Chapter 13 Introduction to Game Theory. -...
Transcript of Chapter 13 Introduction to Game Theory. -...
Chapter 13 – Introduction to
Game Theory.Goals:
+ Set Up a Game.
+ Solve a Game:
Dominant Strategy Equilibrium.
Nash Equilibrium.
Sub-game Perfect Nash Equilibrium.
Describing the Game.
Elements of a game:
◦ The players
◦ Timing of a game: Simultaneous or Sequential.
◦ Information Availability: Perfect or Imperfect.
◦ The list of possible strategies for each player.
◦ The payoffs associated with each combination
of strategies.
◦ Repetition.
◦ The decision rule.
Presentation of a Game.
Extensive-form presentation (a game tree) of
a game:
◦ The players
◦ Their possible actions
The set of possible actions
The sequence of actions
The information the players have when they decide
◦ The outcome following the actions, i.e. payoffs for
all players
Presentation of a Game – extensive
form. Example 1: A sequential game.
Presentation of a Game – extensive
form. Example 2: A simultaneous-move game.
Presentation of a Game.
Normal-form or strategic-form or pay-off
matrix presentation.
◦ The players
◦ Their possible strategies
◦ The outcome following these strategies
Presentation of a Game – Normal
Form Example 3: The normal-form representation
TOSHIBA
DOS UNIX
IBM DOS (600,200) (100,100)
UNIX (100, 100) (200,600)
Common Games.
Prisoner’s Dilemma
PRISONER’s DILEMMA. Prisoner 2
Stay Silent Betray
Prisoner 1 Stay Silent Each serves 1
year
Prisoner 1:
Serves 20 years.
Prisoner 2: Go
free.
Betray Prisoner 1: Go
free.
Prisoner 2:
Serves 20 years
Each serves 10
years.
Common Games.
Battle of Sexes
Battle of Sexes Wife
Hockey Game Ballet
Performance
Husband Hockey Game (100,50) (20,20)
Ballet
Performance
(0,0) (50,100)
Equilibrium of a Game.
Equilibrium of a game: a situation in which no
player wishes to change his/her strategy.
We study three equilibrium concepts in
Econ 203:
◦ Dominant strategy equilibrium.
◦ Nash equilibrium.
◦ Sub-game perfect Nash equilibrium.
Equilibrium of a Game – Dominant
Strategy Equilibrium. Dominant strategy equilibrium.
◦ Consider the following game:
◦ The players: Player 1 and Player 2
◦ Timing of a game: Simultaneous.
◦ Information Availability: Perfect.
◦ Possible strategies: Player 1 (L;R) and Player 2(U;D)
◦ The payoffs associated with each combination of strategies (table
below)
◦ Repetition: Non - repeated.
◦ The decision rule: Max own payoff.
◦ Equilibrium: (Player 1, Player 2) = (R;D) with payoff (6,3)
Player 2
(U) (D)
Player 1 L (2,2) (4,4)
R (0,1) (6,3)
Equilibrium of a Game – Nash
Equilibrium. Dominant Strategy does not always exist.
The Nash equilibrium (NE)◦ Set of strategies (one for each player) such that
no player wishes to change her strategy given the strategies of the other players The strategy of each player is a so-called best response
to the given strategies of the other.
Equilibrium of a Game – Nash
Equilibrium. Nash equilibrium.
◦ Consider the following game:
◦ The players: Player 1 and Player 2
◦ Timing of a game: Simultaneous.
◦ Information Availability: Perfect.
◦ Possible strategies: Player 1 (L;R) and Player 2(U;D)
◦ The payoffs associated with each combination of strategies (table
below)
◦ Repetition: Non - repeated.
◦ The decision rule: Max own payoff.
◦ Nash Equilibria: (Player 1, Player 2) = (R;U) and (L;D) with
payoff (3;6) and (6;3)
Player 2
(U) (D)
Player 1 L (0,2) (6,3)
R (3,6) (0,2)
Equilibrium of a Game – Nash
Equilibrium.◦ Consider the following game:
◦ The players: Player 1 and Player 2
◦ Timing of a game: Sequential with Player 1 moves first.
◦ Information Availability: Perfect.
◦ Possible strategies: Player 1 (L,R) and Player 2(UU,UD,DU,DD)
◦ The payoffs associated with each combination of strategies (table
below)
◦ Repetition: Non - repeated.
◦ The decision rule: Max own payoff.
◦ Equilibria: (Player 1, Player 2) = (R;U,U), (L;U,D), (L;D,D)
with payoff (3;6), (6;3) and (6;3)
◦ 1 equilibrium is not valid…which one?
Equilibrium of a Game – Nash
Equilibrium. Extensive Presentation – Game Tree.
Normal Form Presentation – Payoff Matrix.
Player 2
(U,U) (U,D) (D,U) (D,D)
Player 1 R (3,6) (3,6) (0,2) (0,2)
L (0,2) (6,3) (0,2) (6,3)
Equilibrium of a Game – Sub-Game
Perfect Nash Equilibrium Sub-game Perfect Nash equilibrium
◦ A player’s best response to a given strategy played by another
player.
Q: How do I find a sub-game perfect Nash equilibrium?
A: Take game tree and use method called backward
induction
◦ Remember, that the SPNE is a set of strategies, not an outcome
or a sequence of actions
The SPNE: (Player 1, Player 2) = (L;U,D)
What are the other two NE?
◦ They are non-credible threats.
Equilibrium of a Game – Sub Game
Perfect Nash Equilibrium. Extensive Presentation – Game Tree.
Normal Form Presentation – Payoff Matrix.
Player 2
(U,U) (U,D) (D,U) (D,D)
Player 1 R (3,6) (3,6) (0,2) (0,2)
L (0,2) (6,3) (0,2) (6,3)
SPNE
Game Theory
What can we conclude?
◦ Always use normal form/pay-off matrix to solve a
simultaneous game.
First identify if there exists dominant strategy for any of the
players.
If there is, the equilibrium must contain such dominant strategy.
If no dominant strategy exists, solve the game using the very
definition of Nash Equilibrium: my best response given your
action.
◦ Always use the game tree to solve a sequential game
using backward induction.
The equilibrium is a sub-game perfect nash equilibrium.
Eliminate all the non-credible threats.
Duopoly - Revisited
Consider the following game:
◦ The players: Firm 1 and Firm 2 with same cost functions.
◦ Timing of a game: Simulteneous
◦ Information Availability: Perfect.
◦ Possible strategies: Firm1 (PB,PM) and Firm 2(PB,PM)
◦ The payoffs associated with each combination of strategies
(table below)
◦ Repetition: Non - repeated.
◦ The decision rule: Max own payoff.
Nash Equilibrium: (Player 1, Player 2) = (PB;PB) with
payoff (1,1) Non-cooperate the lowest
welfare outcome.
How to induce cooperation?
Duopoly - Revisited
Penalty:
◦ Suppose there is a enforceable penalty of 5 charged to the firm
who does not comply with the agreement (always play PM) and is
given to the one that suffers from that noncompliance.
Result: Cooperative outcome.
Firm 2
PB PM
Firm1 PB (1,1) (10-5=5,0+5=5)
PM (0+5=5,10-5=5) (6,6)
Duopoly - Revisited
What if the penalty of 5 is not enforceable?
◦ Punishment imposed by the other firm.
Tit-for-Tat: My action in this game depends on your action in the
previous game.
Grim Strategy: once a cheater always a cheater. Punishment sustains
for the rest of the game.
Consider the previous game with number of repetition (round) = 10.
Firm 1 chooses to cooperate or not at round 10.
If cooperate = 6 (and the other cooperates)
If non-cooperate = 10 (and the other cooperates)
Non-cooperate at round 10 for both firms.
Move backward. Firm 1 chooses to cooperate or not at round 9
If cooperate = 6 + 0 = 6
Non cooperate = 10 + 0 = 10
Non-cooperate at round 9 and 10 for both firms.
Keep moving backward and we will see that two firms will not cooperate in any of the round under grim strategy with 10 rounds.
How to get a cooperative outcome?
Duopoly - Revisited◦ Grim Strategy: once a cheater always a cheater.
Punishment sustains for the rest of the game.
Consider the previous game with number of repetition (round) =
infinity.
Firm 1 chooses to cooperate or not at round 1.
If cooperate = 6 + 6 + 6 + 6 + … (and the other cooperates)
If non-cooperate = 10 + 0 + 0 + 0 + … (and the other cooperates)
Cooperate at round 1 for both firms.
Move forward. Firm 1 chooses to cooperate or not at round 2 If cooperate = 6 + 6 + 6 + … = infinity
Non cooperate = 10 + 0 + 0 + … = 10
Cooperate at round 1 and 2for both firms.
Keep moving forward and we will see that two firms will always cooperate under grim strategy with infinity number of rounds.