Theory of Games
Unit-I
Prof. Narinder Verma
School of Business Management prof Narinder Verma2
Strategic Behavior- Uncertain
A lot of what we do, involves optimizing against various alternatives:
What should I do after 10+2?
Should I do MBA after graduation?
What should be my major specialization?
What kind of job be my dream job?
Should I join family business?
School of Business Management prof Narinder Verma3
Strategic Behavior- Uncertain
Or at times against nature in general:
Should I take an umbrella today?
What crops should I plant this season?
How do we treat this disease or injury?
How do I fix my out of order car?
We sometimes imagine it as a game against opponents
School of Business Management prof Narinder Verma4
Prisoners DilemmaTwo suspects arrested for a crime were put into
prison separately.
prisoners decide whether to confess or not.
If both confess, both sentenced to 5 years of jail
If neither confesses, both will be sentenced to 1 year of jail
If one confesses and the other does not, then the confessor goes free (no jail) and the non-
confessor sentenced to 9 years of jail
What should each prisoner do?
School of Business Management prof Narinder Verma5
Battle of Seaes
A couple deciding how to spend the evening
Wife would like to go for a movie
Husband would like to go for a cricket match
Both however want to spend the time together
Scope for strategic interaction
School of Business Management prof Narinder Verma6
Understanding the Game
Game refers to a situation of conflict
and competition in which two or more
competitors are involved in the
decision-making process in anticipation
of certain outcomes over a period of
time.
1. Competitors are called players
2. Action is called Strategy
3. Anticipated outcome is the payoff
School of Business Management prof Narinder Verma7
Understanding the Game
1. pricing of products is affected by
the price of the competitor.
2. Success of TV programme
depends upon the presence of
competing TV programmes in the
same time slot.
A basic feature here is that the final
outcome depends primarily upon
the combination of strategies
School of Business Management prof Narinder Verma8
Game theory is a series of mathematical
models that deal with interactive
decision-making situations under the
conditions of conflict and competition.
The study of oligopolies
The study of cartels; e.g. OPEC
The study of military strategies.
Study of effect of promotional campaigns
Defining Game Theory
School of Business Management prof Narinder Verma9
Game Theory
Game theoretic notions go back thousands of years
Talmud and Sun Tzu's writings.
Modern theory credited to John von Neumann and Oskar Morgenstern 1944.
Theory of Games and Economic Behavior.
John Nash (A Beautiful Mind fame) generalized these results and provided the basis of the modern field.
School of Business Management prof Narinder Verma10
Classification of Games
2-person Game:
A game that involves exactly two players
n-person Game
A game that involves exactly n players
Zero-sum Game:
When the sum of gains to one player is exactly equal to the losses to other player, so that the sum of the gains and the losses is equal to zero(0).
Non-zero sum Game:
A game whose sum is not equal to zero
School of Business Management prof Narinder Verma11
Classification of Games
Simultaneous Move Game:
Each player has to take action simultaneously like Racing, Stone-paper-Scissors, Chidiya Ud etc
Sequential Move Game
One player moves
Second player observes and then moves
Like Chess, Tic-Tac-Toe (naughts and crosses), Discuss Throw, promotional campaigns by competitors
School of Business Management prof Narinder Verma12
Elements of a Game
A game consists of:
A set of players
A set of strategies for each player
The payoffs to each player for every possible strategy
Strategy is a course of action that a
player adopts for every payoff
School of Business Management prof Narinder Verma13
Strategies of Games
Optimal Strategy:
The particular strategy that optimizes a players gains or losses, without knowing the competitors courses of action
Pure Strategy:
A particular strategy that a player chooses
to play again and again regardless of other
players strategies
It is a deterministic situation
School of Business Management prof Narinder Verma14
Strategies of GamesMixed Strategy:
A set of strategies that a player chooses on a particular move of the game with some fixed probabilities
It is a probabilistic situation
Value of the Game:
It is the expected gain or loss in a game
when a game is played a large number of
times
It is represented by V
School of Business Management prof Narinder Verma15
payoff matrix to player A
a11 a12 a1n
a21 a22 ........ a2n
.
. aij
am1 am2 . amn
A1
A2
.
.
Am
Two-person Zero-sum Game
B1 B2 Bn
Strategies of player BS
tra
teg
ies
of
pla
yer
A
School of Business Management prof Narinder Verma16
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 Aiand B plays Bj (aij is ve means B got |aij| from A)
Two-person Zero-sum Game
School of Business Management prof Narinder Verma17
Two boutiques A and B in Solan compete
such that each gets 50% of clientele. As gain would be Bs loss. B moots an idea to gain market share by:
1. Discount coupons - B1
2. Decreasing price - B2
3. TV Advertisement - B3
4. 10% Cash Back - B4
Two-person Zero-sum Game
School of Business Management prof Narinder Verma18
A has an idea that B might do this. So A
also moots an idea to gain market share
by:
1. Discount coupons - A1
2. TV Advertisement - A2
3. 10% Cash Back - A3
The payoff matrix in Rupees for boutique
A is given as:
Two-person Zero-sum Game
School of Business Management prof Narinder Verma19
8 -6 2 1
4 9 4 5
7 -5 3 -7
B1 B2 B3 B4
A1
A2
A3
Strategies of player B
Str
ate
gie
s o
f
pla
yer
A
1. Find the optimal strategies for both A and B.
2. Find the value of the game.
School of Business Management prof Narinder Verma20
Maximin and Minimax principle
Maximin principle:
For each row, find the minimum, then
The maximum out of these minimums is the Maximin value for A
Minimax principle:
For each column, find the maximum, then
The minimum out of these maximums is the Minimax value for B
School of Business Management prof Narinder Verma21
Maximin and Minimax principle
Saddle point:
When Maximin value is equal to the Minimax value then the game is said to have an equilibrium point.
Equilibrium point is called the Saddle point.
The corresponding strategies are called optimal strategies. These are the pure ones.
At saddle point,
Maximin = Minimax = Value of Game (V)
School of Business Management prof Narinder Verma22
Maximin and Minimax principleA game may have more than one saddle point.
A game may not have a saddle point at all.
In general,
Maximin Value V Minimax Value
A game is said to be a fair game if the
Maximin = Minimax = 0 (zero)
A game is said to be strictly determinable if
the Maximin = Minimax = Value of the game (V)
School of Business Management prof Narinder Verma23
8 -6 2 1
4 9 4 5
7 -5 3 -7
B1 B2 B3 B4
A1
A2
A3
MaxCol 8 9 4 5
Row min
-6
4
-7
minimax
maximin
Illustration Continues
School of Business Management prof Narinder Verma24
Solution is based on the principle of securing
the best of the worst for each player. If the
player A plays strategy A1, then whatever
strategy B plays, A will get at least -6 (loses
at most Rs. 6).
Thus to maximize its minimum returns, A
should play strategy A2.
If A plays strategy A2, then whatever B
plays, will get at least 4. and if A plays
strategy A3, then he will get at least -7(loses
at most Rs. 7) whatever B plays.
School of Business Management prof Narinder Verma25
Now if B plays strategy B1, then whatever
A plays, he will lose a maximum of 8.
Similarly for strategies B2,B3,B4. (These
are the maximum of the respective
columns).
Maximin = Minimax = 4 = Saddle point
Thus here, 4 is value of the game and
appropriate strategies are A2, B3
Thus to minimize this maximum
loss, B should play strategy B3.
School of Business Management prof Narinder Verma26
Two boutiques A and B in Solan compete
such that each gets 50% of clientele. As gain would be Bs loss. B moots an idea to gain market share by:
1. Discount coupons - B1 2. Decreasing price - B23. TV Advertisement - B34. 10% Cash Back - B45. Newspaper Inserts - B5
Illustration 2
School of Business Management prof Narinder Verma27
A has an idea that B might do this. A also
moots an idea to gain market share by:
1. Discount coupons - A12. Decrease price - A23. TV Advertisement - A34. 10% Cash Back - A4
The payoff matrix in Rupees for boutique
A is given as:
Illustration 2
School of Business Management prof Narinder Verma28
3 -1 4 6 7
-1 8 2 4 12
16 8 6 14 12
1 11 -4 2 1
B1 B2 B3 B4
A1
A2
A3
A4
Strategies of player B
Str
ate
gie
s o
f
pla
yer
A
1. Find the optimal strategies for both A and B.
2. Find the value of the game.
B5
School of Business Management prof Narinder Verma29
maxCol 16 11 6 14 12
Row min
-1
-1
6
-4
minimax
maximin
Illustration 2
3 -1 4 6 7
-1 8 2 4 12
16 8 6 14 12
1 11 -4 2 1
B1 B2 B3 B4 B5
A1
A2
A3
A4
School of Business Management prof Narinder Verma30
Illustration 3
The following game gives As payoff. Determine p, q that will make the entry a22 a saddle point.
1 q 6
p 5 10
6 2 3
A1
A2
A3
B1 B2 B3
Col max 6 5 10
Row min
1
5
2
School of Business Management prof Narinder Verma31
Since a22 must be a saddle point,
There for p has to be at least as large
as 5, so
And also q has to be at least as small
as 5, so
5p
5q
Illustration 3
Find row minimums and column
maximums with considering the
values of unknowns
School of Business Management prof Narinder Verma32
Specify the range for the value of the
game in the following case assuming
that the payoff is for player A.
3 6 1
5 2 3
4 2 -5
A1
A2
A3
B1 B2 B3
Col max 5 6 3
Row min
1
2
-5
Illustration 4
School of Business Management prof Narinder Verma33
maximin minimax (2 3)
Hence the game has no saddle point.
When there is no saddle point, then
Maximin Value V Minimax Value i.e., 2 V 3.
Thus the value of the game lies
between 2 and 3.
Illustration 4
School of Business Management prof Narinder Verma34
Principles of DominanceThese are used to reduce the size of the payoff matrix by deleting certain inferior rows and or columns
Dominance rules are especially used for the evaluation of the two-person zero-sum games without a saddle point
1. For player B who is assumed to be a loser, if each element in a column Cr is greater than or epual to the corresponding element in another column Cs then the column Cr is said to be dominated by Cs (or inferior to) and there column Cr can be deleted
School of Business Management prof Narinder Verma35
Principles of Dominance
2. For player A who is assumed to be a gainer, if each element in a row Rr is less than or equal to the corresponding element in another row Rs then the row Rr is said to be dominated by Rs (or inferior to) and there row Rr can be deleted
3. A strategy k can also be dominated if it is inferior to an average of two or more other pure strategies
Dominance rules are framed assuming that the payoff matrix is a profit matrix for player A
School of Business Management prof Narinder Verma36
8 6 2 8
8 9 4 5
7 5 3 5
B1 B2 B3 B4
A1
A2
A3
Strategy A3 is dominated by the
strategy A2 and so can be eliminated.
Eliminating the strategy A3 , we get:
Principles of Dominance:Ex1
School of Business Management prof Narinder Verma37
8 6 2 8
8 9 4 5
B1 B2 B3 B4A1
A2
Eliminating the strategies B1 , B2, and
B4 we get the reduced payoff matrix:
For player B, strategies B1, B2, and B4 are
dominated by the strategy B3.
Principles of Dominance:Ex1
School of Business Management prof Narinder Verma38
2
4
B3
A1
A2
Now , for player A, strategy A1 is dominated
by the strategy A2
Eliminating the strategy A1
Principles of Dominance:Ex1
School of Business Management prof Narinder Verma39
4
B3
A2
We thus see that A should always play A2 and B always B3 and the value of the game
is 4 as before.
Principles of Dominance:Ex1
School of Business Management prof Narinder Verma40
-5 10 20 8
5 -10 -10 6
5 -20 -20 7
B1 B2 B3 B4
A1
A2
A3
Strategy B3 is dominated by the
Strategies B1 and B2 and so can be
eliminated
B4 is inferior to B1 so is deleted., we get:
Principles of Dominance:Ex2
School of Business Management prof Narinder Verma41
-5 10
5 -10
5 -20
B1 B2A1
A2
Eliminating the strategy A3 we get the
reduced payoff matrix:
For player A, strategy A3 is dominated by
the strategy A2.
Principles of Dominance:Ex2
A3
School of Business Management prof Narinder Verma42
B1 B2A1
A2
The game has no saddle point as
was the case originally.
Principles of Dominance:Ex2
-5 10
5 -10
School of Business Management prof Narinder Verma43
Used for (2 a n) or (m a 2) games, i.e.,
for two-person zero-sum games where
at least one player has only 2 strategies
It is assumed that the player with two
strategies, chooses a mixture of both
the strategies with some fixed but
unknown probabilities, to be calculated
Mixed Strategies:
Graphical Method
School of Business Management prof Narinder Verma44
a11 a12 . a1n
a21 a22 . a2n
B1 B2 . Bn
A1
A2
player A selects two strategies A1 and A2
with probabilities p1 and p2 respectively
Graphical Method: 2 x n Game
Probability
p1
p2
player Bplayer A
Probability p1 p2 .. pn
School of Business Management prof Narinder Verma45
Such that 1,2 0 and 1 + 2 = 1
Graphical Method: 2 x n Game
Expected Payoff for Player A
Bs pure Strategies As Expected Payoff
B1 + B2 + B3 + . .
. .
. .
Bn +
School of Business Management prof Narinder Verma46
Now plot probabilities on a- axis and
Expected payoff on Y- axis
Choose the lower envelope if initial
payoff was for A, and
player A has two strategies (as here)
Graphical Method: 2 x n Game
School of Business Management prof Narinder Verma47
Graphical: Illustration 1
2 4 3 8 4
5 6 3 7 8
10 5 9 8 7
4 2 8 4 3
B1 B2 B3 B4 B5
A1
A2
A3
A4
A1 and A4 are dominated by A3
School of Business Management prof Narinder Verma48
Graphical: Illustration 1
5 6 3 7 8
10 5 9 8 7
B1 B2 B3 B4 B5
A2
A3
B1 is dominated by B3,so B1 gets deleted
School of Business Management prof Narinder Verma49
Graphical: Illustration 1
6 3 7 8
5 9 8 7
B2 B3 B4 B5
A2
A3
Probability
p1
p2
Assume that player A selects Strategies
A2 and A3 with probabilities p1 and p2respectively
School of Business Management prof Narinder Verma50
Such that 1,2 0 and 1 + 2 = 1
Graphical : Illustration 1
Expected Payoff for Player A
Bs pure Strategies As Expected Payoff
B2 + B3 + B4 + B5 +
School of Business Management prof Narinder Verma51
M
As Expected payoff, A1
probability
B2
9
8
7
6
5
4
3
2
1
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1(0, 0)
Graphical : Illustration 1
9
8
7
6
5
4
3
2
1
p1=1, p2=0 p1=0, p2=1
(0, 0)
As Exapected payoff, A2B3B4
B5
L
N
School of Business Management prof Narinder Verma52
Out of three points L, M and N; M
represents Maximin, so M is the
optimal mixed strategy
Solve payoff equations for B2 and B3
+ = + , and
+ =
Graphical : Illustration 1
School of Business Management prof Narinder Verma53
+ ( ) = + ( )
= + = +
= + = +
= =
There for =
and =
and =
Graphical : Illustration 1
School of Business Management prof Narinder Verma54
Illustration 1: Calculation for B
6 3
5 9
B2 B3
A2
A3
probability q2 q3
School of Business Management prof Narinder Verma55
Such that 2, 3 0 and 2 + 3 = 1
Graphical : Illustration 1
Expected Payoff for Player B
As pure Strategies Bs Expected Payoff
A2 + A3 +
School of Business Management prof Narinder Verma56
+ ( ) = + ( )
= + = +
= + = +
= =
There for =
and =
and =
Graphical : Illustration 1
School of Business Management prof Narinder Verma57
Simplex Method
a11 a12 a1n
a21 a22 ........ a2n
.
. aij
am1 am2 . amn
A1
A2
.
.
Am
B1 B2 Bn
Strategies of player B
Str
ate
gie
s o
f p
layer
AProbability
1
2
Probability q1 q2 qn
School of Business Management prof Narinder Verma58
= =
= , , , . . ,
Simplex Method
,
:
School of Business Management prof Narinder Verma59
Simplex Method
a11 p1+ a21 p2+ ... + am1 pm V
a12 p1 + a22 p2 + ... + am2 pm V
:
a1n p1 + a2n p2 + ... + amn pm V
p1 + p2 + ... + pm = 1
pi 0 (Non-negativity constraints) = 1,2, 3, . . ,
, ,
School of Business Management prof Narinder Verma60
Simplex MethodDivide all constraints by V
a111
+ a21
2
+ ... + am1
1
a121
+ a22
2
+ ... + am2
1
:
a1n1
+ a2n
2
+ ... + amn
1
1
+ 2
+ ... +
= 1
0 = 1,2, 3, . . ,
School of Business Management prof Narinder Verma61
Simplex Method
Let
= so that
a11 1+ a21 2+ ... + am1 1
a12 1+ a22 2+ ... + am2 1
:
a1n 1+ a2n 2+ ... + amn 1
1+2+ ... += 1
0 = 1,2, 3, . . ,
School of Business Management prof Narinder Verma62
Simplex Method
=
, there for,
=
= 1+2+ ... + ,
a11 1+ a21 2+ ... + am1 1
a12 1+ a22 2+ ... + am2 1
:
a1n 1+ a2n 2+ ... + amn 1
0 = 1,2, 3, . . ,
School of Business Management prof Narinder Verma63
Simplex Method
a11 a12 a1n
a21 a22 ........ a2n
.
. aij
am1 am2 . amn
A1
A2
.
.
Am
B1 B2 Bn
Strategies of player B
Str
ate
gie
s o
f p
layer
AProbability
1
2
Probability q1 q2 qn
School of Business Management prof Narinder Verma64
= =
= , , , . . ,
Simplex Method
,
:
School of Business Management prof Narinder Verma65
Simplex Method
a11 q1+ a12 q2+ ... + a1n qn V
a21 q1 + a22 q2 + ... + a2n qn V
:
am1 q1 + am1 q2 + ... + amn qn V
q1 + q2 + ... + qn = 1
qj 0 (Non-negativity constraints) = 1,2, 3, . . ,
, ,
School of Business Management prof Narinder Verma66
Simplex MethodDivide all constraints by V
a111
+ a12
2
+ ... + a1n
1
a211
+ a22
2
+ ... + a2n
1
:
am11
+ am2
2
+ ... + amn
1
1
+ 2
+ ... +
= 1
0 = 1,2, 3, . . ,
School of Business Management prof Narinder Verma67
Simplex Method
Let
= so that
a11 1+ a12 2+ ... + a1n 1
a21 1+ a22 2+ ... + a1n 1
:
am1 1+ am2 2+ ... + amn 1
1+2+ ... += 1
0 = 1,2, 3, . . ,
School of Business Management prof Narinder Verma68
Simplex Method
=
, there for,
=
= 1+2+ ... + ,
a11 1+ a12 2+ ... + a1n 1
a21 1+ a22 2+ ... + a2n 1
:
am1 1+ am2 2+ ... + amn 1
0 = 1,2, 3, . . ,
School of Business Management prof Narinder Verma69
Solve for ys as there are only slack variables
X values will be calculated from dual
values of slack in the primal for y
For x values, take the absolute value of
(cj-zj), i.e., | cj -zj| against slack variables
For x1 take | cj -zj| against s1, for x2 take
| cj -zj| against s2 and like wise
Simplex Method
School of Business Management prof Narinder Verma70
Note: As
= 0 and
= 0
For this, value must be non-negative, i. e., 0 ,there for all aij 0
For this, add {| largest negative aij | +1}to all aij . If value of the game now is .
Original value of the game ( ) now is
= { | | +1}
Simplex Method
School of Business Management prof Narinder Verma71
1 -1 3
3 5 -3
6 2 -2
B1 B2 B3
A1
A2
A3
MaxCol 6 5 3 5
Row min
-1
-3
-2
minimax
maximin
Simplex Method: Illustration
School of Business Management prof Narinder Verma72
Clearly there is no saddle point
Value of the game follows
Here V can be negative, so add
{| largest negative aij | +1} i.e., {| -3| +1}= 4 to all aij
Simplex Method: Illustration
School of Business Management prof Narinder Verma73
Modified Payoff Matrix
Probability 1 2 3
5 4 7
7 9 1
10 6 2
B1 B2 B3
A1
A2
A3
Probability
1
2
3
Simplex Method: Illustration
School of Business Management prof Narinder Verma74
F ; ,, and + + =
Simplex Method : Illustration
Expected Payoff for Player A
Bs pure Strategies As Expected Payoff
B1 + + B2 + + B3 + +
School of Business Management prof Narinder Verma75
Simplex Method
5 p1+7 p2+ 10 p3 V
3 p1+9 p2+ 6 p3 V
7 p1 + p2 + 2 p3 V
p1 + p2 + p3 = 1
p1, p2 , p3 0
, ,
School of Business Management prof Narinder Verma76
Simplex MethodDivide all constraints by V
5
+ 7
+ 10 1
3
+ 9
+ 6 1
7
+
+ 2 1
+
+
=
,
, 0
School of Business Management prof Narinder Verma77
Simplex Method
Let
= ,
= and
= so that
5 + 7 + 10 1
3 + 9 + 6 1
7 + + 2 1
++ =
,, 0
School of Business Management prof Narinder Verma78
Simplex Method
=
, there for,
=
= ++ ,
5 + 7 + 10 1
3 + 9 + 6 1
7 + + 2 1
,, 0
School of Business Management prof Narinder Verma79
F ; ,, and + + =
Simplex Method : Illustration
Expected Payoff for Player B
As pure Strategies Bs Expected Payoff
A1 + + A2 + + A3 + +
School of Business Management prof Narinder Verma80
Simplex Method
5 q1 + 3 q2 + 7 q3 V
7 q1 + 9 p2 + q3 V
10 q1 + 6 q2 + 2 q3 V
q1 + q2 + q3 = 1
q1, q2 , q3 0
, ,
School of Business Management prof Narinder Verma81
Simplex MethodDivide all constraints by V
5
+ 3
+ 7 1
7
+ 9
+ 1
10
+ 6
+ 2 1
+
+
=
,
, 0
School of Business Management prof Narinder Verma82
Simplex Method
Let
= ,
= and
= so that
5 + 3 + 7 1
7 + 9 + 1
1 + 6 + 2 1
++ =
,, 0
School of Business Management prof Narinder Verma83
Simplex Method
=
, there for,
=
= ++ ,
5 + 3 + 7 1
7 + 9 + 1
1 + 6 + 2 1
,, 0
School of Business Management prof Narinder Verma84
Simplex Method: Illustration
The Standard form of LPP with constraints is:
= 1+ 2 + + 01 + 02 + 03
Subject to
5 1+ 3 2+ 7 3+ 1 1+ 0 2+ 0 3 = 1
7 1+ 9 2+ 3+ 0 1+ 1 2+ 0 3 = 1
10 1+ 6 2+ 2 3+ 0 1+ 0 2+ 1 3 = 1
1 0, 2 0, 3 0, 1 0, 2 0, 3 0
School of Business Management prof Narinder Verma85
Final Tableau with optimal solution:
Cj 1 1 1 0 0 0
CB
Basic
Variable
(B)
Basic
Soln(YB)y1 y2 y3 s1 s2 s3
1 y3 1/10 2/5 0 1 3/20 -1/10 0
1 y2 1/10 11/15 1 0 -1/60 7/60 0
0 s3 1/5 24/5 0 0 -1/5 -3/5 1
17/15 0 0 2/15 1/15 0
-2/15 0 0 -2/15 -1/15 0
Zj
(Net Evaluation)Cj - Zj
Simplex Method: Illustration
Zq= 1/5
School of Business Management prof Narinder Verma86
Optimal strategy for B:
= , =
=
=
There for =
=
= there for = = =
= there for = =
=
= there for = =
=
Simplex Method: Illustration
School of Business Management prof Narinder Verma87
As =
=
Original value of the game ( ) now is
= { | | +1}
= { } =
Simplex Method: Illustration
School of Business Management prof Narinder Verma88
X values will be calculated from dual
values of slack in the primal for y
For x values, take the absolute value of
(cj-zj), i.e., | cj -zj| against slack variables
For x1 take | cj -zj| against s1, for x2 take
| cj -zj| against s2 and like wise
=
, =
=
Simplex Method
School of Business Management prof Narinder Verma89
Optimal strategy for A:
=
, =
=
= there for = =
=
= there for = =
=
= there for = = =
= (
,
, ), = ,
,
, = 1
Simplex Method: Illustration
School of Business Management prof Narinder Verma90
End of Unit - I