Sei sulla pagina 1di 75

GAME THEORY

Life is full of conflict and competition. Numerical examples involving adversaries in conflict include games, military battles, political campaigns, advertising and marketing campaigns by competing business firms and so forth. A basic feature in many of these situations is that the final outcome depends primarily upon the combination of strategies selected by the adversary.

Game theory is a mathematical theory that deals with the general features of competitive situations. It places particular emphasis on the decisionmaking processes of the adversaries.
Game theory deals with decisions and situations in which intelligent opponents have conflicting objectives. Game theory is a decision theory in which ones 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.

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.

Basic Terminologies

Game: A competitive situation is called as game if it has the following properties: Finite no. of participants called players, finite no. of alternatives (activities) available to each participant, Every game results in an outcome. Two person Game: involves only two players n-person game: involves n players

Basic Terminologies

Strategy: list of all possible actions available to each player.


Pure Strategy: a player knows exactly what the other player is going to do,

Deterministic situation, Objective is to maximize the gain or to minimize the loss Decision rule to select a particular strategy irrespective of the strategy others may choose

Mixed

Strategy: all players are guessing as to which strategy is to be selected on a particular occasion with some fixed probability,

Probabilistic situation, Objective is to maximize the expected gain or to minimize the expected loss Decision rule to select the combination of strategies with fixed probabilities

Basic Terminologies

Pure Strategy is a particular case of Mixed Strategy: probability of the selected strategy = 1 and probability of other strategies = 0 The outcome resulting from a selection of a particular strategy is also known to the players in advance and is expressed in terms of numerical values. Optimal Strategy: the particular strategy by which a player optimizes his gains or losses without knowing the competitors strategies. Value of the Game: the expected outcome per play when players follow their optimal strategy.

Basic Terminologies

Payoff : outcome associated with each pair of strategies that one player receives from other players, or a quantative measure of satisfaction a player gets at the end of each play Let vi be the payoff to the player Pi; 1 i n, in n person game. Zero Sum Game: one player wins whatever the other one loses, so that the sum of their net winnings is zero. If vi = 0, then the game is said to be a zero sum game. Two person zero sum game: (Rectangular Game): two player game and the gain of one player is equal to the loss of the other player.

In general, a two-person game is characterized by The strategies of player 1. The strategies of player 2. The pay-off matrix.

Thus the two person zero sum game is represented by the payoff matrix to player A as
B1 A1 A2 . . a11 a21 . . am2 . amn B2 a12 a22 ........ Bn a1n a2n

Am am1

Here A1,A2,..,Am are the strategies of player A


B1,B2,...,Bn are the strategies of player B aij is the payoff to player A (by B) when the player A plays strategy Ai and B plays Bj (aij is ve means B got |aij| from A)

Basic assumptions of the Game:


1. Each Player has finite no. of strategies. The list may not be same for each player.

2. Player A attempts to maximize gains and player B attempts to minimize losses.


3. The decisions of both players are made individually, prior to play with no communication between them. 4. The decisions are made simultaneously and also announced simultaneously so that neither player has an advantage resulting from direct knowledge of other players decision.

5. Both the players know not only the possible payoffs to themselves but also of each other.

Simple Example: Consider the game of the odds and evens. This game consists of two players A,B, each player simultaneously showing either of one finger or two fingers. If the number of fingers matches, so that the total number for both players is even, then the player taking evens (say A) wins Rs.1 from B (the player taking odds). Else, if the number does not match, A pays Rs.1 to B. Thus the payoff matrix to player A is the following table:

B 1 1 A 1 -1 2 -1 1

Optimal solution of two-person zero-sum games The solution of the game is based on the principle of securing the best of the worst for each player.

Each player will act so as to maximum his minimum gain or minimum his maximum loss.
Known as minimax criterion

Developed by J. Von Neuman (father of Game theory

Rule:
1.Select the min. element of each row. 2. Select the max element of each column. 3. Select the max element from the column of the min. 4. Select the min element from the row of the max.

Saddle point: position of an element in the payoff matrix which is min in maxima row and max in minima column.
If max min aij min max aij ars
i j j i

then payoff matrix is said to have a saddle point (r,s). Game is with saddle point Optimal Strategy for player A = rth strategy Optimal Strategy for player B = sth strategy Value of Game (v): payoff ars

Fair Game: if v = 0.

Problem: Determine the saddle-point solution, the associated pure strategies, and the value of the game for the following game. The payoffs are for player A .

B1 A1 8 A2 8 A3 7

B2 6 9 5

B3 2 4 3

B4 8 5 5

Row min 2 4 3

Col 8 Max

max min

min max

If the player A plays strategy 1, then whatever strategy B plays, A will get at least 2.
Similarly, if A plays strategy 2, then whatever B plays, will get at least 4. and if A plays strategy 3, then he will get at least 3 whatever B plays. Thus to maximize his minimum return, he should play strategy 2.

Now if B plays strategy 1, then whatever A plays, he will lose a maximum of 8. Similarly for strategies 2,3,4. (These are the maximum of the respective columns). Thus to minimize this maximum loss, B should play strategy 3. And max (row minima)= min (column maxima) = 4 = value of the game.
Position (2,3) is called the saddle-point.

Problem The following game gives As payoff. Determine p,q that will make the entry (2,2) a saddle point.
B1 A1 A2 A3 Col max 1 p 6 max(p,6) B2 q 5 2 max(q,5) B3 6 10 3 10 Row min

min(1,q)
min(p,5) 2

Since (2,2) must be a saddle point,

p 5 and q 5

Problem Specify the range for the value of the game in the following case assuming that the payoff is for player A.
B1 A1 A2 A3 Col max 3 5 4 5 B2 6 2 2 6 B3 1 3 -5 3 Row min

1
2 -5

Thus max( row min) min (column max)


The game has no saddle point.

Thus,

2 value of the game 3.

Here both players must use random mixes of their respective strategies so that A will maximize his minimum expected return and B will minimize his maximum expected loss.
Game without saddle point then Maximin value value of Game Minimax value

Solution of mixed strategy games

Whenever a game does not possess a saddle point, game theory advises each player to assign a probability distribution over his/her set of strategies. Mathematically speaking,

Let xi = probability that player A will use strategy Ai (i = 1,2,m) yj = probability that player B will use strategy Bj (j = 1,2,n)

In this context, the minimax criterion says that a given player should select the mixed strategy that minimizes the maximum expected loss to himself; equivalently that maximizes the minimum expected gain to himself.

Thus player As expected payoff


ai1 xi
i 1
m

when B plays strategy B1 when B plays strategy B2

ai 2 xi
i 1

. . .
ain xi
i 1 m

when B plays strategy Bn

Thus A should
maximize [min{ ai1 xi , ai 2 xi ,... ain xi }]
i 1 i 1 i 1 m m m

where x1+x2++xm=1, xi 0 Similarly B should minimize [max{ a1 j y j , a2 j y j ,... amj y j }]


j 1 j 1 j 1 n n n

where y1+y2++yn=1, y j 0

Graphical solution of mixed strategy games Consider the following problem in which player A has only two strategies. The matrix is payoff matrix for player A:
B1 B2 B3

A1
A2

1
2

-3
4

7
-6

Let x1 be the probability with which player A plays the strategy 1 so that 1-x1 is the probability with which he will play the strategy 2.

As expected payoff when B plays the


Pure strategy B1 is 1 x1 + 2 (1-x1) = - x1 + 2 B2 is -3 x1 + 4 (1-x1) = -7x1+4 B3 is 7 x1 6 (1-x1) = 13x1- 6

Hence he should maximize


min { -x1+2 , -7x1+4 , 13x1-6 }

Now we draw the graphs of the straight lines:


v = -x1+2, v = -7x1+4, v = 13x1-6 for 0 x 1

B2 B1
(1/2)

x1

B3

We find that the minimum of 3 expected payoffs correspond to the lower portion of the graph (marked by vertical lines). Thus the maximum occurs at x1 = and the value of the game is v = (the corresponding ordinate). Now let B play the strategies with probabilities y1, y2, y3.
By the graph above we find B should play the strategy B1 with probability 0 (otherwise A will get a higher payoff).

Thus Bs expected payoff to A are:


1*0- 3y2 + 7(1-y2) = -10y2 + 7 when A plays strategy 1 and 2*0+ 4y2 - 6(1-y2) = 10y2 - 6 when A plays strategy 2 For optimal strategy -10y2 + 7 = 10y2 - 6 or 20 y2 = 13 Therefore y2=13/20 and y3=7/20. Value of the game = -10*(13/20)+7 =1/2???

Problem: The payoff matrix for A is given by


B1 A1 A2 1 -1 B2 -1 1

Find the optimal solution by graphical method. Bs pure strategy As expected payoff

1
2

x1 - (1-x1) = 2x1 - 1
- x1 + (1-x1) = - 2x1+1

B2
(1/2, 0) 1

B1

Thus A and B play the strategies with probabilities 0.5, 0.5 and the value of the game is 0.

Problem
Consider the game having the following pay-off (to A) table:
Player B Strategy 1 Player A 2 -1 2 1 3 2 -2

Use the graphical procedure to determine the value of the game and the optimal strategy for each player according to the minimax criterion. When B plays strategy 1 As Expected pay-off 3 x1-1(1-x1) = -1 + 4x1

-2 x1+2(1-x1) = 2 4x1

B2
(3/8, 1/2) 0 1 x1

B1

Thus A should play his strategies 1, 2 with probabilities 3/8, 5/8. And the value of the game = 1/2. Now if Bs optimal strategies are y1* and (1-y1*), then
3 y1* - 2(1- y1* ) = 1/2 Or 5 y1* = 5/2 , i.e., y1* = 1/2 = y2* Thus B should play his strategies 1, 2 with probabilities 1/2, 1/2.

Q.

B1 B2 B3 A1 0 0 0

B4 0

B5 0

B6 0

A2 4
A3 4 A4 4 A5 4 A6 4

2
3 3 3 3

0
1 7 4 -1

2
3 -5 -1 -2

1
2 1 2 2

1
2 2 2 2

Dominance Rule:

1. If all the elements of a row (kth) the corresponding element of any other row (rth), then kth row is dominated by rth row. 2. If all the elements of a column (kth) the corresponding element of any other column (rth), then kth column is dominated by rth column.

3. Omit dominated rows and columns.


4. If some linear combination of some rows dominates ith row, then ith row will be deleted. 5. Similar argument follows for column.

B1 B2 B3 A1

B4

B5

B6

A2
A3 4 A4 4 A5 4 A6 3 3 3 1 7 4 3 -5 -1 2 1 2 2 2 2

B1 B2 B3 A1

B4

B5

B6

A2
A3 4 A4 4 A5 4 A6 3 3 3 1 7 4 3 -5 -1 2 1 2 2 2 2

B1 B2 B3 A1

B4

B5

B6

A2
A3 A4 A5 A6 1 7 4 3 -5 -1 2 1 2

B1 B2 B3 A1

B4

B5

B6

A2
A3 A4 A5 A6 1 7 4 3 -5 -1

B1 B2 B3 A1

B4

B5

B6

A2
A3 A4 A5 A6 1 7 3 -5

Thus As expected payoff to B are:


0x1 + 0x2+x3+7x4+0x5+0x6 = v 0x1 + 0x2+3x3-5x4+0x5+0x6 = v x3+x4 = 1 And Bs expected payoff are: 0y1 + 0y2+y3+3y4+0y5+0y6 = v 0y1 + 0y2+7y3-5y4+0y5+0y6 = v y3+y4 = 1

Therefore x3=6/7, x4= 1/7 and v = 13/7 and y3=4/7 and y4=3/7. Value of the game = v =13/7

Problem
Consider the game having the following pay-off (to A) table: Strategy

Player B 1 2
1 2 3 4 1 0 3 -3 -1 4 -2 6

3 3 1 5 -2

Player A

Use the graphical procedure to determine the value of the game and the optimal strategy for each player.

It is clear that the strategy 3 of player B is dominated by the strategy 1 of player B, in the sense that Bs pay-off to A is less whatever strategy A plays. (Note 1 < 3, 0 < 1, 3 < 5, -3 < -2). Thus B should never play the strategy 3 and the pay-off matrix becomes

Player B
Strategy 1 2 3 4 1 1 0 3 -3 2 -1 4 -2 6

Player A

Thus we first solve Bs problem graphically.

Let B play the strategies 1, 2 with probabilities y1 and (1-y1). Thus


When A plays strategy Bs expected pay-off (to A)

y1 (1-y1) = -1 + 2y1

2
3

4(1-y1) = 4 - 4y1
3y1 2(1-y1) = -2 + 5y1

-3y1 +6 (1-y1) = 6 - 9y1

A4
A2

(2/3, 4/3)
y1

A1 A3

Thus B should play strategy 1 with prob 2/3 and strategy 2 with prob 1/3 and the value of the game = 4/3.

Now we find As prob distribution. From the graph we see that A should never play the strategy 1. Thus x1 = 0, x4= 0. If x2*, x3* are the optimal prob with which he plays the strategies 2, 3, then we find
Then, y1 = 2/3, y2 = 1/3. v = 4/3 And x3* = 4/9 and so x2* = 5/9

Question: In the given example, for player A, strategy A3 is dominated by the strategy A2 and so can be eliminated.
B1 B2 B3 B4

A1 8
A2 8

6
9

2
4

8
5

A3 7

Eliminating the strategy A3 , we get the

following reduced payoff matrix:


B1 A1 8 A2 8 B2 6 9 B3 2 4 B4 8 5

Now , for player B, strategies B1, B2, and B4 are dominated by the strategy B3. Eliminating the strategies B1 , B2, and B4 we get the reduced payoff matrix:

following reduced payoff matrix:


B3 A1 A2

2
4

Now , for player A, strategy A1 is dominated by the strategy A2. Eliminating the strategy A1 we thus see that A should always play A2 and B always B3 and the value of the game is 4.

Remark: If there are more than two lines passing though the maxmin (minmax) point P, this would imply that alternative optimal mixed strategy exist for the player.

Any two lines having opposite signs of their slopes will define an alternative optimum solution.

Problem
Consider the game having the following pay-off (to A) table: Strategy Player A 1 2 1 1 0

Player B 2 3
-1 4 3 1

4 -1 6

Use the graphical procedure to determine the value of the game and the optimal strategy for each player.

Let A play the strategies 1, 2 with probabilities x1 and (1-x1). Thus


When B plays strategy As expected pay-off

2x1 + 4(1-x1) = 4- 2x1

2
3

2x1 + 3(1-x1) = 3- x1
3x1 + 2(1-x1) = 2 + x1

-x1 +6 (1-x1) = 6 7x1

B4 B1
(1/2)

B2 B3

x1

We find that the minimum of 3 expected payoffs correspond to the lower portion of the graph (marked by vertical lines). Thus the maximum occurs at x1 = and the value of the game is v = 5/2 (the corresponding ordinate). Now let B play the strategies with probabilities y1, y2, y3, y4.
By the graph above we find B should play the strategy B1 with probability 0 (otherwise A will get a higher payoff).

Thus Bs expected payoff to A are:


0y1 + 0y2 +3y3 - 1(1-y3) = 4y3 -1 when A plays strategy 1 and 0y1 + 0y2 +2y3 + 6(1-y3) = -4y3 +6 when A plays strategy 2 For optimal strategy 4y3 - 1 = -4y3 + 6 Therefore y3=7/8 and y4= 1/8. Value of the game = 5/2
Alternative Solution Exist: y2=1/2 and y3= 1/2

From the graph, we see that A should never play the strategy 1. Thus x1 = 0, x4= 0. If x2*, x3* are the optimal prob with which he plays the strategies 2, 3, then we find
Then, y1 = 2/3, y2 = 1/3. v = 4/3 And x3* = 4/9 and so x2* = 5/9

Solution by LP method As problem is:


Max [min{ ai1 xi , ai 2 xi ,... ain xi }]
i 1 i 1 i 1 m m m

where x1+x2++xm=1, xi 0

Similarly Bs Problem is:


Min [max{ a1 j y j , a2 j y j ,... amj y j }]
j 1 j 1 j 1 n n n

where y1+y2++yn=1, y j 0

Let

v [min{ ai1 xi , ai 2 xi ,... ain xi }]


i 1 i 1 i 1

This implies

a
i 1

i1 i

x v

for all j = 1,2,n

Thus As problem becomes Maximize z = v m Subject to v aij xi 0, j 1, 2,...n


i 1

x1 x2 ... xm 1, x j 0, v unrestricted in sign

In Bs Problem, Putting
v max aij y j
i n

Bs problem becomes
Minimize w = v Subject to v a y 0, i 1, 2,..., m ij j
j 1 n

j 1

y1 y2 ... yn 1, yi 0, v unrestricted in sign

We easily see that Bs (LP) problem is the dual of As (LP) problem. Hence the optimal solution of one problem automatically yields the optimal solution of the other.
Problem Solve the following problem by LPP
B1 A1 A2 2 0 B2 0 0 B3 0 4

A3

Note that max (Row Min) = 0 and min (column Max) = 2.

Thus the game has no saddle point and we have to go in for mixed strategies.

Thus As problem is:


Maximize z = v Subject to

v 2 x1 0 v 3x3 0 v 4 x2 0 x1 x2 x3 1 x j 0, v unrestricted in sign

And Bs problem is:


Minimize w = v

Subject to v 2 y1 0

v 4 y3 0 v 3 y2 0 y1 y2 y3 1 yi 0, v unrestricted in sign
We now solve As problem by two phase method.

Phase-I
Basic

r s1 s2 s3 R1 r s1 s2 s3 x1

r 1 0 0 0 0 1 0 0 0 0

v+ 0 1 1 1 0 0 1 1 1 0

v0 -1 -1 -1 0 0 -1 -1 -1 0

x1 01 -2 0 0 1 0 0 0 0 1

x2 01 0 0 -4 1 0 2 0 -4 1

x3 01 0 -3 0 1 0 2 -3 0 1

s1 0 1 0 0 0 0 1 0 0 0

s2 0 0 1 0 0 0 0 1 0 0

s3 0 0 0 1 0 0 0 0 1 0

R1 -10 0 0 0 1 -1 2 0 0 1

Sol

01 0 0 0 1 0 2 0 0 1

Phase - II
Basic

z s1 s2 s3 x1 z s1 v+ s3 x1

z 1 0 0 0 0 1 0 0 0 0

v+ -1 1 1 1 0 0 0 1 0 0

v1 -1 -1 -1 0 0 0 -1 0 0

x1 0 0 0 0 1 0 0 0 0 1

x2 0 2 0 -4 1 0 2 0 -4 1

x3 0 2 -3 0 1 -3 5 -3 3 1

s1 0 1 0 0 0 0 1 0 0 0

s2 0 0 1 0 0 1 -1 1 -1 0

s3 R1 0 -10 0 0 0 0 1 0 0 1 0 0 0 1 0

Sol

0 2 0 0 1 0 2 0 0 1

Phase II (continued)
Basic

z s1 v+ x3 x1 z x2 v+ x3 x1

z 1 0 0 0 0 1 0 0 0 0

v+ 0 0 1 0 0 0 0 1 0 0

v0 0 -1 0 0 0 0 -1 0 0

x1 x2 x3 s1 s2 s3 R1 Sol 0 -4 0 0 0 1 -10 0 0 26/3 0 1 2/3 -5/3 0 2 0 1 0 0 0 -4 0 0 0 -4/3 1 0 -1/3 1/3 0 0 1 7/3 0 0 1/3 -1/3 1 1 12/13 0 0 0 6/13 4/13 3/13 4/13 3/13 3/13 0 1 0 3/26 1/13 -5/26 -5/26 12/13 0 0 0 6/13 4/13 8/13 4/13 0 0 1 2/13 1/13 1/13 6/13 1 0 0 -7/26 2/13 3/26

This is the optimal table and the optimal solution is:


x1 = 6/13, x2 = 3/13, x3 = 4/13 From the optimal table we also read the optimal solution of Bs problem as: y1 = 6/13, y2 = 4/13, y3 = 3/13

And the value of the game is : v = 12/13