Unit II-chapter 2-Game Theory

m

m Introduction
m Characteristics of Games Theory
m Definitions
m Minimax Criteria
m Saddle Point
m Rectangular Games without Saddle Point
m Solution of m x n games by LP Method
m Principles of Dominance
m Applications of Game Theory
KOPPAR & ASSOCIATES, CHARTERED

5/21/2011 ACCOUNTANTS

m Game is defined as an activity between two or more
persons involving activities by each person according to a
set of rules, at the end of which each person receives some
benefit or satisfaction or suffers loss (negative benefit).
m The set of rules defines the game.
m Game theory is a type of decision theory in which oneǯs

choice of action is determined after taking into account all
possible alternatives available to an opponent playing the
same game, rather than just by the possibilities of several
outcomes.


m The game theory finds use in everyday life situations as
life is a series of struggles & competitions.
m The game theory is fundamentally based upon the

Ǯminimaxǯ or Ǯmaximinǯ criterion given by J. Van
Neumann who is called the father of game theory.
m The criterion implies that each player will act so as to

maximize his minimum gain or minimize his
maximum loss. Here it is assumed that each person
will act rationally.

m The various types of games can be classified on the basis of the
following characteristics.
¬

: If in a game, activities are determined by
skill, it is said to be a game of strategy. If they are determined
by chance, it is a game of chance.
¬¬ : A game is called a n-person game if the

number of persons playing is n. The person means an
individual or a group aiming at a particular objective.
¬¬¬
¬¬
¬ : These may be finite or infinite.


¬

¬ ¬ ¬
:
In a particular activity, choices may be finite or infinite. A finite
game has a finite number of activities, each involving a finite
number of alternatives. Otherwise the same is said to be infinite.

¬

¬¬
¬

Completely available, partly available or not available at
all
¬ : A quantitative measure of satisfaction a person gets at
the end of each play is called a pay off. If ϝVi = o , where Vi is
the payoff to the player Pi ( 1 ϐ i ϐ n ) then the game is said to
be a zero-sum game.

M
m
¬
¬: A competitive situation is called a competitive
game if it has the following 4 properties.
[ There are finite number (n) of competitors called players such that
n ϑ 2. In case n=2, it is called a 2 person game and in case n > 2, it is
referred to as n-person game.
[ Each player has a list of finite number of possible activities.
[ A play is said to occur when each player chooses one of his activities.
The choices are assumed to be made simultaneously i.e. no player
knows the choice of the other until he has decided on his own.
[ Every combination of activities determines an outcome (which may

be points, money or anything else whatsoever) which results in a
gain of payments (+ve, -ve, zero) to each player, provided each
player is playing uncompromisingly to get as much as possible.
Negative gain implies the loss of same amount.

M
m £ : Competitive games
are classified according to the number of players involved
i.e. as a two-person game, three-person game etc. Another
distinction is between zero-sum games and non-zero-sum
games. If the players make payments only to each other i.e.
the loss of one is the gain of others, and nothing comes
from outside, the competitive game is said to be zero-sum.
Mathematically, it is represented as ϝVi = o
m A game which is not a zero-sum is called a non zero-sum

game. Most of the competitive games are zero-sum games.

M
m {

: A strategy of a player has been defined as a rule for decision-
making in advance of all the plays he decides the activities he should
adopt.
m The strategy may be of two kinds.
a) {

: If a player knows exactly what the other player is going
to do, a deterministic situation is obtained and the objective function is
to maximize the gain. Therefore, a pure strategy is a decision rule
always to select a particular course of action.
b) ¬!" {

: If a player is guessing as to which activity is to be
selected by the other on any particular occasion, a probabilistic
situation is obtained and the objective function is to maximize the
expected gain.

M
m Thus the mixed strategy is a selection among pure
strategies with fixed probabilities.
m Mathematically, a mixed strategy for a player with m (ϑ 2)

possible courses of action is denoted by the set S of m non-
negative real numbers whose sum is unity, representing
probabilities with which each course of action is chosen.
If xi (i = 1, 2, - - m) is the probability of choosing the course
i, then
S= (x1, x2, x3 - - - xm)
Subject to x1 + x2 + - - xm =1
And x1 ϑ o, x2 ϑ 0, - - xm ϑ 0

M
m M # $%
&:
A game with only 2 players is called a ǲtwo-person, zero-
sum gameǳ if the losses of one player are equivalent to the
gains of the other, so that the sum of their net gains is zero.
m Two-person, zero-sum games are also called rectangular

games as they are usually represented by a pay off matrix in
rectangular form.

M
m
¬!: Suppose the player A has m activities and the player B
has n activities then a payoff matrix can be formed by adopting the
following rules.
m Row designation of each matrix are activities available to player A
m Column designations for each matrix are activities available to player B.
m Cell entry Vij is the payment to player A in Aǯs pay off matrix when A
chooses the activity i and B chooses the activity j.
m With a Ǯzero-sum, two-person gameǯ, the cell entry in the player Bǯs pay
off matrix will be negative of the corresponding cell entry Vij in the
player Aǯs pay off matrix so that the sum of pay off matrix for player A
and player B is ultimately zero.
M
Player Aǯs pay off matrix
'
1 2 --- j -- -n
1 V11 V12 - - - - - - Vij - - Vin
2 V21 V22 - - V2j - - V2n

i Vi1 Vi2 - - Vij - - Vin
m Vm1 Vm2 - - Vmj - - - Vmn

!
"#$%
m The minimax criterion of optimality states that if a player

lists the worst possible outcomes of all his potential
strategies, he will choose that strategy to be most suitable
for him which corresponds to the best of these worst
outcomes. Such a strategy is called an optimal strategy.
m This can be illustrated with the help of the following

example.

"&#
m Table below illustrates a game, where competitors A & B
are assumed to be equal in ability & intelligence. A has a
choice of strategy 1 & strategy 2 while B can select strategy
3 or strategy 4.
Competitor B
{

{

Strategy +4 +6
Competitor A
Strategy +3 +5

"&#
m Both competitors know the pay offs for every possible strategy.
The game favours A since all values are positive. Values that
favour B would be negative. Based upon these conditions, game
is biased against B. However, since B must play the game he will
play to minimize his losses.
m The various possible strategies for the two competitors are
[ A wins the highest game value if he plays strategy 1 all the time since
it has higher values than strategy 2
[ B realizes this situation and plays strategy 3 in order to minimize his
losses since the value of 4 in strategy 3 is lower than the value of 6 in
strategy 4.

"&#
m The game value is 4 since A wins 4 points while B losses 4 points
each time the game is played, the Ǯgame valueǯ is the average
winnings per play over a long number of plays.
m The above is a zero-sum game since A wins as much as B loses.
m From the above we can conclude that
[ The strategy that A should use is one that ensures that his
average gain per play is at least equal to the value of the game
(maximizing his minimum gains)
[ The strategy that B should use is one that his average loss per
play is no more than the value of the game (minimizing his
maximum losses).

$# '#&!
m The concept of saddle point is illustrated with the help of
the following example.
'
P Q Minimum
of row
Player A L -3 3 -3
M -2 4 -2
N 2 3 2
Maximum 2 4 2 [Maximin &
of column Minimax]

$# '#&!
m Player A would gain (-3) if uses strategy L, (- 2) if he uses strategy M &
(+2) if he uses strategy N. So he will play strategy N as it will maximize
his minimum gain. This strategy is maximin strategy and his
corresponding gain is called the maximin value of the game.
m Player B, on the other hand, wants to minimize his losses. If he plays
strategy P, he can lose no more than max (-3, -2, 2) = 2 regardless of Aǯs
selection. If he plays his second strategy Q, he stands to lose max (3, 4,
3) = 4.
m He will, therefore, select that strategy that minimizes his maximum
loss. So he chooses strategy P and his max loss will be (2). This is
called minimax value of the game.
m In the instant case, minimax value = maximin value. When the two
are equal, the corresponding pure strategies are called
¬

¬ and the game is said to have a "" ¬
or
(¬¬¬ ¬
m {"" ¬
¬
¬ ¬
¬ ¬
" ¬
¬ ¬
.

"&#(
m Find the saddle point
'
X Y Z Minimum
of row
P 1 13 11
Player A Q -9 5 -11
R 0 -3 13
Maximum of
column

"&#
Find the saddle point
'
X Y Z Minimum
of row
P 3 -4 8
Player A
Q -8 5 -6
R 6 -7 6
Maximum
of column


%#)$#
m In cases where there is no saddle point, players will resort to
mixed strategies.
m In a 2 x 2 game, the following method provides an easy way of

finding the optimum strategies for each player. It consists of the
following steps.
m Step 1: Subtract the two digits in column 1 and write them under
column 2, ignoring sign.
m Step 2: Subtract the two digits in column 2 and write them under
column 1, ignoring sign.
m Step 3: Similarly proceed for 2 rows. These values are called

oddments. They are the frequencies with which the players must
use their courses of action in their optimum strategies.

"&#(
m In a game of matching coins, player A wins Rs 2 if there are
2 heads, wins nothing if there are two tails & loses Rs 1
when there is one head and one tail. Determine the payoff
matrix, best strategies for each player and the value of game
to A.

$#
m The pay off matrix for A will be
'
H T Oddments
H 2 -1 1 [¼ = 0.25]
Player A
T -1 0 3 [¾ = 0.75]
Oddments 1 3
[¼= 0.25] [¾ = 0.75]

$#
m Thus for optimum gains, player A should use strategy H for 25% of the
time and strategy T for 75% of the time while player B should use
strategy H 25% of the time and strategy T 75% of the time.
m To obtain the value of the game, we can use either Aǯs oddments or Bǯs
oddments.
m m ¬ ' ""
:
B plays H, value of game = V = {1 x 2 Ȃ 3 x 1} / {3 + 1} = -1 / 4
B plays T, V = [1 x -1 + 3 x 0] / [3 + 1] = -1 / 4
m m ¬ ) ""
:
A plays H, V = [1 x 2 Ȃ 1 x 3] / [3 + 1] = -1 / 4
A plays T, V = [-1 x 1 + 0 x 3] / [3 + 1] = -1 / 4
m Thus the final solution of the game is

A (1, 3) B (1, 3) V = - 1 / 4
m This is the value of the game to A, i.e. A gains Rs (-1 / 4) i.e. he loses Rs
1/4 which B, in turn gets.

"&#
m Find the optimal strategies for the games for which the pay off matrices
are given below also find the value of the game.
*
I II
I 1 3
Player 1
II 4 2
+
I II
I -4 6
Player X
II 2 -3

$#
*
I II Oddments
Player 1 I 1 3 2 [2/4]
II 4 2 2[2/4]
Oddments 1 3
[1/4] [3/4]
m ¬ *) ""
:
2 plays I, value of game = V = {1 x 2/4}+{4 x 2/4} = 5/2
2 plays II, V = [3x 2/4]+ [2 x 2/4] = 5/2

"#M*%
m Rule 1: If each element in one row, say rth of the pay off
matrix[ vij ] is less than or equal to the corresponding
element in the other row, say sth, then the player A will
never choose the rth strategy.
m The value of the game and the non-zero choice of
probabilities remain unaltered even if rth row is deleted
from the pay off matrix.
m Such rth row is said to be dominated by sth row.

"#M
#(
m Consider the following matrix

I + , -
II - - +
III - +
m In this case every element of 2nd row is less than or equal to

the corresponding element of 3rd row hence 2nd row is said
to be dominated by 3rd row.
m The value of the game will remain unaltered even if the 2nd
row is deleted from the pay off matrix.

"#M*%
m Rule 2: If each element in one column Cr is greater than or

equal to the corresponding element in the other column,
say Cs, then player B will never use the strategy
corresponding to column Cr.
m In this case column Cs dominates column Cr.

"#M
#
m Consider the same example given above.

I + , -
II - - +
III - +
m Every element in 3rd column is grater than or equal to the

corresponding element in column 1
m Hence column 1 dominates column 3.
m Hence deleting column 3 will not affect the nature of the
game.

"#M*%
m Rule 3: 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 of two or more other pure
strategies.
m In general, if some convex linear combination of some rows
dominates the ith row, then ith row will be deleted.
m Conversely if the ith row dominates the convex linear
combinations of some other rows, then one of the rows
involved in the combination may be deleted.
m Similar arguments for columns also.

"#M
#-
m Consider the following game

I a
II ( . ,
III ( -
m None of the pure strategies of player A is inferior to any other of

his pure strategies. However, the average of the players Aǯs first
and second pure strategies gives
m { 5-1/2, 0+8/2, 2+6/2} (2, 4, 4)
m This is superior to player Aǯs third pure strategy; so the third row
may be deleted from the matrix.

"#M"&#
m Solve the following game by using the rules of dominance
'
I II III
I ( /
Player A
II , /
III a ,

"#M"&#
$#
m Consider the rows first, row 3 is inferior to the row 2, so row 3 can be
deleted from the pay off matrix. Now consider the columns, column 3
is dominated by column 1, therefore, column three will be deleted.
m The resultant pay off matrix will be
'
Player A I II
I ( /
II ,

"#M"&#
$#
m This is a 2x2 game without saddle paint; this can be solved
either by LPP(simplex method) or by system of
simultaneous equations.
m x1+6x2 = v ; x1+x2 = 1; 7x1+2x2 = v [for player A]
m y1+y2 = v; 6y1 + 2y2 = v; y1+y2=1 [for player B]
m x1 = 2/5, x2 = 3/5
m so strategy for A = (2/5, 3/5, 0)
m y1 = 1/2, y2 = 1/2
m So strategy for B = (1/2, 1/2, 0)

$#%&
0 $"#!
m The procedure to solve the game by simplex method is
described below by means of an example.
m Solve the following 3 x 3 games by simplex method
'

- ( -
Player A
- - (
+ - -

$#
m {
,First apply Minimax (Maximin) criteria to find the value of the
game
%
¬
- ( - -3
- - ( -3 Maximin
= -3
+ - - -4
Column 3 3 3
Max
Minimax
=3

$#
m Since maximin value is -3, the value of the game may be negative or
Zero as -3< v <3
m A constant C is added to all the elements of the matrix which is at least
equal to the negative of the maximin value.
m i.e., C ϑ 3; let c = 5, then the matrix can be rewritten as
'

. +
Player A
. +
( .

$#
m Players Bǯs LPP is
m Maximize , y0 = Y1 + Y2 + Y3 Subject to
8Y1 + 4Y2 + 2Y3 ϐ 1
2Y1 + 8Y2 + 4Y3 ϐ 1
Y1 + 2Y2 + 8Y3 ϐ 1
Y1, Y2, Y3 ϑ 0
m {
* Introduce slack variables and construct the simplex
table
8Y1 + 4Y2 + 2Y3 + S1 = 1
2Y1 + 8Y2 +4Y3 + S2 = 1
Y1 + 2Y2 + 8Y3 +S3 = 1

$#
CJ = 1 1 1 0 0 0
* * *

$( ( + ( (1.
$ ( . + (
$- ( ( . ( (
( ( (

$#
CJ = 1 1 1 0 0 0
* * *

V


$#
CJ = 1 1 1 0 0 0
* * *

(
$

-


$#
* * *

( ( (1(+ (
( ((1(2,

- ( a1+2
+a1(2,


$#
m Solution for Bǯs Problem is
m y1 = Y1/y0 = 1/14 / 45/196 = 14/45
m y2 = Y2/y0 = 11/196 / 45/196 = 11/45
m y3 = Y3/y0 = 5/49 / 45/196 = 20/45
m The optimal strategies for player A are given by duality rules

m x0 = y0 = 45/196
m X1 = ϙ4 = 5/49
m X2 = ϙ5 = 11/196
m X3 = ϙ6 = 1/14
m x1 = X1/x0 = 20/45
m x2 = X2/x0 = 11/45
m x3 = X3/x0 = 14/45
m Therefore V = [1/y0]Ȃ C = [196/45] -5 = -29/45

m Value of the game = -29/45 [for B]
m Value of the game for A = 29/45.


Unit II-chapter 2-Game Theory

Enviado por

Dados do documento

Descrição original:

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Unit II-chapter 2-Game Theory

Enviado por

Direitos autorais:

Formatos disponíveis

m 

KOPPAR & ASSOCIATES, CHARTERED

m The set of rules defines the game.

m Game theory is a type of decision theory in which oneǯs

KOPPAR & ASSOCIATES, CHARTERED

m The game theory is fundamentally based upon the

m The criterion implies that each player will act so as to

¬¬     : A game is called a n-person game if the

KOPPAR & ASSOCIATES, CHARTERED

KOPPAR & ASSOCIATES, CHARTERED

[ Each player has a list of finite number of possible activities.

[ Every combination of activities determines an outcome (which may

KOPPAR & ASSOCIATES, CHARTERED

m A game which is not a zero-sum is called a non zero-sum

KOPPAR & ASSOCIATES, CHARTERED

m The strategy may be of two kinds.

KOPPAR & ASSOCIATES, CHARTERED

m Mathematically, a mixed strategy for a player with m (ϑ 2)

KOPPAR & ASSOCIATES, CHARTERED

m Two-person, zero-sum games are also called rectangular

KOPPAR & ASSOCIATES, CHARTERED

m Row designation of each matrix are activities available to player A

m Column designations for each matrix are activities available to player B.

1 V11 V12 - - - - - - Vij - - Vin

2 V21 V22 - - V2j - - V2n

i Vi1 Vi2 - - Vij - - Vin

m Vm1 Vm2 - - Vmj - - - Vmn

KOPPAR & ASSOCIATES, CHARTERED

m The minimax criterion of optimality states that if a player

m This can be illustrated with the help of the following

KOPPAR & ASSOCIATES, CHARTERED

KOPPAR & ASSOCIATES, CHARTERED

KOPPAR & ASSOCIATES, CHARTERED

KOPPAR & ASSOCIATES, CHARTERED

KOPPAR & ASSOCIATES, CHARTERED

KOPPAR & ASSOCIATES, CHARTERED

KOPPAR & ASSOCIATES, CHARTERED

KOPPAR & ASSOCIATES, CHARTERED

m In a 2 x 2 game, the following method provides an easy way of

m Step 3: Similarly proceed for 2 rows. These values are called

KOPPAR & ASSOCIATES, CHARTERED

KOPPAR & ASSOCIATES, CHARTERED

KOPPAR & ASSOCIATES, CHARTERED

m Thus the final solution of the game is

KOPPAR & ASSOCIATES, CHARTERED

KOPPAR & ASSOCIATES, CHARTERED

KOPPAR & ASSOCIATES, CHARTERED

KOPPAR & ASSOCIATES, CHARTERED

m In this case every element of 2nd row is less than or equal to

KOPPAR & ASSOCIATES, CHARTERED

m Rule 2: If each element in one column Cr is greater than or

m In this case column Cs dominates column Cr.

KOPPAR & ASSOCIATES, CHARTERED

m Every element in 3rd column is grater than or equal to the

KOPPAR & ASSOCIATES, CHARTERED

m Rule 3: Dominance need not be based on the superiority of

KOPPAR & ASSOCIATES, CHARTERED

m None of the pure strategies of player A is inferior to any other of

KOPPAR & ASSOCIATES, CHARTERED

KOPPAR & ASSOCIATES, CHARTERED

KOPPAR & ASSOCIATES, CHARTERED

KOPPAR & ASSOCIATES, CHARTERED

KOPPAR & ASSOCIATES, CHARTERED

m

¬¬ : A game is called a n-person game if the

- ( a1+2