CHAP 1: NORMAL FORM GAMES

What is a static game of complete information?

 In a static game, a set of players independently choose once-and-for-all actions, which in turn cause
the realization of an outcome. Thus a static game can be thought of as having two distinct steps:
Step 1: Each player simultaneously and independently chooses an action.
Players take their actions without observing what actions their counterparts take and without
interacting with other players to coordinate their actions.

Step 2: Conditional on the players’ choices of actions, payoffs are distributed to each player.
Once the players have all made their choices, these choices will result in a particular outcome,
or probabilistic distribution over outcomes. The players have preferences over the outcomes of
the game given by some payoff function over outcomes.
For example, if we are playing rock-paper-scissors and I draw paper while you simultaneously
draw scissors, then the outcome is that you win and I lose, and the payoffs are $0.10 for
example.

 What do we mean by complete information?

All players understand the environment they are in—that is, the game they are playing—in every way.

Games of Complete Information

A game of complete information requires that the following four components be common knowledge
among all the players of the game:
1. all the possible actions of all the players,
2. all the possible outcomes,
3. how each combination of actions of all players affects which outcome will materialize, and
4. the preferences of each and every player over outcomes.

Note : what exactly common knowledge means is by no means common knowledge. To make it clear,

Definition 1 An event E is common knowledge if (1) everyone knows E, (2) everyone knows that
everyone knows E, and so on ad infinitum.

Example : Pink hats

Requiring common knowledge : I want to predict how you will make your choice, given my belief that
you understand the game. Your understanding incorporates your belief about my understanding, and so
on.
Definition 2 A Static Game of Complete Information is a static game in which the way players’ actions
affect each player’s payoff is common knowledge to the players of the game.

Section 1 : Normal-Form Games with Pure Strategies

A strategy is defined as a plan of action intended to accomplish a specific goal.

Definition 3 A pure strategy for player i is a deterministic plan of action. The set of all pure strategies
for player i is denoted Si. A profile of pure strategies s = (s1, s2,. . ., sn), si ∈ Si for all i = 1, 2,. . ., n,
describes a particular combination of pure strategies chosen by all n players in the game.

=
Note 1: Attention is restricted to the case in which players choose deterministic actions: “pure”
strategies: An alternative to choosing one of your pure strategies would be for you to choose actions
stochastically → Chap 6 of TADELIS
Note 2: The concept of strategy is useful for games in which there will be relevance to conditioning
one’s actions on events that unfold over time, → Chapter 7 of TADELIS. In static games players
choose actions once and for all and simultaneously, so strategies are actions

Definition 3 A normal-form game includes three components as follows:

1. A finite set of players, N = {1, 2,..., n}.

2. A collection of (possibly infinite) sets of pure strategies, {S1, S2,..., Sn}.

3. A set of payoff functions, {v1, v 2 , . . . , vn}, each assigning a payoff value to each combination of
chosen strategies, that is, a set of functions vi : S1 × S2 ×. . . × Sn → R for each i ∈ N .

Thus from now on the normal-form game will be a triple of sets:

Example: The Prisoner’s Dilemma

Players: N = {1, 2}.

Strategy sets: Si = {M, F } for i ∈ {1, 2}. M; Mums, F: Finks

Payoffs: Let vi(s1, s2) be the payoff to player i if player 1 chooses s1 and player 2 chooses s2. We can
then write payoffs as

v1(M, M) = v2(M, M) = −2
v1(F, F) = v2(F, F) = −4
v1(M, F) = v2(F, M) = −5
v1(F, M) = v2(M, F) = −1.
Example: Cournot Duopoly
A variant of this example was first introduced by Augustin Cournot (1838).
Two identical firms, players 1 and 2, produce some good.

Assume that there are no fixed costs of production, and let the variable cost to each firm i of producing
quantityqi ≥ 0 be given by the cost function, ci(qi) = qi2 for i = 1, 2

Demand is given by the function q = 100 – p where

i q = q1 + q 2

The actions are choices of quantity, and the payoffs are the profits. Hence the following represents the
normal form:

Players: N = {1, 2}.

Strategy sets: Si = [0, ∞] for i ∈ {1, 2} and firms choose quantities si ∈ Si .
Payoffs: For i, j ∈ {1, 2}, i ≠ j ,

We are implicitly assuming that prices cannot fall below zero, so that if the firms together produce a
quantity that is greater than 100, the price will be zero and each firm’s payoffs are its costs.

Section 2 : Matrix Representation: Two-Player Finite Game

In many cases, players may be constrained to choose one of a finite number of actions. This is the case
for the Prisoner’s Dilemma, rock-paper-scissors, and many more strategic situations.
Even when players have infinitely many actions to choose from, we may restrict attention to a finite
number of actions. If we think of the Cournot duopoly example, the demand function p = 100 - q
suggests that flooding the market with more than 100 units will cause the price of the product to drop to
zero. This means that we have effectively restricted the strategy set to a finite number of strategies
(101), to be accurate, for the quantities 0, 1,..., 100).
Definition 3.4 A finite game is a game with a finite number of players, in which the number of
strategies in Si is finite for all players i ∈ N .
Any two-player finite game can be represented by a matrix that will capture all the relevant information
of the normal-form game. This is done as follows:
Rows Each row represents one of player 1’s strategies. If there are k strategies in S1 then the
matrix will have k rows.

Columns Each column represents one of player 2’s strategies. If there are m strategies in S2
then the matrix will have m columns.

Matrix entries Each entry in this matrix contains a two-element vector (v1, v2), where vi is
player i’s payoff when the actions of both players correspond to the row and column of that
entry.

Note, however, that neither the Cournot duopoly nor the voting example (in book) can be represented by
a matrix. The Cournot duopoly is not a finite game (there are an infinite number of actions for each
player), and the voting game has more than two players.
Example: The Prisoner’s Dilemma

Notice that all the relevant information appears in this matrix.

Example: Rock-Paper-Scissors

Remark Such a matrix is sometimes referred to as a bi-matrix. In a traditional matrix, by definition,

each entry corresponding to a row-column combination must be a single number, or element, while here
each entry has a vector of two elements—the payoffs for each of the two players.

Section 3 : Solution Concepts

We want be able to make some prediction, or be able to prescribe behavior, about how players should
(or will) play.

Example : the Battle of the Sexes, introduced by R. Duncan Luce and Howard Raiffa (1957). The game
is summarized in the following matrix:

We need to make assumptions on behavior and on beliefs of the players → in search of a solution
concept, that will result in predictions or prescriptions.
 files that emerge as one of the solution concept’s
An equilibrium is any one of the strategy pro
predictions.

Assumptions and Setup

Equilibrium analysis requires:

1. Players are “rational”: A rational player is one who chooses his action, si ∈ Si , to maximize
his
∈ payoff consistent with his beliefs about what is going on in the game.
2. Players are “intelligent”: An intelligent player knows everything about the game: the actions,
the outcomes, and the preferences of all the players.
3. Common knowledge: The fact that players are rational and intelligent is common knowledge
among the players of the game.
4. Self-enforcement: Any prediction (or equilibrium) of a solution concept must be self-
enforcing. No third party that can enforce any profile of actions that the players will not
voluntarily agree to stick to. The requirement is at the heart of noncooperative game theory.

Evaluating Solution Concepts

→ Three criteria that evaluate a variety of solution concepts: existence, uniqueness, and invariance.
Existence: How Often Does It Apply? We require it to result in the existence of an equilibrium
solution.
Uniqueness: How Much Does It Restrict Behavior? we require that it be meaningful in that it
restricts the set of possible outcomes to a smaller set of reasonable outcomes.
Invariance: How sensitive Is It to Small Changes? That the solution concept be invariant to small
changes in the game’s structure. However, the term “small changes” needs to be qualified more
precisely.It is reasonable to argue that if the payoffs of a game are modified only slightly, then this is a
small change to the game that should not affect the predictions of a “robust” solution concept.

Evaluating Outcomes
Economists use a particular criterion for evaluating whether an outcome is socially undesirable. An
outcome is considered to be socially undesirable if there is a different outcome that would make some
people better off without harming anyone else.
→ criterion of Pareto optimality, which is in tune with the idea of efficiency or “no waste.”
That is, we would like all the possible value deriving from a given interaction to be distributed
among the players.
For example, in the Prisoner’s Dilemma we made the case that (F, F) should be considered as a very
likely outcome. Note it is Pareto dominated by (M, M). (Notice that (M, M) is not the only Pareto-
optimal outcome. (M, F) and (F, M) are also Pareto-optimal outcomes because no other profile
dominates any of them. Don’t confuse Pareto optimality with the best “symmetric” outcome that leaves
all players “equally” happy.)