International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169

Volume: 4 Issue: 4 897 - 905

_______________________________________________________________________________________________
Two-Person Zero-Sum Game using MAX-MIN Approximation

M. D. Khedekar S. N. Yadav M. S. Bapat

Department of Basic Sciences Department of Basic Sciences Department of Mathematics
ADCET, Ashta - 416 301, ADCET, Ashta - 416 301, Willingdon College, Sangli – 416
Maharashtra, India Maharashtra, India 415, Maharashtra, India
Email: khedekarmd@gmail.com Email: snyadavmath@gmail.com Email: msbapat@gmail.com

S. J. Aher
Department of Basic Sciences
ADCET, Ashta - 416 301, Maharashtra, India
Email: suja11_aher86@yahoo.co.in

Abstract:-In this paper we discuss Two-Person Zero-Sum Game with Epsilon-Delta fuzzy payoff matrix is defined and obtained optimal fuzzy
strategies by using MAX-MIN approximation. Criterion of Fuzzy Minimax-Maximin is established. Some examples are given.

Keywords: Epsilon-delta fuzzy number; Fuzzy payoff matrix; Fuzzy saddle point; Fuzzy two person zero sum game, MAX-MIN approximation.
__________________________________________________*****________________________________________________
1. Introduction
The mathematical analysis of competitive problems is based upon ‘Minimax-Maximin Criterion’ of J. Von Neuman. The
simplest type of competitive situations is two-person zero-sum game [4-6]. These games involve only two players; they are called
zero-sum games because one player wins whatever the other player loses [7].
Game theory is a type of decision theory in which one’s choice of action is imprecise after taking into account all
possible alternatives available to an opponent playing the same game. Fuzzy games are intended to model conflict situation with
imprecise information, Payoffs strategies etc. In this paper, we have proved ‘Fuzzy Minimax-Maximin criterion’ by using
Epsilon–delta fuzzy number [1] as cell entries of payoff matrix. We use MAX-MIN approximation [2] for establishing the
Minimax-Maximin criterion.
In section 2 we give basic notions which are used in the sequences. In section 3 fuzzy two person zero sum game is
discussed. In section 4 results are elaborated by empirical examples.
2. Preliminaries
Basic notions of Two-Person Zero-Sum Games
 A two-person game is characterized by the strategies of each player and the payoff matrix.
 The payoff matrix shows the gain (positive or negative) for player 1 that would result from each combination of
strategies for the two players. Note that the matrix for player 2 is the negative of the matrix for player 1 in a zero-sum
game.
 The entries in the payoff matrix can be in any units as long as they represent the utility (or value) to the player.
 There are two key assumptions about the behavior of the players. The first is that both players are rational. The second is
that both players are greedy meaning that they choose their strategies in their own interest (to promote their own wealth).
 The definition of a two-person zero-sum game in normal form amounts to defining sets of strategies A and B of players I
and II respectively, and of the pay-off function H of player I, defined on the set A×B of all situations (the pay-off function
of player II is −H by definition). Formally, a two-person zero-sum game Γ is given by a triplet Γ=⟨A, B, H⟩. Play consists
in the players choosing their strategies a ∈ A, b ∈ B, after which player I obtains the sum H (a, b) from player II. Such a
definition of a two-person zero-sum game is sufficiently general to include all variants of two-person zero-sum games,
897
IJRITCC | April 2016, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 4 Issue: 4 897 - 905
_______________________________________________________________________________________________
including dynamic games and positional games provided that the sets of strategies and the pay-off function are properly
described. A rational choice of actions (strategies) of the players in the course of a two-person zero-sum game is based on
a minimax principle:
If max a∈ A, inf b∈ B, H (a, b) =min b∈ B, sup a ∈ AH (a, b) (1)
or
sup a ∈ A , inf b ∈ B, H(a, b)  inf b ∈B, sup a ∈A H(a, b) (2)
the game Γ has optimal strategies (ϵ-optimal strategies, respectively) for both players. The common value of both parts of
equation (2) is called the value of the game Γ. However, equations (1) or (2) may not be valid even in the simplest cases.
Eg. In a matrix game with payoff matrix
Player B
 1 1
Player A  1 1 
 
The following equalities are valid:
max min aij  1, min max aij  1
i j j i

Definition 1. [8] A fuzzy number A is a subset of the real line  with membership function A :   [0,1] such that A is
normal, A is fuzzy convex and upper semi-continuous i.e. α-cut is closed for all   0, 1 .Support of A is bounded. If left hand
curve and right hand curve are straight lines then the fuzzy number is called triangular fuzzy number.
Definition 2. [1] If r is a real number then ε-δ fuzzy number r , is the triangular fuzzy number for some  ,   ¡ ,(ε,   0) is a
fuzzy set r , : ¡  [0,1] defined by

 x  (r   )
 , if r   < x  r ,


 x  (r   )
r , ( x)   , if r < x  r   ,
  
0, otherwise.


A triangular fuzzy number A  (l , m, n) in above notation is denoted by A  mml ,nm . Also, r ,  (r   , r , r   ) .

1------------------------------

r , ( x)

0 r  r 

Fig.1. Membership functions of the fuzzy number r , .

898
IJRITCC | April 2016, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 4 Issue: 4 897 - 905
_______________________________________________________________________________________________
The support of ε-δ fuzzy number r , is  r   (1   ), r   (1   )  , r  ¡ ,  ,   ¡ and  ,   0 . The α-cut of r , is
denoted by (r , ) = [r  ε(1  α), r   (1  α)],   (0.1] .

The above notation for triangular fuzzy number is simple and may be considered as a family of functions of three
parameters.

Definition 3 [2] Let r1 ,1 and s 2 , 2 be any two epsilon-delta fuzzy numbers where r  s . If r  1  s   2 and

r  1  s   2 then we define
(r1 ,1  s 2 ,2 )  r1  2 ,1 2 and. r1 ,1  s 2 ,2  s1  2 , 1 2

If s   2  r  1 and r  1  s   2 then we define

(r1 ,1  s 2 ,2 )  r 2 ( s r ),1 and (r1 ,1  s 2 ,2 )  s1 ( s r ),2 .

If r  1  s   2 and s   2  r  1 then
(r1 ,1  s 2 ,2 )  r1 ,2 ( s r ) and (r1 ,1  s 2 ,2 )  s 2 ,1 ( s r ) .

If s   2  r  1 and s   2  r  1 then
r1 ,1  s 2 ,2  r 2 -( s-r ), 2 ( s-r ) and (r1 ,1  s 2 ,2 )  s1 ( s r ),1 ( s r ) .
The above evaluations are triangular approximations of max and min operations on fuzzy numbers. Many researchers
have used Dubois and Prade approximation of max and min operations given below
r1 ,1  s 2 , 2  (r  s)1  2 , 1  2 and r1 ,1  s 2 , 2  (r  s)1  2 , 1  2 .

Addition or subtraction of the term r  s may be considered as a correcting factor and thereby gives better approximation.
3. Fuzzy two person zero sum game

The sets of the possible feasible strategies of player I are two fuzzy sets A and B on S1 and S 2 respectively. The two

payoff functions P1 (for player I) and P2 (for player II) from S1  S2  0,1 are fuzzy.
Epsilon-delta Payoff matrix

Definition 4. A quantitative measure of satisfaction of a person expressed in terms of epsilon-delta fuzzy numbers is called
epsilon-delta payoff
Suppose player A has m activities and player B has n activities. Then a fuzzy
Payoff matrix can be formed by adopting following rules.

The player A’s fuzzy payoff matrix

899
IJRITCC | April 2016, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 4 Issue: 4 897 - 905
_______________________________________________________________________________________________
Player B
Strategies 1 2 … j … n

Player A
11 12 1j 1n
1 r 11,11 r 12,12
…. r 1 j ,1 j
….. r 1 n ,1 n

21 22 2j 2n
2 r 21, 21 r 22, 22
….. r 2 j , 2 j
…… r 2 n , 2 n

: : : : :
….. ……
: : : : :
i1 i2 ij in
i r i 1, i 1 r i 2, i 2
…… r  ij , ij
……. r in , n

: : : : :
….. ……
: : : : :
m1 m2 mj mn
m r m1, m1 r m 2, m 2
…… r mj , mj
……. r mn , mn

We denote above fuzzy payoff matrix as follows

 11 12
  1n

 r 11,11 r 12,12
r  r11
1 n ,1 n r12  r1n 
 21 22 2n   21 
 r 22  r 2n 
  r  21, 21 r r r
AE , D 22, 22 2 n , 2 n 
,where  A   ,
           
   m1 
 r m1 r r m2  r mn  mn
 r  
m2 mn

  m1, m1 r m 2, m 2 mn , mn  mn

 11 12  1n   11 12  1n 

  22   2 n    22   2 n 
E   ij    21 and D   ij   
21

           
   
 m1  m 2   mn  mn  m1  m 2   mn  mn

Fuzzy Minimax-Maximin Criterion

Theorem 4. Let, P be the payoff matrix for a two person zero sum game where P : X  0,1 .

If
i 
 r
j
ij
ij , ij
 r and  
j
 i r    r
ij
ij , ij
then r r

i.e.    r       r    .
ij ij
j i ij , ij i j ij , ij

Proof. We have

 r  r
ij ij
and  r    r   , i, j
ij ij
i ij , ij ij , ij j ij , ij ij , ij

Let maximum be obtained at i  i* and minimum be obtained at j  j * . Therefore,

i* j ij*
 r  r , i, j
ij
r 
i* j , i* j ij , ij 
ij* , ij*

900
IJRITCC | April 2016, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 4 Issue: 4 897 - 905
_______________________________________________________________________________________________

  r     r      r    , i, j
i* j ij ij*
j  i* j , i* j  ij , ij i  ij* , ij* 

 

 

ij ij

 r   r
j  i  i  j   

  ij , ij 
 ij , ij 

Fuzzy Saddle Point

Definition 5. If
 

 

ij ij

 r   r
j  i  i  j   

  ij , ij 
 ij , ij 

i.e.If r  r  r  , Then the fuzzy game has a saddle point at the cell  i, j  .
ij
ij , ij

4. Fuzzy optimal strategies

ij
If fuzzy payoff r  is a saddle point the players have fuzzy optimal strategies in pure strategies: Player I have i th and
ij , ij

player II have jth fuzzy optimal strategies respectively.

Value of Fuzzy Game
The fuzzy payoff r 
ij
ij , ij
at saddle point  i, j  is called value of fuzzy game.
Fair Fuzzy Game
 0 ,
ij
A fuzzy game is said to be fair game if saddle point r  ij , ij

Example 6. Consider two-person zero-sum game which represents fuzzy payoff to the player A. Find the optimal strategy if any
Player B
I II III Row Minimum

39,10 27,9 65,9 39,9

I

29,7 04,8 29,7 09,7 Maximin Value

Player A II  r   09,7
513,11 26,8 47,9 411,9
III
Column Maximum
512,11 04,9 65,9

Minimax Value  r   07,9

Solution: Row Minimum

I :  39,10  27,9   65,9  39,9

39,10 27,9 39,9 65,9

12 9 3 2 7 12 1 6 15
901
IJRITCC | April 2016, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 4 Issue: 4 897 - 905
_______________________________________________________________________________________________
No Crossing No Crossing
39,10  27,9  39,9 39,9  65,9  39,9

II :  29,7  04,8   29,7  09,7

04,8 29,7 07,7 29,7

7 4 0 2 8 9 7 0 2 7 9
Left Crossing No Crossing
29,7  04,8  07,7 07,7  29,7  09,7

III :  513,11  27,9   47,9  411,9

27,9 513,11 47,9 213,9

9 8 2 5 7 16 15 11 4 2 5 7
No Crossing Left Crossing
513,11  27,9  213,9 213,9  47,9  411,9
Maximin

 3 9,9  09,7   411,9  09,7

39,9 09,7 411,9 09,7

12 9 3 0 6 7 15 9 5 7
No Crossing No Crossing
39,9  09,7  09,7 411,9  09,7  09,7
In general player A selects the strategies that maximize his minimum gain.

902
IJRITCC | April 2016, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 4 Issue: 4 897 - 905
_______________________________________________________________________________________________
Column Maximum

I :  39,10  29,7   513,11  512,11

39,10 29,7 29,10 513,11

12 7 3 2 7 9 8 7 2 5 12 16
No Crossing Left Crossing
39,10  29,7  29,10 29,10  513,11  512,11

II :  27,9  04,8   27,9  04,9

27,9 04,8 27,9 04,9

9 4 2 0 7 8 9 4 2 0 7 9
No Crossing No Crossing
27,9  04,8  04,9 04,9  27,9  04,9

III :  65,9  29,7   47,9  65,9

29,7 65,9 47,9 65,9

7 2 1 6 9 15 11 4 1 5 6 15
No Crossing No Crossing
65,9  29,7  65,9 65,9  47,9  65,9
Minimax:

5 12,11  04,9   65,9  65,9

04,9 512,11 07,9 65,9

7 4 0 5 9 16 7 0 1 6 9 15
Left Crossing No Crossing
512,11  04,9  07,9 07,9  65,9  07,9
Player B selects the strategies that minimize his maximum loss.
In this case equality holds,

903
IJRITCC | April 2016, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 4 Issue: 4 897 - 905
_______________________________________________________________________________________________
 

 

ij ij

 r   09,7    r   07,9
j  i  i  j 
  ij , ij 
  ij , ij 

i.e. r r.

Thus, fuzzy optimal strategies of player I and II is 2 and fuzzy game have saddle point 08,9 .

Example7. Consider following fuzzy two-person zero-sum game

Player B
I II III Row Minimum

31,2 42,3 81,2 42,2

I

82,1 56,1 62,2 86,1 Maximin Value

Player A II  r   42,2
62,2 71,2 62,2 73,2
III
Column Maximum
61,3 51,3 81,5

Minimax Value  r   51,2

 

r 

ij ij

As discussed in above example,    

 1,2 i 
5 , r   42,2
 j  ij ,ij 
 i  ij , ij 
j
 
 

r 

ij ij

Also,    
 i  r 
 j  ij ,ij 
 i  ij , ij 
j
 

Such fuzzy games are called fuzzy games without saddle point.

Example 8. Consider following fuzzy two-person zero-sum game

Player B
I II Row Minimum

111,12 69,10 111,10

I
Maximin Value
211,14 48,9 211,11
Player A II  r   29,10
214,11 610,9 610,9
III
Column Maximum
211,14 69,10

Minimax Value  r   211,14

904
IJRITCC | April 2016, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 4 Issue: 4 897 - 905
_______________________________________________________________________________________________
 

 

ij ij

As discussed in above example,   r   211,14    r   29,10  0 ,

j  i  i  j 
  ij , ij 
  ij , ij 

The fuzzy optimal solution to the fuzzy game is given by

1) The fuzzy optimal strategy for player A is II and the fuzzy optimal strategy for player B is I
2) The value of the game is 29,10 for player A and 210,9 for player B.

 

 

ij ij

3) Also,   r   r  29,10  0 , ,the fuzzy game is not fair.

j  i  i  j   

  ij , ij 
 ij , ij 

5. Conclusion
In the present paper we considered a fuzzy two-person zero-sum game with fuzzy payoffs having cell entries epsilon-
delta fuzzy numbers. We defined the ordering of a , and b , . The optimal strategies are obtained.
1 1 2 2

References
[1] Bapat, M. S., Yadav, S. N., Kamble, P. N., Triangular approximations of fuzzy numbers, International Journal of
Statistika and Mathematika, ISSN: 2277-2790 E-ISSN:2249-8605, Volume 7, Issue 3, (2013) 63-66.
[2] Bapat, M. S., Yadav, S. N., MAX-MIN Operations on Fuzzy Number,National Conference on Differntial Equation-
2015,Shivaji University Kolhapur,29-30 January 2015
[3] Dubois, D., Prade, H., Fuzzy Sets and Systems: Theory and Applications, Academic Press, Boston, 1980.
[4] Nurmi, H. (1976).on fuzzy games, Eur.Meet.Cybern.Syst.Res.,3rd ,Vienna
[5] Orlovsky,S.A. (1977). On programming with fuzzy constraint sets.Kybernetes 6, pp 197-201
(Reference from II.4.)
[6] Ragade,R.K. (1976). Fuzzy games in the analysis of options.J.Cybern 6, pp213-221
[7] Sharma S D, Operation Research: Theory Methods and Applications, Kedar Nath Ram Nath Publication, 2013, ISBN-13
9789380803388.
[8] Zadeh L. A., Fuzzy sets, Information and Control 8 (1965)338-353

905
IJRITCC | April 2016, Available @ http://www.ijritcc.org
_______________________________________________________________________________________