Game Theory

Lecture Notes By
Y. Narahari
Department of Computer Science and Automation
Indian Institute of Science
Bangalore, India
July 2012

Chapter 1: Introduction to Game Theory

Note: This is a only a draft version, so there could be flaws. If you find any errors, please do send email to
[email protected]. A more thorough version would be available soon in this space.
In this chapter, we will introduce a classical and well studied form of games called strategic form games and
expose the readers to key notions in game theory such as utilities, rationality, intelligence, and common
knowledge. We also present a brief discussion on different types of games.

Game theory may be defined as the study of mathematical models of interaction between rational,
intelligent decision makers [1]. The decision makers are usually referred to as players or agents. The
interaction may include both conflict and cooperation. Game theory provides general mathematical
techniques for analyzing situations in which two or more players make decisions that influence one
anothers welfare. There are many categories of games that have been proposed and discussed in game
theory. We first introduce a class of games called strategic form games or normal form games, which
are perhaps the most commonly studied among all types of games.
Definition 1 (Strategic Form Game): A strategic form game is a tuple hN, (Si )iN , (ui )iN i,
where N = {1, 2, . . . , n} is a finite set of players; S1 , S2 , . . . , Sn are the strategy sets of the players
1, . . . , n, respectively; and ui : S1 S2 Sn R for i = 1, 2, . . . , n are mappings called the utility
functions or payoff functions.
The strategies are also called actions or more specifically pure strategies. We denote by S, the
Cartesian product S1 S2 Sn . The set S is the collection of all strategy profiles of the players.
It is important to note that the utility of an agent depends not only on its own strategy but also on
the strategies of the rest of the agents. Every profile of strategies produces or induces an outcome in
the game. A strategic form game is said to be finite if N and all the strategy sets S1 , . . . , Sn are finite.
A strategic form game captures each agents decision problem of choosing a strategy that will
counter the strategies adopted by the other agents. Each player is faced with this problem and therefore
the players can be thought of as simultaneously choosing their strategies from the respective sets
S1 , S2 , . . . , Sn . We can view the play of a strategic form game as follows: each player simultaneously
writes down a chosen strategy on a piece of paper and hands it over to a referee who then computes the
1

outcome and the utilities. we will be presenting several examples of strategic form games in Chapter
3.

Key Notions in Game Theory

There are certain key notions which are fundamental to game theory. We discuss these notions and a
few related issues.

1.1

Utilities

Utilities (also called payoffs) enable the preferences of the players to be expressed in terms of real
numbers in some utility scale. Utility theory is the science of assigning numbers to outcomes in a
way that captures the preferences of the players. The theory is a foundational contribution of von
Neumann and Morgenstern, who stated and proved in [2] a crucial result called the expected utility
maximization theorem. This theorem establishes for any rational decision maker that there must exist
a way of assigning utility numbers to different outcomes in a way that the decision maker would
always choose the option that maximizes his expected utility. This theorem holds under quite weak
assumptions about how a rational decision maker should behave.

1.2

Intelligence

Another key notion in game theory is that of intelligence of the players. This notion connotes that each
player in the game knows everything about the game that a game theorist knows, and the player is
competent enough make any inferences about the game that a game theorist can make. In particular,
an intelligent player is strategic, that is, would fully take into account his knowledge or expectation
of behavior of other agents in determining what his best response strategy should be. Each player is
assumed to have enough resources to carry out the required computations involved in determining a
best response strategy.
Myerson [1] and several other authors provide the following convincing explanation to show that the
two assumptions of rationality and intelligence are indeed logical and reasonable. The assumption that
all individuals are rational and intelligent may not exactly be satisfied in a typical real-world situation.
However, any theory that is not consistent with the assumptions of rationality and intelligence loses
credibility on the following count: If a theory predicts that some individuals will be systematically
deceived into making mistakes, then such a theory will lose validity when individuals learn through
mistakes to understand the situations better. On the other hand, a theory based on rationality and
intelligence assumptions would be sustainable.

1.3

Common Knowledge

The notion of common knowledge is an important implication of intelligence. Aumann [3] defines
common knowledge as follows: A fact is common knowledge among the players if every player knows
it, every player knows that every player knows it, and so on. That is, every statement of the form
every player knows that every player knows that every player knows it is true ad infinitum. If
it happens that a fact is known to all the players, without the requirement of all players knowing
that all players know it, etc., then such a fact is called mutual knowledge. In game theory, analysis
often requires the assumption of common knowledge to be true; however, sometimes, the assumption

of mutual knowledge suffices for the analysis. A players private information is any information that
the player has that is not common knowledge among all the players.
The intelligence assumption means that whatever a game theorist knows about the game must be
known or understood by the players of the game. Thus the model of the game is also known to the
players. Since all the players know the model and they are intelligent, they also know that they all
know the model. Thus the model is common knowledge.
In a strategic form game with complete information, the set N , the strategy sets S1 , . . . , Sn , and the
utility functions u1 , . . . , un are common knowledge, that is every player knows them, every player knows
that every player knows them, and so on. We will be studying strategic form games with complete
information in this and the next few chapters. We will study games with incomplete information in
Chapter 10.
1.3.1

An Example for Illustrating Common Knowledge

This example is a variant of the one presented by Myerson [1]. Assume that there are five rational
and intelligent mothers A, B, C, D, and E and let a, b, c, d, and e be their daughters (or sons),
respectively. The kids go the school every day, escorted by their respective mothers and the mothers get
an opportunity everyday to indulge in some conservation. The conversation invariably centers around
the performance and behavior of the kids. Everyday when the five mothers meet, the conversation
protocol is the following. If a mother thinks her kid is well behaved, she will praise the virtues of her
kid. On the other hand, if a mother knows that her kid is not well behaved, she will cry. All mothers
follow this protocol.
The fact is that none of the kids is well behaved but their behaviors is unknown to their respective
mothers. However, whenever a mother finds that the kid of another mother is not well behaved, she
would immediately report it to all mothers except the kids mother. For example, if A finds b badly
behaved, then A would report it to C,D, and E, but not to B. This protocol is also known to all the
mothers. Since none of the kids is well behaved, the fact that a kid is not well behaved is common
knowledge among all the mothers except the kids own mother.
Since each mother does not know that her kid is badly behaved, it turns out that every mother
keeps praising her kid everyday. On a fine day, the class teacher meets all the mothers and makes the
following statement: one of the boys is not well behaved. Thus the fact that one of the boys is not a
well behaved is now common knowledge among all the mothers. Subsequently, when the five mothers
meet the next day, all of them praise their respective kids; the same happens on the 2nd day, 3rd , and
the 4th day. On the 5th day, however, all the mothers cry together because all of them realize that
their respective kids are not well behaved.
Note that the announcement made by the class teacher is common knowledge and that is what
makes all the mothers cry on the fifth day.

1.4

Bounded Rationality

Osborne and Rubinstein [4] have the following to say regarding the important notion of bounded
rationality.
Game theory, in its most common form, assumes that all the players are symmetric; that is,
they have identical capabilities of perception and computation. It does not model asymmetries
in abilities or perceptions of situations.

 A good example of this is the game of chess. When analyzed using game theory, the game of
chess can be solved using an algorithm. This was in fact shown by Zermelo in 1913. Thus chess
becomes a trivial game for rational players. Zermelos result is that the game of chess has a
unique equilibrium outcome with the property that a player who follows the suggested strategy
is guaranteed that the outcome will be at least as good as the equilibrium outcome. Only the
existence of the equilibrium outcome has been shown by Zermelo. In fact, the actual equilibrium
outcome is yet to be computed. A game theoretic model of chess therefore reveals an important
fact about the game and suggests that it is not at all interesting for rational players. However,
the game theoretic model does not capture the asymmetric abilities of the players which is what
makes chess an interesting game.
Bounded rationality grapples with the problem of asymmetric abilities of the players and the fact that
the players do not have infinite computational resources at their disposal.

Classification of Games

Any subject like game theory which has been around for more than eight decades now will abound in
numerous kinds of games being defined and studied. There are innumerable classes of games and we
only provide a listing of some of the well known ones here.
Non-cooperative Games and Cooperative Games
Non-cooperative games are those in which the actions of individual players are the primitives; in
cooperative games, joint actions of groups of players are the primitives. John Harsanyi (1966) [5]
explained that a game is cooperative if commitments (agreements, promises, threats) among players
are enforceable and that a game becomes non-cooperative if the commitments are not enforceable.
Different Representational Forms
A strategic form game (also called simultaneous move game or normal from game), which we
have got introduced to in this chapter, is a model or a situation where each player chooses the
plan of action once and for all and all players exercise their decisions simultaneously.
An extensive form game specifies a possible order of events and each player can consider his plan
of action whenever a decision has to be made by him.
A coalitional form game or characteristic form game is one where every subset of players of players
is represented, by associating a value with each subset of players. This form is appropriate for
cooperative games.
Games with Perfect Information and Games with Imperfect Information
When the players are fully informed about the entire past history (each player, before making a move,
knows the past moves of all other players as well as his own past moves), the game is said to be of
perfect information. Otherwise the game is said to be with imperfect information.

Complete Information and Incomplete Information

A game with incomplete information is one in which, at the first point in time when the players can
begin to plan their moves, some players have private information about the game that other players
do not know. In a game with complete information, every aspect of the game is common knowledge.
Other Categorizations
There are many other categories of games, such as dynamic games, repeated games, stochastic games,
games with communication, multi-level games (Stackelberg games), differential games, etc. These
games are beyond the scope of our discussion but can be looked up in the references at the end of this
chapter.

To Probe Further

The material discussed in this chapter draws upon mainly from three sources, namely the books by
Myerson [1], Mascolell, Whinston, and Green [6], and Osborne and Rubinstein [4].
The classic treatise by John von Neumann and Oskar Morgenstern [2], published in 1944, provides a
comprehensive foundation for game theory. To this day, even after seven decades of its first appearance,
the book continues to be a valuable reference.
The first chapter of Myersons book [1] presents a detailed treatment of decision theory and utility
theory. We will be treating utility theory in detail in Chapter 6.
For an undergraduate level treatment of game theory, we recommend the books by Osborne [7],
Straffin [8], and Binmore [9]. For a graduate level treatment, we recommend the books by Myerson
[1] and Osborne and Rubinstein [4].

References
[1] Roger B. Myerson. Game Theory: Analysis of Conflict. Harvard University Press, Cambridge,
Massachusetts, USA, 1997.
[2] John von Neumann and Oskar Morgenstern. Theory of Games and Economic Behavior. Princeton
University Press, 1944.
[3] Robert J. Aumann. Agreeing to disagree. The Annals of Statistics, 4(6):12361239, 1976.
[4] Martin J. Osborne and Ariel Rubinstein. A Course in Game Theory. Oxford University Press,
1994.
[5] John C. Harsanyi. Games with incomplete information played by Bayesian players. Part I: The
basic model. Management Science, 14:159182, 1967.
[6] Andreu Mas-Colell, Michael D. Whinston, and Jerry R. Green. Micoreconomic Theory. Oxford
University Press, 1995.
[7] Martin J. Osborne. An Introduction to Game Theory. The MIT Press, 2003.
[8] Philip D. Straffin Jr. Game Theory and Strategy. The Mathematical Association of America, 1993.
[9] Ken Binmore. Fun and Games : A Text On Game Theory. D. C. Heath & Company, 1992.