Now we return to simultaneous-move games. We resolve the issue of non-existence of Nash equilibrium. in pure strategies through intentional mixing.

Similar documents
Sequential-move games with Nature s moves.

CMPSCI 240: Reasoning about Uncertainty

6.254 : Game Theory with Engineering Applications Lecture 3: Strategic Form Games - Solution Concepts

Introduction to Multi-Agent Programming

Game Theory: Minimax, Maximin, and Iterated Removal Naima Hammoud

UC Berkeley Haas School of Business Game Theory (EMBA 296 & EWMBA 211) Summer 2016

Elements of Economic Analysis II Lecture X: Introduction to Game Theory

Player 2 H T T -1,1 1, -1

G5212: Game Theory. Mark Dean. Spring 2017

Game Theory - Lecture #8

Game Theory. Analyzing Games: From Optimality to Equilibrium. Manar Mohaisen Department of EEC Engineering

Introduction to Game Theory

The Ohio State University Department of Economics Econ 601 Prof. James Peck Extra Practice Problems Answers (for final)

10.1 Elimination of strictly dominated strategies

MS&E 246: Lecture 2 The basics. Ramesh Johari January 16, 2007

CS711 Game Theory and Mechanism Design

University of Hong Kong

Games with incomplete information about players. be symmetric or asymmetric.

Economics 171: Final Exam

Game Theory: Global Games. Christoph Schottmüller

Rationalizable Strategies

S 2,2-1, x c C x r, 1 0,0

Best-Reply Sets. Jonathan Weinstein Washington University in St. Louis. This version: May 2015

Economics 109 Practice Problems 1, Vincent Crawford, Spring 2002

Mixed Strategies. In the previous chapters we restricted players to using pure strategies and we

SI 563 Homework 3 Oct 5, Determine the set of rationalizable strategies for each of the following games. a) X Y X Y Z

Mixed Strategy Nash Equilibrium. player 2

CS 331: Artificial Intelligence Game Theory I. Prisoner s Dilemma

Game Theory: Additional Exercises

Math 135: Answers to Practice Problems

B w x y z a 4,4 3,3 5,1 2,2 b 3,6 2,5 6,-3 1,4 A c -2,0 2,-1 0,0 2,1 d 1,4 1,2 1,1 3,5

The Ohio State University Department of Economics Second Midterm Examination Answers

CS 798: Homework Assignment 4 (Game Theory)

Notes for Section: Week 7

January 26,

CUR 412: Game Theory and its Applications Final Exam Ronaldo Carpio Jan. 13, 2015

Basic Game-Theoretic Concepts. Game in strategic form has following elements. Player set N. (Pure) strategy set for player i, S i.

Chapter 10: Mixed strategies Nash equilibria, reaction curves and the equality of payoffs theorem

Elements of Economic Analysis II Lecture XI: Oligopoly: Cournot and Bertrand Competition

Advanced Microeconomics

ANASH EQUILIBRIUM of a strategic game is an action profile in which every. Strategy Equilibrium

Test 1. ECON3161, Game Theory. Tuesday, September 25 th

Name. Answers Discussion Final Exam, Econ 171, March, 2012

Problem 3,a. ds 1 (s 2 ) ds 2 < 0. = (1+t)

Extensive-Form Games with Imperfect Information

Game Theory. VK Room: M1.30 Last updated: October 22, 2012.

Introduction to Game Theory

Microeconomics II. CIDE, MsC Economics. List of Problems

Prisoner s Dilemma. CS 331: Artificial Intelligence Game Theory I. Prisoner s Dilemma. Prisoner s Dilemma. Prisoner s Dilemma.

that internalizes the constraint by solving to remove the y variable. 1. Using the substitution method, determine the utility function U( x)

Iterated Dominance and Nash Equilibrium

Problem Set 2 - SOLUTIONS

Auction is a commonly used way of allocating indivisible

TTIC An Introduction to the Theory of Machine Learning. Learning and Game Theory. Avrim Blum 5/7/18, 5/9/18

Game Theory. Important Instructions

Simon Fraser University Fall Econ 302 D200 Final Exam Solution Instructor: Songzi Du Wednesday December 16, 2015, 8:30 11:30 AM

Francesco Nava Microeconomic Principles II EC202 Lent Term 2010

ECON 803: MICROECONOMIC THEORY II Arthur J. Robson Fall 2016 Assignment 9 (due in class on November 22)

HW Consider the following game:

Game theory and applications: Lecture 1

6.207/14.15: Networks Lecture 9: Introduction to Game Theory 1

Simon Fraser University Spring 2014

Continuing game theory: mixed strategy equilibrium (Ch ), optimality (6.9), start on extensive form games (6.10, Sec. C)!

PAULI MURTO, ANDREY ZHUKOV

6.207/14.15: Networks Lecture 10: Introduction to Game Theory 2

ECE 586GT: Problem Set 1: Problems and Solutions Analysis of static games

PROBLEM SET 6 ANSWERS

Corporate Control. Itay Goldstein. Wharton School, University of Pennsylvania

In the Name of God. Sharif University of Technology. Microeconomics 2. Graduate School of Management and Economics. Dr. S.

Markets with Intermediaries

Markets with Intermediaries

Repeated Games with Perfect Monitoring

Mixed Strategies. Samuel Alizon and Daniel Cownden February 4, 2009

Math 152: Applicable Mathematics and Computing

Expectations & Randomization Normal Form Games Dominance Iterated Dominance. Normal Form Games & Dominance

Evolution & Learning in Games

Bayesian Nash Equilibrium

Finding Mixed-strategy Nash Equilibria in 2 2 Games ÙÛ

MA200.2 Game Theory II, LSE

Today. Applications of NE and SPNE Auctions English Auction Second-Price Sealed-Bid Auction First-Price Sealed-Bid Auction

Outline for today. Stat155 Game Theory Lecture 13: General-Sum Games. General-sum games. General-sum games. Dominated pure strategies

Regret Minimization and Security Strategies

1 R. 2 l r 1 1 l2 r 2

Econ 302 Assignment 3 Solution. a 2bQ c = 0, which is the monopolist s optimal quantity; the associated price is. P (Q) = a b

CS 7180: Behavioral Modeling and Decision- making in AI

Preliminary Notions in Game Theory

March 30, Why do economists (and increasingly, engineers and computer scientists) study auctions?

Strategy -1- Strategy

Static Games and Cournot. Competition

Homework #2 Psychology 101 Spr 03 Prof Colin Camerer

CHAPTER 9 Nash Equilibrium 1-1

Solution Problem Set 2

In the Name of God. Sharif University of Technology. Graduate School of Management and Economics

Not 0,4 2,1. i. Show there is a perfect Bayesian equilibrium where player A chooses to play, player A chooses L, and player B chooses L.

Ph.D. Preliminary Examination MICROECONOMIC THEORY Applied Economics Graduate Program June 2015

Version 01 as of May 12, 2014 Author: Steven Sagona To be submitted to PRL (not really)

Game Theory with Applications to Finance and Marketing, I

Noncooperative Oligopoly

Economics 703: Microeconomics II Modelling Strategic Behavior

Econ 101A Final exam Mo 18 May, 2009.

Transcription:

Econ 221 Fall, 2018 Li, Hao UBC CHAPTER 7. SIMULTANEOUS-MOVE GAMES: MIXED STRATEGIES Now we return to simultaneous-move games. We resolve the issue of non-existence of Nash equilibrium in pure strategies through intentional mixing. Real life examples of mixing: sports, free riding, tax audit. 1

7.1 What is a mixed strategy A penalty taker with a powerful left-sided shot. Keeper Left Right Kicker Left 1, 0 0.4, 0.6 Right 0, 1 1, 0 This is a zero-sum game, but unlike Matching Pennies, it is asymmetric. 2

There is no Nash equilibrium in pure strategies. Each strategy of each player is rationalizable. The only way to win, or equivalently not to lose, is to keep the opponent guessing by mixing between your own pure strategies. Mixing is therefore making your own choice of strategy intentionally random from perspective of opponent. 3

A mixed strategy of a player assigns a probability of playing each pure strategy, such that sum of assigned probabilities is equal to 1 over all the strategies of the player. Mixing doesn t mean assigning same probability to each pure strategy. With only two pure strategies, a mixed strategy can be represented by one number between 0 and 1. A pure strategy is a degenerate mixed strategy, with 1 assigned to the pure strategy and 0 to all other pure strategies. 4

Expected payoff in the presence of mixing. Expected payoff to a player is probability-weighted sum of the player s payoffs, with probabilities determined by corresponding mixing. Mixing is independent across players. Example: expected payoff to Kicker from using mixed strategy of Left with probability p and Right with 1 p against Keeper s mixed strategy of Left with probability q and Right with probability 1 q is given by following: pq 1+ p(1 q) 0.4+(1 p)q 0+(1 p)(1 q) 1. 5

7.2 Mixing moves Benefit of mixing. Suppose that Keeper plays a mixed strategy of Left with probability q and Right with probability 1 q. Expected payoff to Kicker is q 1+(1 q) 0.4 from Left, q 0+(1 q) 1 from Right, and p times the first payoff plus (1 p) times the second from a mixed strategy of Left with probability p and Right with probability 1 p. 6

Kicker s best response is p = 1 if q > 3/8 and p = 0 if q < 3/8. If q = 3/8, Kicker is indifferent between Left and Right, and expected payoff from best response is minimized. Benefit of mixing to Keeper: since this is a zero-sum game, Keeper should mix between Left and Right with exactly q = 3/8, and thus increases guaranteed payoff from 0 to 3/8. 7

Payoff 1 5/8 Kicker s expected payoff from Left Kicker s expected payoff from Right 0 3/8 1 q Figure 1. Benefit of mixing to Keeper. 8

Kicker should mix with p of Left and 1 p of Right in such a way that Keeper is indifferent between Left and Right. Equating expected payoff of Keeper p 0+(1 p) 1 from Left to expected payoff p 0.6+(1 p) 0 from Right gives p = 5/8. Benefit of mixing to Kicker: by choosing p = 5/8, Kicker increases guaranteed payoff from 0.4 (from choosing p = 1) to 5/8. 9

Nash equilibrium in mixed strategies. With only two pure strategies, a mixed strategy can be represented by one number between 0 and 1, so mixed strategies are special case of continuous strategies. Nash equilibrium can be found by intersection of best response functions. 10

p 1 Kicker s best response function 5/8 Nash equilibrium Keeper s best response function 0 3/8 1 q Figure 2. Nash equilibrium in Penalty Kick. 11

7.3 Nash equilibrium as a system of beliefs and responses In a Nash equilibrium, each player plays a best response against the equilibrium strategies of other players. Recall two features of Nash equilibrium: non-cooperative and correct beliefs. Nash equilibrium in mixed strategies requires players to have correct beliefs in that they know equilibrium mixtures of all players. 12

7.4 Mixing in non-zero-sum games Free-rider problem in crime-reporting. Setup: two witnesses of crime decide whether to report it or not; reporting it costs c individually; each witness receives b > c if at least one of them reports it. Game table. Witness 2 Report Don t Witness 1 Report b c, b c b c, b Don t b, b c 0, 0 13

This is the same as Hawk-Dove, with two Nash equilibria in pure strategies. Witness 2 Report Don t Witness 1 Report b c, b c *b c, b* Don t *b, b c* 0, 0 There is a mixed strategy equilibrium. 14

Principle of making your opponent indifferent: your opponent is willing to mix between two pure strategies only when you mix your own two pure strategies in a way such that your opponent is indifferent. Apply the principle: if Witness 2 chooses Report with probability q (and Don t with probability 1 q), then expected payoff of Witness 1 is(b c) from Report, and qb from Don t, so q = (b c)/b; symmetrically, Witness 1 has to choose Report with probability p = (b c)/b to make Witness 2 indifferent. 15

p 1 Witness 1 s best response function (b c)/b Witness 2 s best response function 0 (b c)/b 1 q Figure 3. Three Nash equilibria in Reporting a Crime. 16

7.5 General discussion of mixed-strategy equilibria Weak sense of equilibrium. Nash equilibrium in pure strategies is typically strict in that each player plays the unique best response to equilibrium strategies of opponents. Principle of making your opponent indifferent means that each player s equilibrium mix is not a strict best response to opponent s equilibrium mix. 17

Counterintuitive comparative statics. Penalty Kick again: a smaller α means an improvement in Keeper s skill in reducing Kicker s advantage. Keeper Left Right Kicker Left 1, 0 α, 1 α Right 0, 1 1, 0 Mixed-strategy Nash equilibrium depends on α, and can be found by applying the principle of making your opponent indifferent. 18

To make Kicker indifferent, Keeper chooses Left with q such that q 1+(1 q) α = q 0+(1 q) 1, which gives q = (1 α)/(2 α). To make Keeper indifferent, Kicker chooses Left with p such that p 0+(1 p) 1 = p (1 α)+(1 p) 0, which gives p = 1/(2 α). When α decreases, Keeper gets better with Right but in equilibrium achieves a higher payoff by using Right less often. 19

7.6 Mixing when one player has three or more pure strategies Penalty Kick with Panenka. Keeper Left Right Left 1, 0 0.4, 0.6 Kicker Middle 0.6, 0.4 0.8, 0.2 Right 0, 1 1, 0 Keeper still has only two pure strategies so we continue to represent any mixed strategy with probability p that Keeper chooses Left. 20

Payoff 1 Kicker s expected payoff from Left 5/8 Kicker s expected payoff from Middle Kicker s expected payoff from Right 0 3/8 1 q Figure 4. Penalty Kick with Panenka. 21

To minimize Kicker s payoff from best response, Keeper has to keep Kicker indifferent between Left and Middle: equating q 1+(1 q) 0.4 to q 0.6+(1 q) 0.8 gives Keeper s equilibrium mix q = 0.5. Kicker does not use Right in equilibrium. In order for Keeper to mix, Kicker needs to choose Left with probability p l and Middle with probability 1 p l to make Keeper indifferent: equating p l 0+(1 p l ) 0.4 to p l 0.6+(1 p l ) 0.2 gives Kicker s equilibrium mix p l = 0.25. 22

Modified principle of making your opponent indifferent when at least one player has more than two pure strategies: in any mixed-strategy equilibrium, a player has to be indifferent among all pure strategies used in equilibrium, and prefers any of them to the ones unused in equilibrium. When only one player has three or more strategies in a zero-sum game, only two will be used in equilibrium, and which two is determined by the opponent mixing to minimize the former s best payoff. 23

Payoff 1 5/8 Kicker s expected payoff from Left Kicker s expected payoff from Middle Kicker s expected payoff from Right 0 3/8 1 q Figure 5. Penalty Kick: Panenka is used with Right. 24

Payoff 1 5/8 Kicker s expected payoff from Left Kicker s expected payoff from Middle Kicker s expected payoff from Right 0 3/8 1 q Figure 6. Penalty Kick: Panenka is never used. 25

7.6 Mixing when both players have three or more strategies Only strategies that survive iterated elimination of never best responses can be used in equilibrium So we apply modified principle of making your opponent indifferent to reduced game. Again, some pure strategies may not be used in equilibrium, but mixing among three or even more pure strategies can occur in equilibrium. Verifying equilibrium is easier than finding equilibrium. 26

3-by-3 Penalty Kick. Keeper Left Middle Right Left 1, 0 1, 0 0, 1 Kicker Middle 1, 0 0,1 1, 0 Right 0, 1 1, 0 1, 0 Kicker and Keeper both choosing each strategy with equal probability is a Nash equilibrium. 27