Mexican Journal of Operations Research, Vol. 2, No. 2, Jul-Dec 2013, pp. 10-20. ISSN: 2007-5138.

ALTERNATIVE APPROACH TO GAME THEORY PROBLEMS

Navneet Kumar Lamba
Assistant Professor, Department of Mathematics,
Priyadarshini Institute of Engineering and Technology, Nagpur, INDIA.
Email-navneetkumarlamba@gmail.com

Abstract. In this paper, a new approach to the solution of Game theory problem is suggested,
which is based on the iterative procedure. The proposed alternative method is computationally
more effective and easy as compared to the traditional solution of Game problems.

Key words and phrases: Optimum Solution, Game theory, Simplex method.

1. INTRODUCTION
Antoine Cournot (1838) described the earliest example of a formal game-theoretic analysis in the study of a
duopoly. Emile Borel (1921) suggested a formal theory of games in, which was furthered by the John von
Neumann (1928) in a theory of parlor games.

John Nash (1950) demonstrated that finite games always have an equilibrium point, at which all players choose
actions which are best for them given their opponents choices. Game theory is the formal study of conflict and
cooperation. The concepts of game theory provide a language to formulate structure, analyze, and understand
strategic scenarios.

In the present paper, an attempt has been made to solve the game theory problems by the alternative approach
method which provides better result as compared to previous one.

Received Feb 20, 2013 / Accepted Sep 25, 2013

Editorial Acadmica Dragn Azteca (EDITADA.ORG)

Navneet Kumar Lamba / Alternative approach to game theory problems. MJOR, Vol. 2, No. 2, Jul-Dec
2013, pp. 10-20. EDITADA. ISSN: 2007-5138.

2. NEW ALGORITHM APPROACH

Step 1. For the

m n rectangular game when either m or n or both are greater than equal to three , linear

programming approach is as follows:-

Let the two person zero sum game be defined as follows:

Player A has m course of action

A1 , A2 ,........., Am and

player B has n course of the action

B1, B2 ,........., Bn . The pay-off to the player A if he selects strategy

Mixed strategy for player A is defined by the probabilities
mixed strategy for player B is defined by

Ai and player B select

p 1 ,.......... p m , where p1 .......... p m 1 and

q 1 ,..........q n where q 1 .......... q n 1 .

Let the game can be defined as a linear programming problem as given below:

Player A
Minimize

1
or

y1 y2 ........... yn

Subject to the constraints:

a11 y 1 a12 y 2 .......... ..... a1n y n 1

a 21 y 1 a 22 y 2 .......... ..... a 2n y n 1
.......... .......... .......... .......... .......... ....
.......... .......... .......... .......... .......... ....
a m1 y 1 a m2 y 2 .......... ..... a m n y m 1
Player B
Maximize

1
or x1 x2 ...........xn

Subject to the constraints:

a11 x1 a12 x 2 .......... ..... a1n x n 1

a 21 x1 a 22 x 2 .......... ..... a 2n x n 1
.......... .......... .......... .......... .......... ....
.......... .......... .......... .......... .......... ....
a m1 x 1 a m2 x 2 .......... ..... a m n x m 1

11

B j is aij .

Navneet Kumar Lamba / Alternative approach to game theory problems. MJOR, Vol. 2, No. 2, Jul-Dec
2013, pp. 10-20. EDITADA. ISSN: 2007-5138.
The steps for the computation of the optimal solution are as follows:
Step 2: Formulate the linear programming model of the real world problem that is obtained a mathematical
representation of the problems objective function and constraints as stated below.

M c1 x1 c2 x2 .......... cn xn
Subject to constraints:

Maximize

a11x1 a12 x2 .......... a1n xn b1

a21x1 a 22 x2 .......... a2n xn b2
.......... .......... .......... .......... .......... ......
.......... .......... .......... .......... .......... ......
a m1 x1 a m2 x 2 .......... a m n x n bm

x1 , x2 , x3 ,......... .., xn 0
If the objective function is minimized, then convert it into a problem of maximizing by using the rule
Minimum M= -(Maximum (-M))
all b is , i 1,2,......,m must be non negative. If any one of

bi is negative, multiply corresponding inequality

by (-1), So as to get all b is , i 1,2,......, m non-negative.

Step 3: Convert all inequations of the constraints into the equations by introducing slack variables in the left
hand side of constraints and assign a zero coefficient to the corresponding variable in the objective function.
Thus we can reformulate the problem in terms of equation as follows: Maximize M c1 x1 c2 x2
Subject to constraints:

.......... cn xn 0 p1 0 p2 ....... 0 pm

a11x1 a12 x2 .......... a1n xn p1 b1

a 21x1 a 22 x2 .......... a 2n xn p 2 b2
.......... .......... .......... .......... .......... .......... ........
.......... .......... .......... .......... .......... .......... ........
a m1 x1 a m2 x 2 .......... a m n x n p m bm
where x1 , x 2 , x3 ,......... .., x n 0 and p1 , p 2 , p3 ,......... .., p m 0

12

Navneet Kumar Lamba / Alternative approach to game theory problems. MJOR, Vol. 2, No. 2, Jul-Dec
2013, pp. 10-20. EDITADA. ISSN: 2007-5138.
Step 4: An initial basic feasible solution is obtained by setting x1
Thus we get p1

x2 x3 ........... xn 0 .

b1 , p2 b2 ,, pm bm .

Step 5: For computational, efficiency and simplicity, the initial basic feasible solution, the constraint of the
standard linear programming problem as well as the function can be displayed in a tabular form, called the new
simplex tableau as solve below:NEW TABLEAU:
Row(R)

Basi
c

y1

y2

Initial (R*)

c1

c2

First (R1)

p1

a11

a12

Second (R2)

p2

a21

a22

a m1

a m2

mth
Row (Rm)

pm

yn

p1

p2

...

pm

R.H.S

cn

a1n

b1

a 2n

b2

am n

bm

The interpretation of data of the above tableau is given as under: (i).The Initial row (R*), called the objective row of the new simplex table indicates the Coefficients of variables
(m+n) in the objective function.
(ii).From 1st to mth row is called constraint row of the table represents the coefficient of the constraints.
(iii). The first column labeled (R) denotes the m rows also known as objective column, the second column
labeled Basic variable points out the basic variables, and in the initial tableau these variable are the slack
variables, the third column indicates the coefficient of z whose value for the first row is 1 and for all other
rows it is zero.
(iv). The body matrix (under non-basic variables) in the initial tableau consists of the decision variable in the
constraint set.
(v). The Identity matrix in the initial tableau represents the coefficients of the slack variables that have been
added to the original inequality to make them equations.
Step 6:Test the Solution for optimality: Examine the initial row (R*) of the above tableau.
(i). If all the elements in the initial row (R*) are positive then the current solution is optimal.
(ii). If there exists some negative number, the current solution can be improved by entering the column, which
contain less negative term, let it be y r where

p2 . Then

(a). Select the maximum positive term in the corresponding column of y r , let it be a rj , 1 j n , which is
13

Navneet Kumar Lamba / Alternative approach to game theory problems. MJOR, Vol. 2, No. 2, Jul-Dec
2013, pp. 10-20. EDITADA. ISSN: 2007-5138.
our new pivot element and find out the corresponding identity element, let it be lies in the column p k
where 1 k m , next drop this column p k .
(b). Repeat the above process till all the coefficient of the Initial Row (R*) becomes positive, if any one
coefficient of the row R1 is negative then repeat the above process to find the optimum solution.

3. STATEMENT OF THE PROBLEM

Problem 1. Two companies A and B are competing for the same product. Their different strategies are given in
the following pay-off matrix:

Company B
Company A

1
2

4
3
4

7
7

Use new alternative method to determine the best strategies for both players.
Solution. The given game does not posses the saddle point, and its value lies from 2 and +3, Thus adding a
constant k=3 to all the elements of pay-off matrix to matrix, then the pay-off matrix become:

Company B
Company A

4
1

1
5

0 4
7
10

Let the strategies of the two players be

A1
SA
p1

A2
p2

A3
A1
, SB
p3
q1

A2
q2

A3

q3

Where p 1 .......... p m 1 and q 1 .......... q n 1

Maximize Z

or

y1 y 2 y3

Subject to the constraints:

6 y1 y 2 5 y 3 1

4 y1 4 y 3 1

y 1 7 y 2 10 y 3 1 and y j 0 , j 1,2,3

Minimize Z

or

x1 x2 x3

Subject to the constraints:

14

Navneet Kumar Lamba / Alternative approach to game theory problems. MJOR, Vol. 2, No. 2, Jul-Dec
2013, pp. 10-20. EDITADA. ISSN: 2007-5138.

6 x1 4 x 2 x 3 1
x1 7 x 3 1

5x1 4x 2 10 x 3 1 and x j 0 , j 1,2,3

By introducing slack variables p1 0 , p2 0 and p3 0 respectively, the set of constraints of the LPP
are converted into the system of equations; the iterative New tableau are:
Initial table:
Row

Basic

y1

y2

y3

p1

p3

p2

RHS

R1

-1

-1

-1

R2

p1

-1

R3

p2

-4

R4

p3

10

Where p1 , p2 and p 3 denotes the slack variables and R1, R2, R3, R4 represents the first, second, third and
fourth row respectively. From the above table it clear that the least negative coefficient of z lies in column

y1 , y 2 and y 3 which is 1(either we can choose any one), let consider y 2 (since the no. of iteration to
obtain the solution becomes less), which will enters in the basis, pivot element is 7(arbitrary in column y 2 )
and the corresponding identity element lies in the p 3 column, hence we will drop column vector p 3 .
First Iteration: Introduce y 2 and drops p 3 , we get
Row

Basic

y1

y2

y3

p1

p3

p2

RHS

R1

-6/7

3/7

1/7

1/7

R2

p1

43/7

45/7

1/7

8/7

R3

p2

-4

R4

y2

1/7

10/7

1/7

1/7

Since y1 in the initial row (R1) is only one negative i.e. it enters the basis and corresponding pivot element is
43/7, next drops the corresponding identity element which lies in column p1 .

15

Navneet Kumar Lamba / Alternative approach to game theory problems. MJOR, Vol. 2, No. 2, Jul-Dec
2013, pp. 10-20. EDITADA. ISSN: 2007-5138.
Second Iteration: Introduce y1 and drops p1 , we get
Row

Basic

y1

y2

y3

p1

p2 p 3

RHS

R1

57/43

6/43

7/43

91/301

R2

y1

45/43

7/43

1/43

8/43

R3

p2

-352/43

-28/43

-4/43

11/43

R4

y2

55/43

-1/43

6/43

5/43

Since all the elements in the initial row are positive then the current solution is optimal, and its value is

1
8
5
x1 x2 x3
0
x1 75 / 4 , x2 50 / 4 and

43
43
43
Therefore the value of the game for the modified matrix,
13 ,
qi
y j , qi y j
Since

8 43 8
5 43 5
q1 y1

q2 y 2

q y3 0
,
43 13 13
43 13 13 and 3
Company As best strategies appear in the initial row R1, under
positive signs.
Therefore x1

Thus

p1 , p2 and p 3 respectively with

6
7
, x2 0 and x3
43
43

p1 x1

6 43 6

, p2 x2 0 and p3 x3 7 43 7
43 13 13
43 13 13

Hence optimal strategies for company A: - are (6/13, 0, 7/13),

for company B:- are ( 8/13 , 5/13 ,0) and the value of game is :- 43/13-3 = 4/13.
Problem 2. Solve the following game by using new alternative technique:

Player B
Player A

3
6

1
3

5 3
2 2

Solution. Add a suitable constant to make all the entries of the above payoff matrix to ensure them all positive.
Thus, adding the constant k = 4 to each element, we get the payoff matrix:

Player B
16

Navneet Kumar Lamba / Alternative approach to game theory problems. MJOR, Vol. 2, No. 2, Jul-Dec
2013, pp. 10-20. EDITADA. ISSN: 2007-5138.

Player A

7
10

1
2

3
9
6

Let the strategies of the two players be

A1
SA
p1

A2
p2

A3
A1
, SB
p3
q1

A2
q2

A3

q3

where p 1 .......... p m 1 and q 1 .......... q n 1

Maximize Z

x1 x2 x3

or

Subject to the constraints:

5x1 7 x 2 10 x 3 1

3x1 9 x 2 6 x 3 1

7 x1 x 2 2x 3 1 and x j 0 , j 1,2,3

Minimize Z

or

y1 y 2 y3

Subject to the constraints:

5 y1 3 y 2 7 y 3 1
7 y1 9 y 2 y 3 1

10 y1 6 y 2 2 y 3 1 and y j 0 , j 1,2,3
where x j

j 1,2,3 ; u is minimum expected gain to A and is the minimum expected

,

yj

pj
and

loss to B.
By introducing slack variables p1 0 , p2 0 and p3 0 respectively, the set of constraints of the LPP are
converted into the system of equations; the iterative New simplex tableau are:
Initial table:
Row

Basic

y1

y2

y3

p1

p2

p3

RHS

R*

-1

-1

-1

R1

p1

R2

p2

R3

p3

10

Where p1 , p2 and p 3 denotes the slack variables and R*, R1, R2, R3 represents the initial, first, second and
third row respectively. From the above table it clear that the least negative coefficient of z lies in column

y1 , y 2 and y 3 which is (1)(either we can choose any one), let consider y 3 (since the no. of iteration
17

Navneet Kumar Lamba / Alternative approach to game theory problems. MJOR, Vol. 2, No. 2, Jul-Dec
2013, pp. 10-20. EDITADA. ISSN: 2007-5138.
require to obtain the solution becomes less), which will enters in the basis, pivot element is 7(arbitrary in
column y 3 ) and the corresponding identity element lies in the p1 column, hence we will drop column
vector p1 .
First Iteration: Introduce y 3 and drops p1 , we get

Row

Basic

y1

y2

y3

p1

p3

p2

RHS

R*

-2/7

-4/7

1/7

1/7

R1

y3

5/7

3/7

1/7

1/7

R2

p2

44/7

60/7

-1/7

6/7

R3

p3

60/7

36/7

-2/7

5/7

Since y1 in the initial row (R*) is next least negative i.e. it enters the basis and corresponding pivot element is
60/7, next drops the corresponding identity element which lies in column p 3 .
Second Iteration: Introduce y1 and drops p 3 , we get
Row

Basic

y1

y2

y3

p2 p 3

p1

RHS

R*

-2/5

2/15

1/30

1/6

R1

y3

1/6

-1/12

1/12

R2

p2

24/5

1/15

-11/15

1/3

R3

y1

3/5

-1/30

7/60

1/12

Since y 2 in the initial row (R1) is next least negative i.e. it enters the basis and corresponding pivot element is
3/5, next drops the corresponding identity element which lies in column y1 .
Third Iteration: Introduce y 2 and drops y1 .
Row

Basic

y1

y2

y3

p1

p2

p3

RHS

R*

2/3

1/9

I/9

2/9

R1

y3

1/6

-1/12

1/12

R2

p2

-8

1/3

-5/3

-1/3

R3

y2

5/3

-1/18

7/36

5/36

Since all the elements in the initial row (R*) are positive then the current solution is optimal,

and its value is

x1 x2 x3 0 5 / 36 1 / 12 2 / 9
x2 5 / 36 , x3 1 / 12 and

18

Navneet Kumar Lamba / Alternative approach to game theory problems. MJOR, Vol. 2, No. 2, Jul-Dec
2013, pp. 10-20. EDITADA. ISSN: 2007-5138.

Therefore the value of the game for the modified matrix,

Since

qi

9
2,

y j , qi y j

q1 y1 0 , q2 y2

5 9 5
1 9 3
and q3 y3

36 2 8
12 2 8

Company As best strategies appear in the initial row R*, under

positive signs.
Therefore x1

p1 , p2 and p 3 respectively with

1
1
, x2 0 and x3
9
9

Thus p1 x1

1 9 1
1 9 1
, p2 x2 0 and p3 x3
9 2 2
9 2 2

Hence optimal strategies for company A are (0, 5/8, 3/8),

for company B are ( 1/2 , 0 ,1/2) and the value of game is :- 9/2-4 = 1/2.

4. CONCLUSION
We observed that the solution of Game Theory problem has been obtained by new alternative technique very
easily and requires less or at the most equal number of iterations than traditional simplex method. This
technique is very useful to apply on numerical problems, reduces the labour work, and provides more accuracy
and improved optimum solution. Therefore this alternative approach is more powerful in solving Game Theory
problems as compare to traditional method.

REFERENCES
[1] Brown, G.: Iterative Solution of Games by Fictitious Play," in Activity Analysis of Production and
Allocation, John Wiley & Sons, New York, 1951.
[2] Borgers, T.: `Weak Dominance and Approximate Common Knowledge,' Journal of Economic Theory Vol.
64, No. 1 (1994) 265-276.
[3] Fudenberg, D. and Levine, D.: The Theory of Learning in Games, Boston: MIT Press, 1998.
[4] O'Neill, B.: Non metric Test of the Minimax Theory of Two-person Zero-sum Games," Proceedings of the
National Academy of Sciences, 84, (1987), 2106-2109.
[5] Rasmussen, E.: Games and information: an introduction to game theory. Wiley-Blackwell; 4th Edition
edition, 2006.

19

Navneet Kumar Lamba / Alternative approach to game theory problems. MJOR, Vol. 2, No. 2, Jul-Dec
2013, pp. 10-20. EDITADA. ISSN: 2007-5138.
[6] Skyrms, B.: Evolution of the Social Contract. Cambridge University Press: Cambridge, 1996.
[7] Stinchcombe, M. B.: General Normal Form Games. Working paper, Department of Economics, University
of Texas at Austin, 2001.
[8] Tang F.F: Anticipatory Learning in Two-person Games: Some Experimental Results," Journal of Economic
Behavior and Organization, Vol. 44 (2001) 221-232.
[9] Von Neumann, J., Morgenstern, O.: The Theory of Games and Economic Behavior, Princeton: Princeton
University Press, 1944.
[10] Weibull, J: Evolutionary Game Theory, Cambridge: MIT Press, 1995.
[11] Lamba, N.K., Khobragade, N.W and Khot, P.G: Alternative Approach to Wolfes Modified Simplex
Method for Quadratic Programming Problems, Int. J Latest Trend Math, Vol. 2, No. 1 (2012) 19-24.

20