Game Theory

Decision under uncertainty

 Decision making under uncertainty involves
alternative actions whose payoffs depend on
the (random) states of nature.
 Specifically, the payoff matrix of a decision
problem with m alternative actions and n
states of nature can be represented as given in
the following Table.
Payoff matrix
s1 s2 sn
a1 v(a1,s1) v(a1,s2) … v(a1,sn)
a2 v(a2,s1) v(a2,s2) … v(a2,sn)

am v(am,s1) v(am,s2) … v(am,sn)

 The element ai represents action i and the element sj

represents state of nature j.
 The payoff or outcome associated with action ai and
state sj is v(ai,sj).
Decision under uncertainty
 The difference between making a decision
under risk and under uncertainty is that in the
case of uncertainty, the probability distribution
associated with the states sj, j=1,2,…,n, is
either unknown or cannot be determined.
 This lack of information has led to the
development of the following criteria for
analyzing the decision problem:
 1. Laplace 2.Minimax 3.Savage 4. Hurwicz
Laplace Method
 The Laplace criterion is based on the optimistic
assumption that all states are equally likely to
occur,
i.e.; P( s1) = P(s2) = …=P(sn) = 1/n.
If the payoff v(ai,si) is gain, then we select the
action,  n
1


max   v(ai , s j ) 
a
i n j 1
 
 

If v(ai,si) is loss, then minimization replaces

maximization.
Minimax (Maximin)
 The minimax (maximin) criterion is based on the
conservative attitude of making the best of the
worst possible conditions. If v(ai,si) is loss, then
we select the action,


min max v(ai , s j )
ai sj 
 If v(ai,si) is gain, we use the maximin criterion

 
given by
max min v(ai , s j )
ai sj
Example – 1
 National Outdoors School (NOS) is preparing a summer
campsite in the heart of Alaska to train individuals in
wilderness survival. NOS estimates that attendance can
fall into one of four categories: 200, 250, 300, and 350
persons. The cost of the campsite will be the smallest
when its size meets the demand exactly. Deviations
above or below the ideal demand levels incur additional
costs resulting from building surplus (unused) capacity
or losing income opportunities when the demand is not
met. Letting a1 to a4 represent the sizes the campsites
(200,250,300, and 350 persons) and s1 to s4, the level
of attendance, the following table summarizes the cost
matrix (in thousands of dollars) for the situation.
Example – 1(Continued)
Example – 1(Continued) : Minimax Criteria

s1 s2 s3 s4 Rowmax
a1 5 10 18 25 25
a2 8 7 12 23 23
a3 21 18 12 21 21 (Minimax)
a4 30 22 19 15 30
Game Theory
 Game theory deals with decision situations in which two
intelligent opponents with conflicting objectives are
trying to outdo one another.
 Examples: Launching advertising campaigns for
competing products and planning strategies for warring
armies.
 In a game conflict, two opponents, known as players,
will each have a (finite or infinite) number of alternatives
or strategies. Associated with each pair of strategies is
a payoff that one player receives from the other.
 Such games are known as two-person zero-sum games
because a gain by one player signifies an equal loss to
the other.
Game Theory
 It suffices, then, to summarize the game
in terms of the payoff to one player.
Designating the two players as A and B
with m and n strategies, respectively, the
game is usually represented by the payoff
matrix to player A as
Game Theory
B1 B2 Bn
A1 a11 a12 … a1m
A2 a21 a22 … a2m

…… … … …
Am am1 am2 … amn

 The representation indicates that if A uses strategy i

and B uses strategy j, the payoff to A is aij, which
means that the payoff to B is –aij.
Optimal solution of Two-person Zero-sum
Games (Example – 2)
 Two companies, A and B, sell two brands of flu
medicine. Company A advertises in radio (A1),
television (A2), and newspapers (A3). Company
B, in addition to using radio(B1), television (B2),
and newspapers(B3), also mails brochures(B4).
Depending on the effectiveness of each
advertising campaign, one company can
capture a portion of the market from the other.
The following matrix summarizes the
percentage of the market captured or lost by
company A.
Example – 2 (Continued…)

B1 B2 B3 B4 Row.Min
A1 8 -2 9 -3 -3
A2 6 5 6 8 5 (Maximin)
A3 -2 4 -9 5 -9
Col. Max 8 5 9 8
Minimax
Example – 2 (Continued…)
 The optimal solution of the game calls for
selecting strategies A2 and B2 , which means
that both companies should use television
advertising. The pay of will be in favour of
company A, because its market share will
increase by 5 %. In this case, we say that the
value of the game is 5%, and that A and B are
using a saddle–point solution.
Solution of Mixed Strategy Games
 Games with mixed strategies can be
solved either graphically or by linear
programming.
 The graphical solution is suitable for
games in which at least one player has
exactly two pure strategies. The method
is interesting because it explains the idea
of a saddle point graphically.
 Linear progamming can be used to solve
any two-person zero-sum game.
Graphical solution of Games
 We start with the case of (2xn) games in which
player A has two strategies.

y1 y2 … yn
B1 B2 … Bn
x1: A1 a11 a12 … a1m
1- x1: A2 a21 a22 … a2m
Graphical solution of Games
 The game assumes that player A mixes
strategies A1 and A2 with the respective
probabilities x1 and 1-x1, 0<x1<1. Player B
mixes strategies B1 to Bn with the probabilities
y1,y2, …, and yn, where yj > 0 for j=1,2,…,n and
y1+y2+…+yn=1.
 In this case, A’s expected payoff corresponding
to B’s jth pure strategy is computed as
(a1j-a2j)x1+ a2j, j=1,2,…,n
 Player A thus seeks to determine the value of
x1 that maximizes the minimum expected
payoffs – i.e. max
1

x

min (a1j-a2j)x1+ a2j
j
Example – 3
 Consider the following 2 x 4 game. The payoff
is for Player A.

B1 B2 B3 B4
A1 2 2 3 -1
A2 4 3 2 6
Example – 3 (Continued…)
 This game has no pure strategy solution. A’s expected
payoffs corresponding to B’s pure strategies are given
as

B's pure strategy A's expected payoff

1 -2x1+4
2 -x1+3
3 x1+2
4 -7x1+6
Example – 3 (Continued…)
 Graphical Solution
y 9 9
8 8
7 7
B4 6 6
5 5
B1 4 4
B2 3 3
B3 2 2
1 1
0 0.25 0
x
-1 -1