Game Theory

• Definition of a game:
• A set of N players
• A set of strategies for each player
• For every player, a payoff function mapping a strategy
profile to a real number
 The Prisoner’s Dilemma
Prisoner B
Defect Cooperate
-4, -4 0, -5
Defect
Prisoner A
Cooperate -5, 0 -1, -1

• What will each player do?

• Strictly dominant strategy for each player - Defect
• Maximin strategy maximizes the minimum payoff
• Staggering moves does not change the outcome
 Nash Equilibrium - 1
• Nash equilibrium is a solution concept for a game
• Consider a game in which the players have to select
their strategies simultaneously at the beginning of the
game
• In such a game, a Nash equilibrium is a vector of
strategies – one for each player – such that each player’s
strategy is optimal for her (i.e., maximizes her payoff)
given the specified strategies of all other players
 Nash Equilibrium - 2
• A player’s “best-response correspondence” is the
mapping which identifies the player’s “best-response
set” for every conjecture that she can hold about rival
behaviour
• A common way to justify Nash equilibria as plausible
predictions of actual play is to appeal to a process of
pre-play communication between players, possibly with
the mediation of an ‘impartial consultant’
• The interpretation of a Nash equilibrium as a pre-play
self-enforcing strategy recommendation by an
impartial observer has been one of the strongest
justifications for the equilibrium concept
 The Prisoner’s Dilemma
Prisoner B
Defect Cooperate
1, 1 5, 0
Defect
Prisoner A
Cooperate 0, 5 4, 4
• Unique Nash equilibrium – (Defect, Defect)
• Equilibrium payoffs – 1 to each
 Nash Equilibrium - 3
However, we should probably be more interested in how a rational player
thinks while playing a game? Does her “rational introspection” about how
the other players will play the game lead her inexorably to play a Nash
equilibrium strategy? Unfortunately, there are good enough reasons to
think that the answer will oftentimes be “no”. In fact, in a survey article on
Game Theory, Elon Kohlberg (a leading game theorist at Harvard
University) writes thus:
Given a game, can we predict what rational players will actually do? Except
when the game is unusually simple, our answer must be “No!” The reason,
of course, is that in determining his own choice each player must take into
account what the others will do. This leads to an endless chain of
deductions about the other players, about the other players’ deductions,
etc.
The main idea of game theory, due to Cournot (1838) and Nash (1951), is
to make a bold simplification and, instead of asking how the process of
deductions might unfold, asks where its rest points might be.
 Coordination Games
• In many strategic situations, the critical issue for players
is to achieve coordination in strategy choices
- in such games, “failing to coordinate” generally
gives worse payoffs as compared to coordinated outcomes
• Coordination games have multiple Nash equilibria (as
there are typically multiple ways to coordinate)
• Different kinds of coordination games differ in
the “extent of conflict” inherent in coordination, and in
the extent to which uncoordinated outcomes are worse
than coordinated outcomes
 Pure Non-conflicting Coordination
Driver B
Left Right
2, 2 0, 0
Left
Driver A
Right 0, 0 2, 2

• Nash equilibria – (Left, Left) and (Right, Right)

• Equilibrium payoffs – 2 to each player in each eqbm
 Pure Non-conflicting Coordination
• Multiple Pareto-equivalent coordinated Nash equilibria
• For a few familiar players, ‘pre-play communication’
and/or ‘any staggering of order of moves’ will enable
coordination but it’s a true coordination problem for
many anonymous players
 Ranked Non-conflicting Coordination
Typist B
Dvorak Qwerty
2, 2 0, 0
Dvorak
Typist A
Qwerty 0, 0 1, 1

• Nash equilibria: (Dvorak, Dvorak), (Qwerty, Qwerty)

• Eqbm payoffs: 2 to each player in (Dvorak, Dvorak),
1 to each player in (Qwerty, Qwerty)
 Ranked Non-conflicting Coordination
• Multiple Pareto-ranked coordinated Nash equilibria
• For a few familiar players, ‘pre-play communication’
and/or ‘any staggering of order of moves’ will enable
coordination but it’s a true coordination problem for
many anonymous players
 Conflicting Coordination I:
Battle of the Sexes
Wife
Spy movie Concert
5, 3 2, 2
Spy movie
Husband
Concert 1, 1 3, 5

• Nash equilibria: (Spy movie, Spy movie), (Concert,

Concert)
• Eqbm payoffs: 5 to Husband, 3 to Wife in (Spy
movie, Spy movie), 3 to Husband, 5 to Wife in
(Concert, Concert)
 Conflicting Coordination I:
Battle of the Sexes
• Conflict in coordinating on ‘identical actions’: multiple
Pareto-incomparable coordinated Nash equilibria
• If players can stagger their moves, that coordinated
outcome which favours the ‘first-mover’ will obtain as the
unique SPNE
• Pre-play communication before a one-time interaction
can lead to ‘haggling’ among players to select a specific
coordination
• Similar issues arise when firms have to adopt a common
‘standard’, and every standard is proprietary to a specific
firm (e.g., Sony’s Blu-Ray and Toshiba’s HD DVD
technologies)
 Conflicting Coordination II:
Minority Game
You
El Farol bar Another bar
2, 2 5, 3
El Farol bar
Me
Another bar 3, 5 1, 1

• Nash equilibria: (El Farol bar, Another bar), (Another

bar, El Farol bar)
• Eqbm payoffs: 5 to Me, 3 to You in (El Farol bar,
Another bar), 3 to Me, 5 to You in (Another bar, El
Farol bar)
 Conflicting Coordination II:
Minority Game
• Conflict in coordinating on ‘dissimilar actions’ which are
differentially attractive: multiple Pareto-incomparable
coordinated Nash equilibria
• Issues of ‘move staggering’ and ‘pre-play
communication’ are similar to those encountered in the
Battle of the Sexes
• A similar problem arises when two firms decide on entry
in a ‘new’ lucrative market – each wants to be the only
one to enter
 Conflicting Coordination III:
Game of Chicken
Driver B
Hold course Swerve
-100, -100 5, 1
Hold course
Driver A
Swerve 1, 5 2, 2

• Nash equilibria: (Hold course, Swerve), (Swerve,

Hold course)
• Eqbm payoffs: 5 to Driver A, 1 to Driver B in (Hold
course, Swerve), 1 to Driver A, 5 to Driver B in
(Swerve, Hold course)
 Conflicting Coordination III:
Game of Chicken
• Coordinating on differentially attractive ‘dissimilar actions’:
multiple Pareto-incomparable coordinated Nash eqbm
• Issues of ‘move staggering’ and ‘pre-play communication’
are the same as in Battle of the Sexes Game and the
Minority Game
• Games of Chicken are ubiquitous in social/professional
situations where one player has to “volunteer” to perform
an onerous task
• A similar problem arises in a situation where 2 firms find
themselves competing in ‘too small’ a market, and ‘exit’ is
individually more profitable than continual duopoly
competition