ECO 220 Game Theory Simultaneous Move Games Objectives Be able to structure a game in normal form Be able to identify a Nash equilibrium Agenda Definitions Equilibrium Concepts Dominance Coordination Games Applications
Definitions Simultaneous Move Games Games in which the players move within the same information set without knowing what the other players have done Occasionally this means moving at the same time (Rock-Paper-Scissors) or may mean moving at different times Games of imperfect information are played as simultaneous move games Definitions Discrete vs Continuous Strategies In games with discrete strategies there is a finite number of distinct choices that the players can make (enter or not enter) In continuous strategy games the actions that the players can take actions along a continuum (price of a product could be $1.00, $1.01, $1.02 and so on Definitions Pure Strategies vs Mixed Strategies In pure strategy games the player does one action or the other with certainty In mixed strategy games the players randomly choose their actions (such as in Rock-Paper- Scissors) We will focus initially on pure strategy games with discrete choices
Definitions Normal (or strategic) form of the game is a matrix representing the actions and payoffs to the players The game tree that we looked at in the sequential games is what is known as the extensive form A simple game We will use the game Rock-Paper- Scissors to illustrate these concepts Bart and Lisa have to decide how to spend $50 that they got from their grandfather each knows what he or she wants to do If they do what Lisa wants, her payoff is +10 and Bart s is 2 If they do what Bart wants, his payoff is 8 and Lisa s is -3 A Simple Game Bart and Lisa cannot agree on a course of action so they decide to play Rock-Paper- Scissors. Whoever wins RPS chooses what to do with the $50 The game in normal form is shown on the next slide
A Simple Game Lisa Bart Paper Scissors Rock Paper Scissors Rock (0,0) (2,10) (8,-3) (8,-3) (2,10) (0,0) (8,-3) (2,10) (0,0) Bart s Payoff Lisa s Payoff Actions and Strategies In simultaneous move games actions are the same as strategies In the sequential games players had to enumerate all of their choices that they could make once they knew what the other had done In simultaneous move games they move together so there is only one thing to do Equilibrium Concepts The most frequently referred to equilibrium concept is that of a Nash Equilibrium There are other special cases that we will look at but they are all Nash equilibriums An outcome is a Nash equilibrium if no player has any incentive to deviate from his or her course of action given what the other players are doing
Finding An Equilibrium There are several ways to find Nash Equilibriums in pure strategies, when they exist The most basic is a cell-by-cell analysis The best-response analysis in the textbook is really a refinement of this Let s see if there is a Nash Equilibrium in the RPS game played by Lisa and Bart A Simple Game There is no Nash equilibrium in this game Bart Lisa Rock Paper Scissors Rock (0,0) (2,10) (8,-3) Paper (8,-3) (0,0) (2,10) Scissors (2,10) (8,-3) (0,0) Bart has no incentive to deviate but Lisa does. NOT NASH At (Rock, Rock) both players have an incentive to choose differently. Therefore, NOT NASH Lisa has no incentive to deviate but Bart does. NOT NASH Finding Equilibrium In the cell-by-cell analysis we go through each cell and see if one or more players would be better off choosing a different action Best Response Analysis is similar In each row, we find Lisa s best response to Bart s choice of that row. We do the same with each column for Bart
Finding Equilibrium Lisa Bart Paper Scissors Rock Paper Scissors Rock (0,0) (2,10) (8,-3) (8,-3) (2,10) (0,0) (8,-3) (2,10) (0,0) Bart s best response to Rock played by Lisa is to play paper Lisa s best response to Paper played by Bart Best Response Analysis In the previous example, there were no cells where that cell was the best response for both players. There is no Nash equilibrium in this game. Application 1 Two players have to choose among four actions each. Row player chooses R 1, R 2, R 3, or R 4 Column player chooses C 1, C 2, C 3, or C 4 The payoffs are shown in the following matrix Use cell by cell comparison as well as best response analysis to find the equilibrium
Application 1 R 1 (2, 4) (2, 9) (1, 7) (0, 12) R 2 (5, 2) (5, 3) (2, 0) (3, 1) R 3 (6, 0) (7, 6) (0, 2) (1, 3) R 4 (1, 16) (4, 8) (4, 10) (4, 6) Dominance Another solution concept that can sometimes isolate a unique equilibrium and at others merely simplify the game for us involves dominance A strategy (action) is dominant if it provides a better outcome regardless of what the other player does The Prisoners Dilemma illustrates this well The Prisoners Dilemma The Dominant Strategy Equilibrium Rocko Rat Quiet Moose Rat Quiet Moose is always better off ratting too so we know that Quiet is strictly dominated and he would never play it. (-10,-10) (0,-15) (-15,0) (-2,-2) For Rocko, regardless of what Moose does, he is better off ratting. Therefore, keeping quiet is strictly dominated and we know he would never play it.
Dominance A dominant strategy equilibrium is always a Nash equilibrium (but not all Nash equilibriums are dominant strategy equilibriums) Eliminating dominated strategies can greatly simplify the game solving process The next game has a quick solution Dominance R 1 (2, 4) (2, 9) (1, 7) (0, 2) R 2 (5, 2) (5, 3) (2, 0) (3, 1) R 3 (6, 0) (7, 6) (10, 2) (7, 3) R 4 (1, 2) (4, 8) (4, 6) (4, 6) Dominance Much more frequently we won t be so fortunate to jump right to the solution in one go In many cases one player will have a dominant strategy but the other will not It seems plausible that a player will recognize that his/her opponent has dominated strategies and will not play them This can allow us to apply iterative elimination of dominated strategies
Iterated Elimination Row does not have a dominant strategy R 1 (12, 4) (2, 9) (1, 7) (0, 2) R 2 (5, 2) (5, 3) (12, 0) (3, 1) R 3 (6, 0) (7, 6) (10, 2) (7, 3) R 4 (1, 2) (4, 8) (4, 6) (14, 6) Iterated Elimination If we look at the previous slide we can see that Row does not have a dominant strategy However, if we look at Column s payoffs, we see that column does have a dominant strategy (so the others are dominated) If we eliminate the dominated strategies for Column, we can find a dominant strategy for Row Iterated Elimination R 1 (12, 4) (2, 9) (1, 7) (0, 2) R 2 (5, 2) (5, 3) (12, 0) (3, 1) R 3 (6, 0) (7, 6) (10, 2) (7, 3) R 4 (1, 2) (4, 8) (4, 6) (14, 6)
Iterated Dominance Application In the next game, there is a dominant strategy equilibrium that requires a few iterations to eliminate all the dominated strategies Find the equilibrium in this game Does it matter whether you start by eliminating Row or Column s strategies first? Iterated Elimination R 1 (12, 4) (2, 9) (1, 7) (0, 2) R 2 (5, 2) (5, 3) (12, 0) (3, 1) R 3 (6, 0) (7, 6) (10, 2) (7, 3) R 4 (1, 2) (4, 8) (4, 6) (14, 6) Iterated Elimination R 1 R 2 R 3 R 4