Professional Documents
Culture Documents
Games and Decision Making
Games and Decision Making
Until now, our decisions were up to us, not influenced by the actions of other players
of the economy. What strategy are we to follow if the possible actions of other players
are to be considered?
EXAMPLE: Suppose Alitalia and Lufthansa are thinking about pricing a round-trip
airfare from Berlin to Rome.
If both airlines charge a price of 300 €, then the profit of Alitalia would be 50 million
€ per year, and the profits of Lufthansa would be 100 million € per year. Both are
thinking in 200 or 300 € ticket price.
The predicted earnings of the two companies with regard to the ticket prices are
displayed in the following table. The first number refers to the earnings of Alitalia,
while the second component refers to the earnings of Lufthansa.
LUFTHA NSA
Fare 300€ 200€
ALIT 300€ (50, 100) (-100, 200)
ALIA 200€ (150, -200) (-10, -10)
If they deviate from the 300 – 300 € prices, at least one of them is going to suffer great
losses, hence the best thing (for them) to do is to fix prices by coordinating their
policies.
This is not what we are going to assume. We consider a model, where participants are
not forming alliances. In this case the defining properties of the model are:
• There are two or more participants,
• Each participant has a (finite) set of alternative choices,
• For each outcome, there is a payoff that each participant gets.
A game formed along the above principles is called a Strategic Form Game. Here each
player knows all the data. We are to decide what strategy is best for each player to
choose?
Two persons, who have committed a crime have a choice of either confessing to the
crime (= to fink), or keep silent (= to mum).
If one keeps mum and the other finks, the one who confessed can go free, while the
one who did not speak gets 10 years. If none confesses, they both get a sentence of 1
year, while if both confess, their sentence is 5 years each.
Practically, both have a stake in keeping mum. (The sum of their prison terms is
smallest.)
On the other hand, if one knows that the other keeps mum, then he has the incentive to
fink.
Pl ay er 2
Stra b1 b2 …. bn
-
tegy
Pl a1 [u1 (a1 , b1 ), u 2 (a1 , b1 )] [u1 (a1 , b2 ), u 2 (a1 , b2 )] …. [u1 (a1 , bn ), u 2 (a1 , bn )]
.
a a2 [u1 (a2 , b1 ), u 2 (a2 , b1 )] [u1 (a2 , b2 ), u 2 (a2 , b2 )] …. [u1 (a2 , bn ), u 2 (a2 , bn )]
y
er ….
1 am [u1 (am , b1 ), u 2 (am , b1 )] [u1 (am , b2 ), u 2 (am , b2 )] …. [u1 (am , bn ), u 2 (am , bn )]
It is never worth for a player to play a strictly dominated strategy. Thus, we throw out
strictly dominated strategies.
Pla yer 2
Strategy L C R
Pla T (1 , 0) (1 , 3) (3 , 0)
yer M (0 , 2) (0 , 1) (3 , 0)
1 B (0 , 2) (2 , 4) (5 , 3)
Pla yer 2
Strategy L C
Pla T (1 , 0) (1 , 3)
yer M (0 , 2) (0 , 1)
1 B (0 , 2) (2 , 4)
In the new table we see that T strictly dominates M for Player 1. Eliminate M!
Pla yer 2
Strategy L C
Pla T (1 , 0) (1 , 3)
yer 1 B (0 , 2) (2 , 4)
Player 2
Strategy C
Pla T (1 , 3)
yer 1 B (2 , 4)
Since B now strictly dominates T for Player 1, we see that the solution of the matrix game is
the pair of strategies B for Player 1 and C for Player 2 with returns 2 and 4 respectively.
EXAMPLE 2. Unfortunately, there are not always strictly dominated strategies. In the game
Pla yer 2
Strategy U W
Pla X (1 , 1) (0 , 0)
yer 1 Y (0 , 0) (1 , 1)
we don’t even have dominated strategies.
Mrs. Smith wants to go to the opera. Mr. Smith wants to go to a box match, still
they prefer to go together. The situation is described in the following table:
Mrs. Smith
Strategy Opera Box
Mr. Opera (1 , 2) (0 , 0)
Smith Box (0 , 0) (2 , 1)
In this case they have no strictly dominated, not even dominated strategies. (The decision is
up to them.)
Theorem 2.4 If a game can be solved using iterated elimination of strictly dominated
strategies, then the solution is the unique Nash equilibrium of the game.
Until now we had only two players and a finite set of strategies for both of them. The
principles of solving a matrix game can be generalized for many players with possibly
infinite strategy sets.
Definition 3.1 A strategic form game (= a game in normal form) is a set of n players
labeled 1, 2 , ...., n such that each player has
a) a choice set S i , the strategy set of player i ,
b) a payoff function ui : S1 × S 2 ×.... S n → R .
A player’s objective is to maximize her payoff. We assume that all players know all
the strategies and payoff functions. A solution, a choice of strategy for each player is
considered to be optimal, if none of the players want to deviate from it. Such a
strategy profile (or strategy combination) is called a Nash equilibrium.
Definition 3.1 A Nash equilibrium of a strategic form game is a strategy profile
( )
s1∗ , s2∗ ,..., s n∗ ∈ S1 × S 2 ×.... S n such that for each player i
ui (s1∗ ,.., si∗ ,., s n∗ ) ≥ ui (s1∗ ,.., si ,..., sn∗ )
for all si ∈ S i .
( ) ( )
In other words, s1∗ , s2∗ ,..., s n∗ is a Nash equilibrium, if s1∗ , s2∗ ,..., s n∗ is the place of a
th
local maximum for the function ui along the i coordinate. If the game has a unique
Nash equilibrium, then it constitutes an optimal strategy profile. Translated to the
language of partial derivatives:
Suppose the strategy sets are open intervals for each player in a strategic form game,
and the payoff functions are twice differentiable. A strategy profile s1∗ , s2∗ ,..., s n∗ is the ( )
unique Nash equilibrium for the game if
( )
a) (u i ) s s1∗ , s 2∗ ,..., s n∗ = 0 for each player i
i
b) For each i : si∗ is the only stationary point for ui in the i th coordinate, and
( )
c) (u i ) s s s1∗ , s 2∗ ,..., s n∗ < 0 for each i .
i i