1 DESCISION THEORY

SAS 402: BAYESIAN INFERENCE

AND DESCISION THEORY1
1 Descision Theory
Descision making situations may be classified into three types, namely:- De-
terministic situations, probabilistic situations and uncertainty situations.
In deterministic situations, the problem data representing the situations
are constant. They do not vary with respect to time or any other basis.
Under probabilistic situations, the problem data representing the situa-
tion do not remain constant. They vary due to chance, and it is possible to
represent the data in the form of some probability distribution. Later, these
probability distributions become components of the descision making model
of the situation.
When the problem data are subjected to variation, and it is not possible
to represent them in the form of any probability distribution, this situation
is called an uncertain situation. Game theory is an example of an uncertain
situation.

1.1 Game Theory

1.1.1 Terminologies
1. Players: There are two players in a game. The players may be any two
companies (Company A and Company B), competing for tenders, two
countries planning for trade gains in a third country or two persons
bidding in a game,. . . etc

2. Strategy: It is a course of action taken by a player, for example, giving

computer furniture free of cost, giving 20% additional hardware, giving
special prices, etc, while selling computer hardware.
The strategies can be further classified into pure stategy and mixed
strategy. Let m be the number of strategies of player A, and let n be the
number of strategies of player B, and pi be the probability of selection
of the alternative i of player A, (i = 1, 2, 3, . . . , m), while qj be the prob-
ability of selection of alternative j of player B, for (j = 1, 2, 3, . . . , n.)
The sum of the probabilities of selectio of various alternatives of each
1
Lecturer: Dr. Edgar Ouko Otumba, School of Mathematics, Statistics and Actuarial
Sciences, Maseno University, Kenya. Email: eotumba@maseno.ac.ke

1
1.1 Game Theory 1 DESCISION THEORY

of the players is equal to 1, that is;

m
X
pi = 1
i=1

and n
X
qj = 1
j=1

1.1.2 Pure strategy

If a player selects a particular strategy with a probability of 1, then that
strategy is known as a pure strategy. This means that the player is selecting
that particular strategy alone, ignoring his remaining strategies.
If a player A follows a pure strategy, then only one of the pi values will
be equal to 1, and the remaining values will be equal to 0. A sample set of
probabilities of selection of the alternatives for player A is equal to 1, that
is;
p1 + p2 + . . . = 1

1.1.3 Mixed strategy

If a player follows more than one strategy, then the player is said to follow a
mixed strategy. But the probability of selection of the individual strategies
will be less than one, and their sum will be equal to one. If player B follows
a mixed strategy, then a sample set of probabilities of selection of mixed
strategy is a s follows;

q1 = 0.65; q2 = 0; q3 = 0.35

so that the sum of the probabilities is equal to one, that is,

q1 + q2 + q3 = 0.65 + 0 + 0.35 = 1.00

1.1.4 Payoff matrix

Each combination of the slternatives of players A and B is associated with
an outcome aij . If aij is positive, then it represents a gain to player A, and
a loss to player B. If it is negative, it represents a loss to A and a gain to B.

2
1.1 Game Theory 1 DESCISION THEORY

1.1.5 Maximin principle

This principle maximises the minimum guaranteed gains of player A. The
minimum gains with respect to different alternatives of A, irrespective of
B’s alternatives are obtained first. The maximum of these minimum gains
is known as the maximin value and the corresponding alternative is called
maximin strategy.

1.1.6 Minimax principle

This principle minimises the maximum losses. The maximum losses with
respect to different alternatives of player B, irrespective of player A’s alter-
natives, are obtained first. The minimum of these maximum losses is known
as the minimax value, and the corresponding alternative is called minimax
strategy.

1.1.7 Saddle Point

In a game, if the maximin value is equal to the minimax value, then the
game is said to have a saddle point. The intersecting cell corresponding to
this value is known as the saddle point. If the game has a saddle point, then
each player has a pure strategy.

1.1.8 Value of the game

If the game has a saddle point, then the value of the cells at the saddle point
is called the value of the game. Otherwise, the value of the game is computed
based on expected value calculations which will be explained later.

1.1.9 Two-person zero sum game

In a game with two players, if the gain of one player is equal to the loss of
another player, then that game is called a two-person zero sum game.

1.1.10 Game with pure strategies

As we know, if a game has a saddle point, then the game is said to have a
pure strategy for each of the players.

3
1.1 Game Theory 1 DESCISION THEORY

Assignment 1.1 Two computer hardware manufacturing companies, Com-

pay A and Company B, competing for supplying computers to a Government
department. Each company has listed its strategy for selling computers;
The strategies for Company A are:

• Giving a special price.

• Giving 20% worth of additional hardware.

• Supplying computer furniture free of cost.

The strategies for Company B are:

• Giving a special price.

• Giving 30% worth of additional hardware.

• Giving free training to the users of the organisation which is buying the
hardware.

The payoff matrix with respect to Company A is as given here (Rows refer to
A and Columns refer to B):
 
20 15 22
35 45 40
18 20 25
Determine the optimal strategies for Companies A and B