Ap25250255

Dr.C.Loganathan, M.S.Annie Christi / International Journal of Engineering Research and
Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 2, Issue 5, September- October 2012, pp.250-255
Fuzzy game value of the Interval matrix
1
Dr.C.Loganathan, 2M.S.Annie Christi
1
Department of Mathematics, Maharaja Arts and Science College
Coimbatore – 641407, India
2
Department of mathematics, Providence College for Women,
Coonoor-643104, India

ABSTRACT:
In this paper we present a method to point) and the corresponding strategies which give
identify the fuzzy game value of the matrix with the saddle point are called optimal strategies.[1] and
interval data. The ideas of crisp matrix games are [6].The amount of payoff at an equilibrium point is
extended to fuzzy matrix games. This result is called the crisp game value of the matrix.
applied to a three player game. In (1), a11 is the crisp game value of the matrix.

Keyword: Crisp game value, Fuzzy game value, That is a11  1.
crisply related, fuzzily related and saddle value.
1.3 Interval Data:
1. Introduction Due to uncertainty, the payoffs in a matrix
Game theory has been used extensively to are not fixed numbers. Hence fuzzy games have
analyze conflict and co-operative situations in been studied. For further studies see [3], [5] and [7].
Economics, sociology, etc. One of the basic To model such uncertainty, we use interval valued
problems in the game theory is the two- player zero matrix.
sum game.
1.4 Definition:
A fundamental problem of classical game theory is Let A= a  be an mxn rectangular interval
ij
that the player makes decisions with crisp data.
valued matrix. This matrix defines a zero sum
However, in a real life world, most games always
interval matrix game provided whenever, the row
take place in uncertain environments. Because of
player X uses his ith strategy and the column player
uncertainty in real world applications, payoffs of a
Y uses his jth strategy. Then X wins and Y loses a
matrix game may not be a fixed number. This
situation forces us to introduce fuzzy matrix game. common value x  aij . 
In this paper, we consider the payoffs as interval 1.5 Example:
numbers.

1.1 Crisp game value of the Matrix:  [1,2] [7,8] [3,1] [5,3]
A game can be expressed as A=
 [4,6] [0,8] [1,3] [4,5] 
 
 1 3 2  [7,3] [2,1] [1,1] [4,5] 
 0 4 3   
A= (1)
 
 1 5 1 
  In this game if X chooses row one and Y chooses
column two, then X wins an amount x  [7, 8] and
Y loses the same amount.
In this game, the players X and Y have 3 and 4
strategies respectively. For player X, minimum
value in each row represents the least gain (payoff) 2. Interval Arithmetic
to him if he chooses his particular strategy. He will The extension of ordinary arithmetic to
then select the strategy that maximizes his minimum intervals is known as interval arithmetic.
gain. Let A  [aL , aR ], B  [bL , bR ] be two
Similarly, for player Y, the maximum value in each intervals.
column represents the maximum loss to him if he Then,
chooses his particular strategy. He will then select (i) A  B  [aL  bL , aR  bR ] .
the strategy that minimizes his maximum losses.
(ii) A  B  [aL  bR , aR  bL ] .
1.2 Definition:
(iii) AB = [min{ a L bL , a L bR , a R bL , a R bR },max {
If the maximum value equals the minimax
value, then the game is said to have a saddle a L bL , a L bR , a R bL , a R bR }]
value(equilibrium

250 | P a g e

(iv)  A= [  a L ,  a R ], if   0 and = [  a R ,
 a L ], if  <0. ab = 1, if a = b or a R < bL , a  b; or a L < bL
Similarly the other binary operations are defined < a R < bR .
0, if bR < a L , bL <
2.1Coimparison of Intervals
Comparison of two intervals is very a L < bR < a R .
important problem in interval analysis. In this bL  a L
section we consider the order relation (≤ and ≥) , if aL <
between intervals. Let a and b be two non-empty aR  aL
intervals. bL < bR < a R .
2.2 Disjoint Intervals: bR  a R
, if bL < a L
Case (i): If a= [ a L , a R ], b= [ bL , bR ] and bR  bL
if a R < bL then a<b or b>a crisply, which is similar < a R < bR .
to the definition of comparisons used in [4]
bL  a L b a
These values and R  R moves from
aR  aL bR  bL
aL aR bL bR
0 to 1 when x moves a L to bL and a R to bR from
2.3 Example: left to right. Only when a=b, a  b takes the value
1.Using simple algebraic operations, it can be seen
a = [2, 3] and b= [5, 6] be two intervals. Here a < b. that the membership value for b  a =1- a  b.

2.4 Equal Intervals: 3.2 Definition:
The binary fuzzy operator ≥ of two
Case (ii): a=b iff a≤b and a≥b. intervals a and b is defined as
a ≥ b= 1 if a = b ;or bR < a L or bL < a L < bR < a R .
2.5: Overlapping Intervals:
0 if a R < bL a  b, a L < bL < a R < bR
If a L < bL < a R < bR then for any x in [ a L , bL
a R  bR
], a<b.That is x  a, is less than every payoff in b. if a L < bL < bR < a R .
If x  [ bL , a R ], then every x  a, is less than or aR  aL
equal to b. Therefore a≤b (crisply). a L  bL
if bL < a L < a R < bR .
Similarly if y  [ a R , bR ], then y in b is greater than bR  bL
or equal to a. That is b≥a crisply.
The relations of two intervals can now be either
3. Nested sub intervals crisp of fuzzy as described below.
In terms of fuzzy membership
bR  a R 3.3 Definition:
a≤b= If the values of a ≤ b is exactly one or zero
bR  bL then we say that a and b are crisply related otherwise
If b  a then we say that they are fuzzily related.
bL  a L
a≤b= 4. Crisp Game value of the Matrix:
aR  aL The ideas and concept of crisp game value
if the matrixes can be extended to the matrix games
Consolidating the above discussions we define the with interval data where entries are crisply related.
fuzzy operations  and  as follows.
4.1 Definition:
3.1 Definition: Let A be an mxn interval game matrix such
The binary fuzzy operator  of two that all the intervals in the same row (or column) of
intervals a and b returns a real number between 0 A are crisply related. If there exists a g
and 1 as follows The binary fuzzy operator  of
ij  A  aij is simultaneously crisply less than or
two intervals a and b returns a real number between
0 and 1 as follows equal to a ik k  1..........n and crisply ≥

251 | P a g e

g lj l  1...m , then the interval g ij is called a 5. Fuzzy game value of the interval matrix
The crisp relativity may not be satisfied for
saddle interval of the game. The value of the saddle
all intervals in the same row (or column) .We now
interval is called the crisp game value of interval
define the fuzzy memberships of an interval being a
matrix game.
minimum and a maximum of an interval vector R
By the definition above, to determine whether an
and then we define the notion of a least and greatest
interval game matrix has a crisp game value, one
interval in R.
needs to verify the following:-
(i) For each row( 1 ≤ j ≤m) find the entry a ij , that is 5.1 Definition:
crisply less than or equal to all other entries of the The least interval of the vector R is defined as max
ith row. {min { ri  r j }} =l
(ii) For each column (1 ≤ j ≤m) find the entry a ij 1≤ i ≤ n ,1 ≤ j ≤m
that is crisply greater than or equal to all other and similarly the greatest interval of the vector R is
entries of the j th column. defined as min {max { ri  r j }} =g
(iii) Determine if there is an entry a ij ,that is 1≤ j≤ m, 1 ≤ i ≤n
simultaneously the minimum of the i th row and the
maximum of the jth column. 5.2 Example:
(iv) If any of the above values cannot be determined Find the least and the greatest interval of
for the matrix game, the crisp game value cannot be the vector R= {[1, 5], [2, 8], [3, 5]}
determined. (i) [1, 5] ≤ [2, 8] =1
[1, 5] ≤ [3, 5] = 1
4.2 Example:
Y Therefore min [1, 5] =1.
A= X
 1, 2 7,8  3,  1  5,  3 (ii) [2, 8] ≤ [1, 5] =0

 4, 6  1
 0, 8  1, 3 4, 5 
[2, 8] ≤ [3, 5] =
6
  7,  3  2,  1 1,1 4, 5  Therefore min [2, 8] =0.
 
  (iii) [3, 5] ≤ [1, 5] =0
1
We see that in the above game matrix a 14 , a 23 , a [3, 5] ≤ [2, 8] =
2
31 are the minimum of rows 1, 2 and3 respectively Therefore min [3, 5] = 0.
Therefore max {1, 0, 0} = 1.
and a 21 , a 12 , a 23 , a 34 are the maximum of column
Hence [1, 5] is the least interval.
1,2,3, and 4 .Therefore a 23 is the saddle interval of
the matrix A. The value of this interval is the crisp (i)[1, 5]  [2, 8] =0
game value of the interval matrix game. The [1, 5]  [3, 5] =0
corresponding strategies which give the saddle Therefore max [1, 5] =0.
interval values are the optimal strategies of the crisp (ii) [2, 8]  [1, 5] =1
game value. That is the optimal strategies for both X 1
and Y in a crisp interval valued matrix game are [2, 8]  [3, 5] = .
defined as
2
Therefore max [2, 8] =1.
(a) X should select any row containing a saddle
interval and
1
(b) Y should select any column containing a (iii) [3, 5]  [1, 5] =
saddle interval. 2
1
4.3 Definition: [3, 5]  [2, 8] = .
The game value of a crisply interval valued 6
matrix game is its saddle interval. This interval 1
game is fair, if its saddle interval is symmetric with Therefore max [3, 5] = .
respect to zero. That is in the form of [-a, a], a≥0. 2
In the above example, the saddle interval is a 23 = [0, 1
Hence min {0, 1, }= 0.
1].Here 0.5 is the midpoint. Hence the game is 2
unfair since the row player has an advantage of 0.5. This corresponds to [1, 5] which is the largest
interval.

252 | P a g e

Hence we define fuzzy game value of interval
matrix as follows.

5.3 Definition:
Let R be an mxn interval game matrix. If
there is an rij  R  rij is simultaneously the least
and the greatest interval for the ith row and jth
column of R , then that interval value is the fuzzy
game value of the interval valued matrix game.
Then as in the above case find the saddle interval of
6. Three player game the vectors {[ c1 , d1 ],[ c 2 , d 2 ],[ c3 , d 3 ],[ c 4 , d 4 ]}.
The above result can be extended to three player
game. This will be a special case where the players If the saddle interval falls into the column of X2,
A and B had the choice of choosing the strategies then the player C has a choice of winning the game
X1 and X2, whereas the player C will have a fixed with the fuzzy membership corresponding to the
strategy. interval. Otherwise he loses the game. This result
For example if the matrix game is given as follows: can be extended to many strategies X1, X2, X3
…Xn provided each time the payoff matrices are
given when the third player C chooses one of the
strategies. This is illustrated in the following
example.

6.1 Example:

Case (i):
When C chooses X1, if the payoff matrix is given as
below, then
Here the players A and b have strategies X1 and
X2.We assume that in the first case the player C
chooses X1 and in the second case he chooses X2.
Case (i):
When the player C chooses X1, if the payoff matrix
is given as:

We find the saddle interval of {[1, 6], [7, 8], [2, 6],
[3, 4]}

Least interval:
Find the saddle interval of the vectors {[ a1 ,b1 ], [
a 2 ,b2 ], [ a3 ,b3 ], [ a 4 ,b4 ]} if it exists. If the I (a) [1, 6]  [7, 8] =1
saddle interval falls into the column corresponding 1
to the strategy X1, then the player C has a choice of (b) [1, 6]  [2, 6]=
winning the game with the fuzzy membership 5
corresponding to that interval. Otherwise, he looses 2
the game. (c) [1, 6]  [3, 4]=
5
Case (ii): 1 2 1
Therefore min {1, , } =
Similarly, when the third player C chooses X2, if the 5 5 5
payoff matrix is given as follows:
II (a) [7, 8]  [1, 6]=0
(b) [7, 8]  [2, 6] =0
(c) [7, 8]  [3, 4]=0

253 | P a g e

Therefore min {0, 0, 0} =0 2 1 2 2
III (a) [2, 6]  [1, 6] =0 Therefore max {, , }= .
5 4 5 5
(b) [2, 6]  [7, 8] =1 2 1 2 2
Hence min { , ,1 , }=
5 2 5 5
1
(c) [2, 6]  [3, 4] =
4 This corresponds to the interval [3, 4].and [1, 6]
Since [3, 4] is the least and the greatest interval, it is
1 the saddle interval whose membership is 0.4.
Therefore min {0, 1,}=0
4 If the third player C chooses X2, he wins and the
2 game value lies in the interval [3, 4] with the fuzzy
IV (a) [3, 4]  [1, 6] = membership 0.4.If he chooses X1, he loses the
5 game.
1
(b) [3, 4]  [2, 6] =
2 Case (ii):
(c) [3, 4]  [7, 8] =1 Similarly when C chooses the strategy X2 if the
payoff matrix is given as below, then in the same
2 1 2 manner we find the saddle interval. If it falls in the
Therefore min { , , 1}= .
5 2 5 second column the player C wins with the fuzzy
1 2 2 membership of that interval. Otherwise he loses the
Hence max { , 0, 0, }= game.
5 5 5
This corresponds to the interval [3, 4]  [1, 6].
Therefore the least interval is [3, 4].

Greatest interval:
I (a) [1, 6]  [2, 6]=0

(b)  [7, 8] = 0
[1, 6]
2
(c) [1, 6]  [3, 4] =
5
2 2
Therefore max {0, 0, }= . We find the saddle interval of the vectors {[ 3,12 ], [
5 5
1 7,10 ],[ 5,10 ],[ 4,5 ]}
II (a) [2, 6]  [1, 6] =
5 4
I (a) [3, 12]  [7, 10] =
(b) [2, 6]  [7, 8] = 0 9
2
1 (b) [3, 12]  [5, 10] =
(c) [2, 6]  [3, 4] = 9
2 1
1 1 1 (c) [3, 12]  [4, 5] =
Therefore max {0, , }= 9
5 2 2 4 2 1 1
III (a) [7, 8]  [1, 6] =1 Min { , , }= .
9 9 9 9
(b) [7, 8]  [2, 6] =1
2
(c) [7, 8]  [3, 4] =1 II (a) [7, 10]  [3, 12]=
9
Therefore max {1, 1, 1} =1
(b) [7, 10]  [4, 5]=0
2 (c) [7, 10]  [5, 10] =0
IV (a) [3, 4]  [1, 6]=
5
1 2
(b) [3, 4]  [2, 6] = Min { , 0, 0} =0
4 9
2
(c) [3, 4]  [7, 8] = .
5

254 | P a g e

2 7. W. Winston, Operations Research –
III (a) [5, 10]  [3, 12]= Applications and Algorithms,
9 Brooks/Cole, 2004.
2 8. S. Wu, and V. Soo, A Fuzzy Game
(b) [5, 10]  [7, 10] =
5 Theoretic Approach to Multi-Agent
(c) [5, 10]  [4, 5] =0 Coordination, LNCS Vol. 1599, (1998),
pp. 76-87.
2 2
Min { , , 0}=0
9 5

7
IV (a) [4, 5]  [3, 12] =
9
(b) [4, 5]  [7, 10] =1

(c) [4, 5]  [5, 10] =1

7 7
Min { , 1, 1} =
9 9
1 7 7
Max {0, ,0 , }=
9 9 9
This corresponds to the interval [4, 5].

7.Conclusion
In this paper we have introduced two
person zero-sum interval valued matrix games. The
fuzzy game value of rectangular matrices can be
obtained by verifying whether the given matrix is
fuzzily related or not.
In some cases, during the comparison of intervals
will satisfy both  and  . But depending upon the
membership values we can draw conclusions. The
concept is extended to a three player game. This can
also be extended to a multi-player game.

REFERENCES
1. W.Dwayne Collins and Chenyi Hu, Fuzzily
determined Interval matrix games.
2. P.Dutta, Strategies and Games: Theory and
Practice, MIT Press, 1999.
3. de Korvin, C. Hu, and O. Sirisaengtaksin,
On Firing Rules of Fuzzy sets of Type II, J.
Applied Mathematics, Vol. 3, No.2, (2000),
pp. 151-159.
4. D. Garagic and J.B. Cruz, An Approach to
Fuzzy Noncooperative Nash Games, J.
Optimization Theory and Applications,
Vol. 118, No.2, (2000), pp. 475-491.
5. R. E. Moore, Methods and applications of
Interval analysis, SIAM Studies in Applied
Mathematics, 1995.
6. S.Russell and W. A. Lodwick, Fuzzy Game
Theory and Internet Commerce: e-Strategy
and Metarationality, NAFIPS 2002
Proceedings, 2002.

255 | P a g e

Ap25250255

More Related Content

What's hot (16)

Viewers also liked (20)

Similar to Ap25250255 (20)

Ap25250255