Game Theory

1
 Game Theory
Players: There are two players (companies) in the game. Ex. Two companies
competing for tenders

Strategy: Course of action taken by a player. Each player has number of options
called strategies. There are two classification of strategy selection

a. Pure Strategy: A player selects a particular strategy with probability of 1.

Probabilities of all other available strategies is zero so that sum of all
probabilities is 1.
b. Mixed Strategy: A player selects more than one strategy , wherein each
strategy is associated with a probability, so that sum of all probabilities is 1.

Payoff matrix: Outcome (aij)of a game when different alternatives are adopted
by the competing players. Positive aij represents gain to player A and loss to
player B. Negative aij represents loss to A and gain to B.

Two person zero sum game: Game of two players wherein the gain of one
player is equal to the loss of other player
2
 Game Theory
Maximin Principle: The player A who is maximising his gains finds his
minimum guaranteed gain from each strategy and selects the maximum
value out of these minimum gains. Player B

•When A plays strategy A1, he may B1 B2 B3 B4

gain 12, 4, 14 or 16 depending upon the A1 12 4 14 6 4
strategy that B plays. A2 10 6 12 16 6
Player A
A3 8 2 -6 10 -6
•A is guaranteed a minimum gain of 4. 12 6 14 16

•Minimum gains across rows are 4, 6, -6

•Of these the maximum is 6.

•i.e. A follows Maximin principle

3
 Game Theory
Minimax Principle: The player B who is minimising his loss finds his
maximum loss from each strategy and selects the minimum value out of
these maximum losses. Player B
B1 B2 B3 B4
•When B plays strategy B1, he may
A1 12 4 14 6 4
incur loss of 12, 10 or 8 depending
upon the strategy that A plays. Player A A2 10 6 12 16 6
A3 8 2 -6 10 -6
•B has a maximum loss of 12. 12 6 14 16

•Maximum loss across columns are 12,

6, 14, 16.

•Of these the minimum is 6.

•i.e. B follows Minimax principle

4
 Game Theory
Saddle point: In a game if Maximin value is equal to Minimax value, then the
game is said to have a saddle point.
•Saddle point is the intersecting cell
Player B
corresponding to these values.
B1 B2 B3 B4
•If the game has a saddle point, then A1 12 4 14 6 4
each player has a pure strategy. A2 10 6 12 16 6
Player A
A3 8 2 -6 10 -6
•A adopts strategy A2 with 12 6 14 16
probability 1. Strategies A1 and A3
have zero probabilities.
•B adopts strategy B2 with probability 1. Strategies B1, B3 and B4 have zero
probabilities.
•Value of the game: Value of the cell at the saddle point. (if the game has
saddle point). Value of the game is 6.
•Else, the value of the game is calculated based on expected value calculations
5
 Game Theory
Game with mixed strategies: A game without a saddle point is said to
have mixed strategies.
Consider a 2 * 2 payoff matrix w.r.t. player A, without a saddle point.
Let pi and qj be the probabilities of B
selection of the alternative of player A B1 B2 Oddments
and B respectively. A1 a b c-d
A
c-d a-b A2 c d a-b
p1 = p2 =
Oddments b- d a-c
a-b + c-d a-b + c-d

b-d a-c a c-d + c a-b

q1 = q2 = V=
a-c + b-d a-c + b- d
a-b + c-d

6
 Game Theory
Problem 1: Consider the payoff matrix w.r.t. player A and solve it
optimally B
p1 4 4 1 2
= =
3+4 7 1
A
6 9 4
2
p2 3 3 8 4 3
= =
3+4 7
5 2
(6 * 4) + (8 * 3) 48
q1 5 5 V= =
= = 4 + 3 7
2+5 7
Strategies :
q2 2 2
= = Player A: [4/7, 3/7]
2+5 7
Player B: [5/7, 2/7]
7
 Game Theory
Dominance Property:

Size of the matrix is reduced by eliminating redundant rows or columns

Dominance property for rows [In a payoff matrix of player A]

1. If all entries of row X ≥ corresponding entries in row Y, then
Row X dominates row Y. Delete row Y.

2. If sum of entries of any two rows (X & Y) ≥ corresponding entries of row Z,

Row X & Y dominates row Z. Delete row Z.

Dominance property for columns[In a payoff matrix of player A]

1. If all entries of column X ≤ corresponding entries in column Y, then
column X dominates column Y. Delete column Y.

2. If sum of entries of any two column (X & Y) ≤ corresponding entries of col. Z,

column X & Y dominates column Z. Delete column Z.
8
 Game Theory
Problem 2: Players A & B play a game in which each player has 3 coins (20p,
25p & 50p). Each of them selects a coin without the knowledge of the other
person. If the sum of the values of the coins is an even number, A wins B’s
coins. If that sum is an odd number, B wins A’s coin.
a. Develop a playoff matrix w.r.t. player A
b. Find the optimal strategies for the players
Solution:
B B
I II III I II III
20p 25p 50p 20p 25p 50p
I 20p 20 -20 50 -20 I 20p 20 -20 50
A II 25p -25 25 -25 -25 A II 25p -25 25 -25
III 50p 20 -50 50 -50
20 25 50 Col. I dominates col.III. Delete col. III.
Row I dominates row III. Delete row III. 9
 Game Theory
B
I II
20p 25p
I 20p 20 -20 50
A II 25p -25 25 40
Oddments 45 45
p1 50 5 (20 * 50) + (-25 * 40)
= = V= = 0
50+40 9 50 + 40

p2 40 4 Strategies :
= =
50+40 9
Player A: [5/9, 4/9, 0]
q1 45 1
= =
45+45 2
Player B: [1/2, 1/2, 0]
q2 45 1
= = V=0
45+45 2 10
 Game Theory
Problem 3: Using the rule of dominance, obtain the optimal strategies
for both the players and determine the value of the game. The payoff
matrix for player A is given
B
B
I II III IV
I II III IV A II 3 4 2 4
I 3 2 4 0 0 III 4 2 4 0
A II 3 4 2 4 2 IV 0 4 0 8

III 4 2 4 0 0 Col. III dominates col. I. Delete col. I.

IV 0 4 0 8 0 B
4 4 4 8 II III IV
A II 4 2 4
Row III dominates row I. Delete col. I. III 2 4 0
IV 4 0 8
No single row or column dominates
11
 Game Theory
B B
II III IV III IV odds
A II 4 2 4
III 2 4 0 A III 4 0 8
IV 4 0 8
IV 0 8 4
Average of Col. III & IV dominates
Col. II. Delete Col. II.
odds 8 4
B
III IV pIII = 8/12
A II 2 4
pIV = 4/12
III 4 0
IV 0 8 qIII = 8/12
V = 8/3
Average of row III & IV dominates
qIV = 4/12
row II. Delete row II. 12
Thank You

13