Professional Documents
Culture Documents
2 - Normal Form Games - 2
2 - Normal Form Games - 2
Game theory: set of tools for the analysis of interactive decision making (conflict-
ing/mutual interests).
• The players
• The actions/strategies
• The payoffs
{
- purposeful;
Important assumption: the players are rational →
- no indecision, no mistakes.
Given the rules of the game, the players optimize a goal expressed through their
payoffs.
Player 2
Heads T ails
Heads − 1, +1 +1, −1
Player 1
T ails + 1, −1 −1, +1
• Penalty kick
Objective: to determine the strategies that the players should play as well as the
outcome(s) that should arise.
1
Chapter I: Strategic games in normal form
• For each i ∈ I, Si is a non-empty set and is called player’s i strategy set (or action
set).
• For each i ∈ I, ui : S → IR is player i’s payoff function. The amount ui (s) is i’s
payoff (or utility) if the joint strategy s is played.
Prisoner 2
Quiet F ink
Quiet − 2, −2 −10, −1
Prisoner 1
F ink − 1, −10 −5, −5
The payoffs are ordinal (only the ranking of the joint strategies matters)
2
Example 3 : Cournot Oligopoly (linear price and linear costs)
Consider n firms operating in the same market. The inverse demand is given by:
{
a − bQ if Q ≤ a/b
P (Q) =
0 if Q > a/b
∑n
where Q = i=1 qi is the total output produced and a, b > 0.
3
I-2 Solution concepts
Dominated strategies
[ ]
Definitions Consider a game Γ = I, (Si )i∈I , (ui )i∈I
1- Player i’s strategy s̄i is weakly dominated if there exists another strategy ŝi ∈ Si
such that:
ui (s̄i , s−i ) ≤ ui (ŝi , s−i ), for any s−i ∈ S−i ,
with the strict inequality for at least one s−i .
• In case the inequality holds strictly for all s−i ∈ S−i , s̄i is said to be strictly
dominated (by ŝi ).
• It is also common to say that ŝi (strictly/weakly) dominates s̄i .
2- Player i’s strategy ŝi is a weakly dominant strategy if it weakly dominates all the
other strategies of i. That is, if for any s̄i ̸= ŝi , we have:
In case ŝi strictly dominates all the other strategies of player i, we say that ŝi is a
strictly dominant strategy.
Prisoner 2
Quiet F ink
Quiet − 2, −2 −10, −1
Prisoner 1
F ink − 1, −10 −5, −5
For each prisoner, the strategy Quiet is (weakly and strictly) dominated by Fink.
The outcome of the procedure IDSDS is thus ( Fink, Fink ).
4
→ Iterated Deletion of Weakly Dominated Strategies (IDWDS)
Player 2
L C R
T 4, 2 2, 2 0, 0
Player 1 M 2, 4 6, 2 4, 2
B 4, −4 2, −2 −2, −2
No dominant strategy. For player, B is weakly dominated (by T). For player 2, R is
weakly dominated (by C).
The outcome depends on the order of deletion. Eliminating B first gives the unique
joint strategy (T, L) as the outcome of IDWDS.
5
Best responses and Nash equilibrium
[ ]
Consider a game Γ = I, (Si )i∈I , (ui )i∈I .
Definition
- For every s−i ∈ S−i , the strategy si ∈ Si is called a best response of player i against
s−i if we have:
ui (si , s−i ) ≥ ui (s′i , s−i ), for any s′i ∈ Si .
- The set of best responses of player i against s−i is denoted by Bi (s−i ) and the multi-
valued function Bi is called the best response correspondence of player i.
Remarks
1- Every element of Bi (s−i ) must solve the problem: max ui (s′i , s−i ).
s′i ∈Si
2- A dominant strategy of player i is necessarily a best response against any s−i ∈ S−i .
Definition
A Nash equilibrium of the game Γ is a joint strategy s∗ = (s∗i )i∈I ∈ S such that, for any
i ∈ I,
ui (s∗i , s∗−i ) ≥ ui (si , s∗−i ), for any si ∈ Si .
6
The Prisoner’s Dilemma (Example 2)
Prisoner 2
Quiet F ink
Quiet − 2, −2 −10, −1
Prisoner 1
F ink − 1, −10 −5, −5
→ If q2 ≥ a−c
b
, check that firm 1’s best response is 0.
For firm 2,
{
a−c
2b
− q21 if q1 < a−cb
b2 (q1 ) =
0 if q1 ≥ a−c .
b {
q1 = b1 (q2 )
By solving the system of equations which defines the Nash equilibrium
q2 = b2 (q1 ),
∗
one obtains the Nash equilibrium q = ((a − c)/3b, (a − c)/3b).
7
I-3 Further illustrations of the Nash equilibrium (NE)
where a > c.
→The best response of firm i solves the profit maximization problem:
∑ ∑
n
[a − b(qi + qj )]qi − cqi if qj ≤ a/b
j̸=i j=1
max πi (qi , q−i ) = ∑
n
qi ≥0
− cqi
if qj > a/b,
j=1
∑ a−c
Similar to the case n = 2. For qj < b
, the FOC for firm i is
j̸=i
∑ a−c 1∑
a − 2bqi − b qj − c = 0, i.e. qi = − qj
j̸=i
2b 2 j̸=i
8
Auctions
→ The players are the n bidders (n ≥ 2), with valuations v1 > v2 > ... > vn .
→ The highest bidder wins the object and pays her bid to the seller. In case the
highest bid is submitted by many bidders, the object is allocated to the one with
the highest valuation.
→ The payoffs: if player i wins the object, her utility is vi − bi ; otherwise her utility
is 0.
There are many NE. One of them is (b1 , b2 ..., bn ) = (v2 , v2 , ..., vn )
→ No incentive for bidder 1 to deviate
– Choosing b1 < v2 would lose her the object (bidder 2 would win)
– Choosing b1 > v2 (against (b2 = v2 , ..., bn = vn )) would win her the object,
and result in the payoff v1 − b1 < v1 − v2
9
Example 5: Second-price, sealed-bid auction (complete information)
→ The players are the n bidders (n ≥ 2), with valuations v1 > v2 > ... > vn .
→ The highest bidder wins the object. In case the highest bid is submitted by many
bidders, the object is allocated to the one with the highest valuation.
→ The payoffs: if player i wins the object, her utility is vi − max bj ; otherwise her
j̸=i
utility is 0.
In fact, bidding one’s valuation is a weakly dominant strategy in the second price
auction.
10
Figure 1: vi is the dominant bid for player i, b̄ = max bj
j̸=i
• Multiple NE
Player 2
L M R
L 2, 2 0, 0 1, 0
Player 1 M 0, 0 −1, −1 0, 0
R 0, 1 0, 0 1, 1
11
• Existence might be a problem (Matching Pennies, Example 1)
Player 2
Heads T ails
Heads − 1, +1 +1, −1
Player 1
T ails +1, −1 −1, +1
12
I-4 Mixed strategies
3- The concept of expected utility is used to extend the payoff function to mixed strategy
profiles. For any a mixed strategy profile σ = (σ1 , ..., σn ), the expected payoff of player
i is
∑
ui (σ) = σ1 (s1 )...σn (sn )ui (s)
s∈S
13
Definitions
[ ]
In what follows we consider the mixed extension I, (∆(Si ))i∈I , (ui )i∈I .
1- Player i’s mixed strategy σ̄i is strictly dominated if there exists another mixed
strategy σ̂i ∈ ∆(Si ) such that:
2- Given any player i and any mixed strategy profile σ−i of the other players, a
strategy σi ∈ ∆(Si ) is called a best response of i against σ−i if it satisfies:
3- A mixed strategy profile σ ∗ is a mixed strategy Nash equilibrium of Γ if, for every
player i, σi∗ is a best response of i against σ−i
∗
.
14
Example: Best responses for Matching Pennies
15
We will say that a game Γ = [I, (Si )i∈I , (ui )i∈I ] is finite if I is finite and Si is finite for
any i ∈ I.
Remarks:
1- Finiteness of the game is important (a game that is infinite may have no NE)
2- There may be many mixed strategy NE.
3- The NE mentioned in the proposition might be constituted of pure strategies, even if
it is the only one.
How can one check that a given strategy profile is a mixed strategy Nash equilibrium?
The following proposition provides the answer.
• every pure strategy si ∈ Si which is given positive probability by σi∗ yields the
∗ ∗
same expected payoff against σ−i , that is to say ui (si , σ−i ) = ui (σ ∗ );
• every pure strategy si ∈ Si which is given zero probability by σi∗ yields no more
∗
than the pure strategies that are assigned positive probability: ui (si , σ−i ) ≤ ui (σ ∗ ).
16
EXAMPLE
Consider the 2x3 matrix game
Player 2
L C R
T 4, 2 0, 1 −1, −2
Player 1
B 4, −2 0, 1 −1, 2
17
Definitions
1- An n-player game in normal form is symmetric if S1 = S2 = ... = Sn and the n players
all have the same payoff function u(si , s−i ).
PROPOSITION
Every finite, symmetric game in normal form has a symmetric mixed strategy Nash
equilibrium.
Remarks:
1-The symmetric NE may be in pure strategies.
2- There may be other asymmetric NE too.
Example:
Player 2
A B
A 2, 2 4, 3
Player 1
B 3, 4 2, 2
The game is symmetric. The only pure strategy NE are (A, B) and (B, A) (neither of
them is symmetric)
18