Game Theory using AI/ML

Syllabus
Syllabus
Brief introduction of Game
theory
 The book

John von Neumann and Oskar

Morgenstern - Game theory
became a field since their
book was published in 1944.

John Nash – Received his Ph.D.

from Princeton University

Invented the notion of Nash

equilibrium.
What is Game Theory?

Game theory – helps to understand situations in

which decision-makers interact
Game – a competitive activity where players
contend with each other based on a set of rules
Players & Applications:
Business firms competing for business
Political candidates competing for votes
Animals fighting over prey
Jury deciding a verdict
Bidders competing in an auction
The theory of rational choice

The theory of rational choice is a component of many

models in game theory.
this theory is that a decision-maker chooses the best
action according to her preferences, among all the
actions available to her.
No qualitative restriction is placed on the decision-maker's
preferences; her "rationality" lies in the consis- tency of her
decisions when faced with different sets of available
actions, not in the nature of her likes and dislikes.
Definition of Game Theory

Game theory is a study of how to mathematically

determine the best strategy for given conditions in
order to optimize the outcome

It is a branch of mathematics concerned with the

analysis of strategies for dealing with competitive
situations where the outcome of a participant's
choice of action depends critically on the actions
of other participants
Definition of Game Theory

The theory is based on a model with two components:

a set A consisting of all the actions that, under some
circumstances, are available to the decision-maker, and
a specification of the decision-maker's preferences.
In any given situation the decision-maker is faced with a
subset of A, from which she must choose a single element.
The decision-maker knows this subset of available choices,
and takes it as given; in particular, the subset is not
influenced by the decision-maker ́s preferences.
Example

(Payoff function representing preferences) A person is faced

with the choice of three vacation packages, to Havana, Paris,
and Venice.
She prefers the package to Havana to the other two, which she
regards as equivalent.
Her preferences between the three packages are represented
by any payoff function that assigns the same number to both
Paris and Venice and a higher number to Havana.
For example, we can set u(Havana) = 1 and u(Paris) u (Venice)
0, or u(Havana) = 10 and u(Paris) = u(Venice) = 1, or u(Havana)
= 0 and u(Paris) u(Venice) = −2.
Key terms in Game Theory
Game: Any set of circumstances that has a result
dependent on the actions of two or more
decision-makers (players)
Players: A strategic decision-maker within the context of
the game
Strategy: A complete plan of action a player will take
given the set of circumstances that might arise within the
game
Payoff: The payout a player receives from arriving at a
particular outcome
Information set: The information available at a given point
in the game
Equilibrium: The point in a game where both players have
made their decisions and an outcome is reached
Types of Games

Perfect vs. Imperfect information

Simultaneous vs. Sequential moves

Zero-sum vs. Non-zero sum

Cooperative vs. Conflict (Non-cooperative)

Perfect vs. Imperfect information

Games with Perfect information –

All information known to every player
No mystery involved
Players not allowed to hide any information
Eg.: Chess, Tic-tac-toe, Checkers

Games with Imperfect information –

Information hidden from players
Mystery involved / gamble and bluff
With available information, players have to evolve
strategy to win
Eg.: Poker / Rummy, 20 questions, Monopoly
Game Theory - Modeling
Game theory consists of a collection of models
Models of game theory are precise expressions of
ideas presented verbally (not precise)
Mathematical modelling is precise and concise
Game theoretic model development – Based on
IDEA related to interaction of decision-makers
Features of the situation that are relevant
Identifying the features is an art
Too many features increases complexity
Too few features cause loss of useful insight into the
situation
Analyze the model
Theory of Rational Choice
Rationality and Why?

Assumptions:
Humans are rational beings
Humans always seek the best alternative in a set
of possible choices

Why assume rationality?

narrow down the range of possibilities

predictability
 Theory of Rational choice
Definition:
Decision-maker chooses the best action according to
his/her preferences, among all the actions available to
him/her
Consistency in preferences for different sets of actions (A, B,
C)
A preferred over B ; B preferred over C ; then A
preferred over C
Preferences represented by “payoff function”
Payoff function / utility function (u) – number associated
with each action – higher the number more the preference
u(a) > u(b) if and only if a is preferred over b
Types of game and Equilibrium

Order Strategic
Extensive game
game
Information
Subgame
Nash
perfect perfect
Equilibrium
equilibrium
Perfect
Bayesian
imperfect Bayesian
equilibrium
equilibrium
Strategic games - three examples

1) Prisoner’s Dilemma

2) Bash or Stravinsky (BoS)

3) Matching Pennies
Prisoner’s Dilemma
Pair of criminals have been
caught
District attorney has
evidence to convict them of
a minor crime
Not enough evidence to
convict for the major crime
Offers both a deal:
• If both blame/confess, then
each one gets 10 years term
• If one confesses and other
doesn’t, the one who
confesses would get a set
free and the other would
get a longer term (20 years)
• If both don’t confess, then
they would be convicted for
minor crime (1 year)
 Prisoner’s Dilemma

Prisoner 1 (Pay-off function - u1)

u1(Blame, Don’t) > u1(Don’t, Don’t) > u1(Blame, Blame) > u1(Don’t, Blame)

Prisoner 2 (Pay-off function – u2)

u2(Don’t, Blame) > u2(Don’t, Don’t) > u2(Blame, Blame) > u2(Blame, Don’t)
Examples of Prisoner’s Dilemma
Other examples of
Prisoner’s Dilemma are:

Duopoly (Pricing of a
product by 2 firms)

Arms race (Building of

nuclear bombs by 2
countries)

Joint project (work

hard or goof off on
project by 2
employees)
 Bach or Stravinsky (BoS)
Two concerts – one music by Bach and one music by Stravinsky

Two friends - Both players prefer different music

If each one goes to different concert – both friends are unhappy

Also called as “Battle of the Sexes”

Eg: Two merging firms using different computer technologies

 Matching Pennies
Two players – toss a penny simultaneously

If same side (heads/tails) – player 2 pays player 1 one dollar

If different sides (heads & tails) – player 1 pays player 2 one dollar

Strictly competitive game

Eg: Appearance of product by new and existing firms – should it

be same or different
Stag Hunt
 Nash Equilibrium
Definition:
Nash equilibrium: Any combination of strategies in which
each player’s strategy is his/her best choice, given the other
player’s choices. Deviating from NE, would result in the
same/lower payout for that person.
Nash Equilibrium is an outcome reached that, once
achieved, means no player can increase payoff by
changing decisions unilaterally.
It can also be thought of as "no regrets" in the sense that
once a decision is made, the player will have no regrets
regarding the decision, considering the consequences.
NE - Prisoner’s Dilemma

Prisoner 2 chooses Blame/Confess

Prisoner 1 - u1(Blame) > u1(Quiet)

Prisoner 1chooses Blame/Confess

Prisoner 2 – u2(Blame) > u2(Quiet)
 NE - Bach or Stravinsky (BoS)
(Bach, Bach) – friend 1 switches to Stravinsky, her/his payoff
decreases from 2 to 0 – friend 2 switches to Stravinsky, her/his payoff
decreases from 1 to 0 – is a NE

(Bach, Stravinsky) - friend 1 switches to Stravinsky, her/his payoff

increases from 0 to 1 – Not a NE

(Stravinsky, Bach) - friend 1 switches to Bach, her/his payoff increases

from 0 to 2 - Not a NE

(Stravinsky, Stravinsky) - friend 1 switches to Bach, her/his payoff

decreases from 1 to 0 – friend 2 switches to Bach, her/his payoff
decreases from 2 to 0 - is a NE
 NE - Matching Pennies
For the pairs of actions (Head, Head) and (Tail, Tail), player 2 is
better off deviating

For the pairs of actions (Head, Tail) and (Tail, Head), player 1 is
better off deviating.

No NE
 NE - Stag Hunt
For the pairs of actions (Stag, Stag) and (Hare, Hare) –
decrease in payoff for both players – is the NE

For the pairs of actions (Stag, Hare) and (Hare, Stag),

payoffs increase for both players – Not the NE
Nash Equilibrium:

Illustrations
Cournot's model of oligopoly
Bertrand's model of oligopoly
Electoral competition
The War of Attrition
Auctions
Nash Equilibrium:

a few key models that use the notion of Nash

equilibrium
economic, political, and biological phenomena.
The discussion shows how the notion of Nash
equilibrium improves our understanding of a wide
variety of phenomena.
It also illustrates some of the many forms strategic
games and their Nash equilibria can take.
The Cournot Model of
oligopoly
It is based on a mathematical model that explains how
firms in an oligopolistic market will compete.
The model predicts the behavior of a group of firms
that are competing with each other.
"oligopoly" (competition between a small number of
sellers), though they involve no restriction on the
number of firms; the label reflects the strategic
interaction they capture.
It is an oligopoly, meaning that there are only a few
sellers of a particular good or service in the market.
The Cournot Model of
oligopoly
It is based on a mathematical model that explains how
firms in an oligopolistic market will compete.
The model predicts the behavior of a group of firms
that are competing with each other.
"oligopoly" (competition between a small number of
sellers), though they involve no restriction on the
number of firms; the label reflects the strategic
interaction they capture.
It is an oligopoly, meaning that there are only a few
sellers of a particular good or service in the market.
The Cournot Model of
oligopoly
A single good is produced by n firms.
The cost to firm i of producing qi units of the
good is Ci(qi), where Ci is an increasing function
(more output is more costly to produce).
All the output is sold at a single price, determined
by the demand for the good and the firms' total
output. Specifically:
if the firms' total output is Q then the market
price is P(Q); P is called the "inverse demand
function".
The Cournot Model of
oligopoly
Assume that P is a decreasing function when it is
positive: if the firms' total output increases, then
the price decreases (unless it is already zero).
Players The firms.Actions Each firm's set of actions
is the set of its possible outputs (nonnegative
numbers).Preferences Each firm's preferences
are represented by its profit,
The Cournot Model of
oligopoly
A single good is produced by n firms.
The cost to firm i of producing qi units of the
good is Ci(qi), where Ci is an increasing function
(more output is more costly to produce).
All the output is sold at a single price, determined
by the demand for the good and the firms' total
output. Specifically:
if the firms' total output is Q then the market
price is P(Q); P is called the "inverse demand
function".