, since the profit due to strat-
egy 𝐴2 is greater than or equal to the profit due to strategy 𝐴3 , re-
gardless of which strategy player B selects. Hence, strategy 𝐴3 (row
3) can be deleted from further consideration. Thus, the reduced pay-
off matrix becomes:
Player B
player 𝐴 𝐵1 𝐵2 Row minimum
𝐴1 -5 10 -5
𝐴2 5 -10 -10
Column maximum 5 10
As shown in the reduced 2 × 2 matrix, the maximin value is not
equal to the minimax value. Hence, there is no saddle point and one
cannot determine the point of equilibrium. For this type of game
situation, it is possible to obtain a solution by applying the concept
of mixed strategies.

34