Outline for
a course on Game Theory.
Instructor: Mike O’Neill, CMC Math department.
References: 1. Fun and Games by Ken Binmore denoted by [B].
2. UCLA course notes by Tom Ferguson denoted by [F].
1. History and background. Applications of game theory to economics, political
Science, biology and number theory. ([B] and some additional sources)
Definitions and vocabulary. Games in extensive form, plays, game trees, information sets, strategies, chance moves, payoffs. ([B], [F])
Example : Game of Pure Strategy (GOPS). The above notions are demonstrated using this simple card game
(1)
2. Two person zero-sum games of complete information.
Examples: Tic-tac-toe. NIM.
Backward induction, Zermelo’s Algorithm. [B]
NIM sum theorem. [F]
(1)
comp. inf. game.) Saddle points. Nash equilibria.
Sub game perfect equilibria [B]
Pure Strategy equilibria: Cournot Competition, Hotelling Competition, Majority
Voting.
(2)
4. Critiques of backward induction and sub-game perfect equilibria.
N-person games, equilibrium n-tuples, Existence theorem for
equilibrium n- tuples in finite games of complete information. This is
another inductive proof. Observe the difference between this theorem and
Nash’s theorem.
Non-existence of pure strategy equilibria. Matching Pennies.
Inspection game.
Team games, multiple equilibria, [B].
(2)
Probability review. Random variable, conditional probability, expectation.
Lotteries, Compound Lotteries.
Game values.
Analysis:
An extensive form for “Duel”.
A simplified form of Parcheesi.
(2)
Chapter 3 of [B] is required reading before this class.
Definitions: ordinal utility, preference relations, utility axioms
Theorems: equivalent lottery theorem.
Existence of a utility function for a given preference relation.
Why do we need to be clear about the notion of utility?
From [B], Russian roulette, ineffectiveness of Zermelo’s algorithm in games of
incomplete information.
Risk aversion, Utility scales and temperature scales, Allais’ Paradox, Bayesian
rationality. [B]
(2)
Payoff functions, Bi-matrix games, Nash equilibria in terms of payoff functions.
A strategic form for Duel.
A strategic form for Russian roulette.
Successive deletion of dominated strategies.
Common Knowledge and problems with successive deletion.
Examples: Prisoner’s dilemma and 2nd Price Auctions
Iterated deletion in the Cournot Model.
Review of Matrix algebra, closing with a proof of the
Separating hyper plane theorem.
(2)
Convexity, supporting lines. Concave, convex and affine functions.
Bargaining: binding agreements, cooperative pay off regions, free disposal.
Bargaining set, Pareto efficiency (optimality). Nash bargaining
solutions.
Finding Nash bargaining solutions. The Nash axioms.
The Nash bargaining theorem.
Example:“Dividing the dollar”. Risk aversion and love of risk.
Bargaining models: Ultimatum game, with two stages, infinite horizon
Game (Rubinstein’s model).
Theorem: Rubinstein’s solutions approximate the Nash solution.
Theorem: (Rubinstein) Infinite horizon has a unique sub game-perfect
equilibrium outcome.
Example: Stationary War of Attrition
(3)
Matrix games (zero-sum or constant sum)
Mixed strategies
Min-Max theorem: Proof by separating hyper planes theorem.
Proof by reduction to Linear programming problem. [F]
Solution of matrix games by linear programming. [F]
Solution by fictitious play.
Battleships, The inspection game. [B]
Rationalizability and Iterated Strict Dominance.
Theorem: they coincide in 2 player games.
Proof: by separating hyper plane theorem.
(3)
Reaction curves (with pure and mixed strategies)
Cournot models, monopoly, duopoly…
Equilibrium selection, Pareto dominance, stability.
Nash demand game, pre-play negotiation, threats
Prisoner’s dilemma, Collusion in Cournot Duopoly.
Theorems:
Fixed point theorems of Kakutani and Brouwer.
Nash’s theorem on existence of equilibria.[B] and [O].
The game of Hex and Brouwer’s theorem.[B] (a proof of Brouwer’s theorem)
Existence of equilibrium pairs of threat strategies. [O] (Another fixed point
proof.)
Example: Fictitious play cannot be used to solve bi-matrix games.
(3)