Principle of Microeconomics: Recitation Week 13

Game Theory and Strategic Play

洪挺智
December 4, 2024
NTU Economics
Contents

1 Simultaneous-Move Games

2 Dominant Strategy Equilibrium

3 Nash Equilibrium

4 Extensive-Form Games

5 Summary

1/51
Simultaneous-Move Games
Simultaneous-Move Games

Introduction
Game Theory is the Study of Strategic Interactions

• Previously, your utility depends only on your own choices.

• However, in many situations, other agents’ decisions affect your payoff.
• Game theory provides tools to analyze these strategic interactions.

2/51
Simultaneous-Move Games: Definition

• In a simultaneous-move game, players make decisions at the same time.

• It requires players must take their actions without observing others’ actions.
• Also, players cannot coordinate their actions.
• Conditional on players’ actions, each player receives a payoff.

3/51
Simultaneous-Move Games

Key Elements of a Game

Key Elements of a Game: Players, Strategies, and Payoffs

Players Decision makers in the game.

Strategies A complete plan describing how a player will act.
Payoffs the outcomes that result from the players’ actions.

4/51
Example: The Prisoner’s Dilemma

The Prisoner’s Dilemma

Two suspects are arrested and questioned separately. Each suspect is told that
if both confess, they will each receive a 5-year sentence. If neither confesses,
they will each receive a 2-year sentence. If one confesses and the other does
not, the confessor will be released and the other will receive a 10-year
sentence.

5/51
Example: The Prisoner’s Dilemma (cont.)

Players Two suspects.

Strategies 𝑆𝑖 = {Confess, Silent} for each suspect 𝑖.
Payoffs Let 𝑢𝑖 (𝑠1 , 𝑠2 ) be the payoff of suspect 𝑖 when they choose 𝑠1 and 𝑠2 .

𝑢1 (Confess, Confess) = −5, 𝑢2 (Confess, Confess) = −5,

𝑢1 (Confess, Silent) = 0, 𝑢2 (Confess, Silent) = −10,
𝑢1 (Silent, Confess) = −10, 𝑢2 (Silent, Confess) = 0,
𝑢1 (Silent, Silent) = −2, 𝑢2 (Silent, Silent) = −2.

6/51
Example: The Battle the Sexes

The Battle of the Sexes

A couple is trying to decide what to do on a Saturday night. Player 2 prefers to
go to a football game, while the Player 1 prefers to go to a opera. If both go to
the football game, Player 2 receives a payoff of 2 and the Player 1 receives a
payoff of 1. If both go to the opera, Player 2 receives a payoff of 1 and the Player
1 receives a payoff of 2. If they go to different events, both receive a payoff of 0.

7/51
Example: The Battle the Sexes (cont.)

Players Player 1 and the Player 2.

Strategies 𝑆𝑖 = {𝐹, 𝑂} for each player 𝑖.
Payoffs Let 𝑢𝑖 (𝑠1 , 𝑠2 ) be the payoff of player 𝑖 when they choose 𝑠1 and 𝑠2 .

𝑢1 (𝐹, 𝐹) = 1, 𝑢2 (𝐹, 𝐹) = 2,
𝑢1 (𝐹, 𝑂) = 0, 𝑢2 (𝐹, 𝑂) = 0,
𝑢1 (𝑂, 𝐹) = 0, 𝑢2 (𝑂, 𝐹) = 0,
𝑢1 (𝑂, 𝑂) = 2, 𝑢2 (𝑂, 𝑂) = 1.

8/51
Simultaneous-Move Games

Matrix Representation of a Game

Matrix Representation of a Game

• Any two-player finite game can be represented by a payoff matrix.

• It capture all the relevant information of the simultaneous-move game.

Rows Each row represents one of player 1’s strategies.

If player 1 has 𝑘 strategies, there are 𝑘 rows.
Columns Each column represents one of player 2’s strategies.
If player 2 has 𝑚 strategies, there are 𝑚 columns.
Entries Each entry contains a vector (𝑢1 , 𝑢2 ), where 𝑢𝑖 is player 𝑖’s payoff
when the actions of both players correspond to the row and column
of that entry.

9/51
Example: The Prisoner’s Dilemma

Player 2
Confess Silent

Confess −5, −5 0, −10

Player 1
Silent −10, 0 −2, −2

10/51
Example: The Battle of the Sexes

Player 2
𝑂 𝐹

𝑂 2, 1 0, 0
Player 1
𝐹 0, 0 1, 2

11/51
Simultaneous-Move Games

Solution Concepts
Solution Concepts: Finding the “Reasonable” Outcomes

• We have described the key elements of a game.

• Now we need to find a way to predict the outcome of a game.
• We need to make assumptions about the behavior and the beliefs of players.
• A solution concept is a method of analyzing games, restricting the set of all
possible outcomes to those that are more reasonable than others.
• Put simply, a method to solve a game.
• An equilibrium is a strategy profile that emerges as a solution concept’s
prediction.

12/51
Assumptions and Setup

Players are rational They act to maximize their payoffs.

Players are intelligent They know everything about the game.
Common knowledge Players know that others are rational and intelligent.
Self-enforcement Given the strategies of other players, no player has an
incentive to deviate from their strategy in equilibrium.

13/51
Dominant Strategy Equilibrium
Dominant Strategy Equilibrium: The Simplest Solution Concept

• Let 𝑆𝑖 and 𝑆−𝑖 be the strategy sets of player 𝑖 and the players other than 𝑖.
• Let 𝑠𝑖 ∈ 𝑆𝑖 and 𝑠𝑖′ ∈ 𝑆𝑖 be two possible strategies of player 𝑖.
• We say 𝑠𝑖′ is strictly dominated by 𝑠𝑖 if for any 𝑠−𝑖 ∈ 𝑆−𝑖 , player 𝑖’s payoff from
′
choosing 𝑠𝑖 is strictly greater than that from choosing 𝑠𝑖 . That is,
′
𝑢𝑖 (𝑠𝑖 , 𝑠−𝑖 ) > 𝑢𝑖 (𝑠𝑖 , 𝑠−𝑖 ) for all 𝑠−𝑖 ∈ 𝑆−𝑖 .

• We say 𝑠𝑖 is a strictly dominant strategy for 𝑖 if every other strategy is strictly

dominated by 𝑠𝑖 .

14/51
Example: The Prisoner’s Dilemma

Player 2
Confess Silent

Confess −5, −5 0, −10

Player 1
Silent −10, 0 −2, −2

• The dominant strategy for both players is to confess.

• The dominant strategy equilibrium is (Confess, Confess).

15/51
Example: The Battle of the Sexes

Player 2
𝑂 𝐹

𝑂 2, 1 0, 0
Player 1
𝐹 0, 0 1, 2

• No dominant strategy for either player. No dominant strategy equilibrium.

16/51
Concluding Remarks on Dominant Strategy Equilibrium

• It is often the case that there is no dominant strategy equilibrium.

• How to analyze these games?

17/51
Nash Equilibrium
Nash Equilibrium

Best Response, Beliefs, and Rationalizability

Best Response and Never a Best Response

• The strategy 𝑠𝑖 ∈ 𝑆𝑖 is player 𝑖’s best response to other players’ strategies if it

maximizes player 𝑖’s payoff given the strategies of the other players; i.e.,
′ ′
𝑢𝑖 (𝑠𝑖 , 𝑠−𝑖 ) ≥ 𝑢𝑖 (𝑠𝑖 , 𝑠−𝑖 ) for all 𝑠𝑖 ∈ 𝑆𝑖 .

• Hence, a rational player who believes that her opponents are playing 𝑠−𝑖 ∈ 𝑆−𝑖
will always choose a best response to 𝑠−𝑖 .
• The best response correspondence of player 𝑖 is the set of all best responses
to the strategies of the other players 𝑠−𝑖 .
• We write BR𝑖 (𝑠−𝑖 ) for the best response correspondence of player 𝑖.
• A strategy 𝑠𝑖 is never a best response if there are no beliefs 𝑠−𝑖 for which 𝑠𝑖 is a
best response.

18/51
Example: The Prisoner’s Dilemma

Player 2
Confess Silent

Confess −5, −5 0, −10

Player 1
Silent −10, 0 −2, −2

• BR𝑖 (𝑠−𝑖 ) = {Confess} for both players.

• The strategy Silent is never a best response.

19/51
Example: The Prisoner’s Dilemma (cont.)

Player 2
Confess Silent

Confess −5, −5 0, −10

Player 1
Silent −10, 0 −2, −2

• Player 1 knows that Silent is never a best response for Player 2, and vice versa.
• The only rationalizable strategy profile is (Confess, Confess).

20/51
Example: The Battle of the Sexes

Player 2
𝑂 𝐹

𝑂 2, 1 0, 0
Player 1
𝐹 0, 0 1, 2

• BR𝑖 (𝑂) = {𝑂} and BR𝑖 (𝐹) = {𝐹} for both players.
• Rationalizability does not help us to predict the outcome.

21/51
Example: The Battle of the Sexes (cont.)

• Why rationalizability does not help us to predict the outcome?

• Player 1 may behave optimally given the belief that Player 2 will choose 𝑂.
• Player 2 may behave optimally given the belief that Player 1 will choose 𝐹.
• Then, the strategy profile is (Opera, Football).
• The players are playing best responses to their beliefs.
• But their beliefs are wrong!

22/51
Nash Equilibrium

Nash Equilibrium
Nash Equilibrium: Playing the Mutual Best Response

• A Nash equilibrium (𝑠1∗ , … , 𝑠𝑛∗ ) is a strategy profile in which each player’s

strategy is a best response to the strategies of the other players; i.e.,
∗ ∗
𝑠𝑖 ∈ BR𝑖 (𝑠−𝑖 ) for all 𝑖.

• Roughly speaking, no player has an incentive to deviate from their strategy

given the strategies of the other players.

23/51
Example: The Prisoner’s Dilemma

Player 2
Confess Silent

Confess −5, −5 0, −10

Player 1
Silent −10, 0 −2, −2

• What is the Nash equilibrium?

24/51
Example: The Prisoner’s Dilemma (cont.)

Player 2
Confess Silent

Confess −5, −5 0, −10

Player 1
Silent −10, 0 −2, −2

• The Nash equilibrium is (Confess, Confess).

25/51
Example: The Battle of the Sexes

Player 2
𝑂 𝐹

𝑂 2, 1 0, 0
Player 1
𝐹 0, 0 1, 2

• What is the Nash equilibrium?

26/51
Example: The Battle of the Sexes (cont.)

Player 2
𝑂 𝐹

𝑂 2, 1 0, 0
Player 1
𝐹 0, 0 1, 2

• There are two Nash equilibria: (𝑂, 𝑂) and (𝐹, 𝐹).

27/51
Example: The 𝑝-Beauty Contest

• Suppose there are 𝑛 players.

• Each player chooses a number between 0 and 100.
• The player whose number is closest to 𝑝 times the average of all numbers wins.
• The Nash equilibrium is that all players choose 0 if 𝑝 < 1.

28/51
Example: Duopoly Competition

• In the next week, we will study duopoly competition, including Cournot and
Bertrand competition.
• Nash equilibrium is a useful tool to analyze these games.

29/51
Example: Matching Pennies

Player 2
𝐻 𝑇

𝐻 1, −1 −1, 1
Player 1
𝑇 −1, 1 1, −2

• Two players simultaneously show a penny.

• If the pennies match, Player 1 wins a dollar from Player 2.
• If the pennies do not match, Player 2 wins a dollar from Player 1.
• What is the Nash equilibrium?

30/51
Nash Equilibrium

Mixed Strategies
Mixed Strategies: Randomizing Over Actions

• So far, we have assumed that players choose a single action.

• These are called pure strategies.
• Now, we allow players to randomize over their actions.
• Why randomize?
• Being predictable can be exploited by opponents.
• If we allow players to randomize, we can always find a Nash equilibrium.
; Applying the Kakutani fixed-point theorem, omitted here.

31/51
Example: Matching Pennies

Player 2
𝐻 𝑇

𝐻 1, −1 −1, 1
Player 1
𝑇 −1, 1 1, −1

• We have learned that this game does not have a pure strategy NE.
• Can we find a mixed strategy NE?

32/51
Example: Matching Pennies (cont.)

𝑢1
• Let 𝑝 and 𝑞 be the probabilities that Player 1 and
Player 2 chooses 𝐻, respectively.
1 𝑢1 (𝐻, 𝑞)
• Given Player 2 has a probability 𝑞 of choosing 𝐻.
• The expected payoff of Player 1 from choosing 𝐻 is

0 𝑞
𝑢1 (𝐻, 𝑞) = 𝑞 × 1 + (1 − 𝑞) × (−1) = 2𝑞 − 1.

• The expected payoff of Player 1 from choosing 𝑇 is

−1 𝑢1 (𝑇, 𝑞)
𝑢1 (𝑇, 𝑞) = 𝑞 × (−1) + (1 − 𝑞) × 1 = 1 − 2𝑞.

33/51
Example: Matching Pennies (cont.)

• We can calculate the best response of Player 1.

• Player 1 prefers 𝐻 if 2𝑞 − 1 > 1 − 2𝑞; i.e., if 𝑞 > 1 . 𝑝
2
• Player 1 prefers 𝑇 if 2𝑞 − 1 < 1 − 2𝑞; i.e., if 𝑞 < 1 .
2 1
• When 𝑞 = 1 , Player 1 is indifferent between 𝐻 and 𝑇.
2
• The best response correspondence of Player 1 is 1
2

𝑝=0 if 𝑞 < 12 , 𝑞
0
1
BR1 (𝑞) = {𝑝 ∈ [0, 1] if 𝑞 = , 1
2
1
2
1
𝑝=1 if 𝑞 > 2
.

34/51
Example: Matching Pennies (cont.)

• Applying the same logic to Player 2.

• Player 2 prefers 𝐻 if 1 − 2𝑝 > 2𝑝 − 1; i.e., if 𝑝 < 1 . 𝑝
2
• Player 2 prefers 𝑇 if 2𝑝 − 1 > 1 − 2𝑝; i.e., if 𝑝 > 1 .
2 1
• When 𝑝 = 1 , Player 2 is indifferent between 𝐻 and 𝑇.
2
• The best response correspondence of Player 2 is 1
2

𝑞=1 if 𝑝 < 12 , 𝑞
0
1
BR2 (𝑝) = {𝑞 ∈ [0, 1] if 𝑝 = , 1
2
1
2
1
𝑞=0 if 𝑝 > 2
.

35/51
Example: Matching Pennies (cont.)

• We have found the best responses curves. 1

∗
• Now, we can find the NE (𝑝 , 𝑞 ).∗
1
• By definition, 𝑝∗ ∈ BR1 (𝑞∗ ) and 𝑞∗ ∈ BR2 (𝑝∗ ). 2

• It is the intersection of the best response curves.

0 𝑞
1 1
2

36/51
Concluding Remarks on Nash Equilibrium

• Unlike dominant strategy equilibrium, Nash equilibrium is always guaranteed

to exist if we allow mixed strategies.
• Compared to rationalizability, Nash equilibrium refines the possible outcomes.
• However, it requires players to have correct beliefs, which is not always the
case.

37/51
Extensive-Form Games
Extensive-Form Games: Players Make Decisions Sequentially

• In extensive-form games, players make decisions sequentially.

• More precisely, it is not the chronological order of play that matters.
• It is what players know when they make their choices that matters!

38/51
Extensive-Form Games

Game Trees
Game Trees: A Visual Representation of Extensive-Form Games

• A game tree is a visual representation of an extensive-form game.

• It consists of nodes and branches.
• Each node represents a decision point.
• Each branch represents a possible action.

39/51
Example: The Sequential-Move Battle of the Sexes

2 ra
Ope (2, 1)

a
er Foo
• Player 1 makes the first decision. 1 Op tbal
l (0, 0)
• Player 2 makes the second decision.
• The payoffs are shown at the terminal nodes. Fo ra
ot
ba Ope (0, 0)
ll
Foo
2 tbal (1, 2)
l

40/51
Extensive-Form Games

Normal-Form Representation
Normal Form Representation of Extensive-Form Games

• Any extensive-form game can be transformed into a normal-form game

• The normal-form representation is useful for finding Nash equilibria.
• However, the dynamic nature of the game is lost in the normal-form
representation.

41/51
Example: The Sequential-Move Battle of Sexes

Player 2
𝑜𝑜 𝑜𝑓 𝑓𝑜 𝑓𝑓

𝑂 2, 1 2, 1 0, 0 0, 0
Player 1
𝐹 0, 0 1, 2 0, 0 1, 2

Strategies 𝑆1 = {𝑂, 𝐹} and 𝑆2 = {𝑜𝑜, 𝑜𝑓, 𝑓𝑜, 𝑓𝑓}, where the first and the second
letter denotes the Player 2’s action when Player 1 chooses 𝑂 and 𝐹.

• What is the Nash equilibrium?

42/51
Example: The Sequential-Move Battle of Sexes (cont.)

Player 2
𝑜𝑜 𝑜𝑓 𝑓𝑜 𝑓𝑓

𝑂 2, 1 2, 1 0, 0 0, 0
Player 1
𝐹 0, 0 1, 2 0, 0 1, 2

• The Nash equilibria are (𝑂, 𝑜𝑜), (𝑂, 𝑜𝑓), and (𝐹, 𝑓𝑓).

43/51
Example: The Sequential-Move Battle of Sexes (cont.)

• Which of NE is more reasonable?

• What is the difference between (𝑂, 𝑜𝑜) and (𝑂, 𝑜𝑓)?
; The difference is what Player 2 plans to do when Player 1 chooses 𝐹, which is not
reached in the equilibrium.
• If for some reason the Player 1 chooses 𝐹, what should Player 2 do?
• A rational Player 2 should choose 𝐹 as well.
• Consider (𝐹, 𝑓𝑓). Player 2 might say “I will choose 𝐹 no matter what”.
• Is his threat credible?

44/51
Extensive-Form Games

Backward Induction
Backward Induction: Sequentially Rational Play

• How to restrict the set of Nash equilibria to those that are more reasonable?
• The key idea is to require that players play rationally at every decision point.
• That is, they are sequentially rational.
• Backward induction is a solution concept for extensive-form games.
• It is a way to solve games by starting at the end and working backwards.

45/51
Example: The Sequential-Move Battle of the Sexes

1. Put yourself in Player 2’s shoes.

2 ra
; If the Player 1 chooses Opera, Player 2 Ope (2, 1)

should choose Opera. a

er Foo
; If the Player 1 chooses Football, Player 2 1 Op tbal
l (0, 0)
should choose Football.
2. Put yourself in the Player 1’s shoes. Fo ra
; She knows Player 2 is rational.
ot
ba Ope (0, 0)
ll
; She should choose Opera.
Foo
2 tbal (1, 2)
• (𝑂, 𝑜𝑓), giving payoffs (2, 1), survives BI. l

46/51
Example: The Sequential-Move Battle of the Sexes

• There are 3 Nash equilibria: (𝑂, 𝑜𝑜), (𝑂, 𝑜𝑓), and (𝐹, 𝑓𝑓).
• However, only (𝑂, 𝑜𝑓) survives backward induction.
• As a solution concept, NE puts no restrictions on the beliefs of players off the
equilibrium path.
• However, we would expect rational players to play optimally whenever they
are called to move.

47/51
Example: Hume’s Farmer’s Dilemma

Alan 𝑐 (3, 3)
• There are two neighboring farmers.
𝐶 𝑑
• Each will require her neighbor’s help in Fiona (0, 4)
harvesting her corn when it ripens.
• Their corn will ripen at different times. (4, 0)
𝐷 𝑐
• Yet the farmers do not help each other. Why?
Alan 𝑑 (1, 1)

48/51
Example: Hume’s Farmer’s Dilemma (cont.)

Alan 𝑐 (3, 3)
• Fiona’s corn ripens later than Alan’s.
• Put yourself in Fiona’s shoes.
𝐶 𝑑
; “I know Alan is rational.” Fiona (0, 4)
; “If I were to help Alan, he would not help
me because he would get more by not (4, 0)
𝐷 𝑐
helping me.”
• She decides not to help Alan.
Alan 𝑑 (1, 1)

49/51
Example: The Centipede Game

1 2 1 2
𝐶 𝑐 𝐶 𝑐
(3, 3)
𝑁 𝑛 𝑁 𝑛

(1, 1) (0, 3) (2, 2) (1, 4)

• Player 1 can terminate the game by choosing 𝑁 or continue by choosing 𝐶.

• If continue, Player 2 faces the same decision, and so on.
• Notice that as long as the game continues, the sum of payoffs increases.
• The BI predicts the game will end at the first step, giving payoffs (1, 1).

50/51
Summary
Summary

Simultaneous-Move Games Nash Equilibrium

• Players, strategies, and payoffs • Best response
• Matrix representation • Nash equilibrium
• Solution concepts • Mixed strategies

Dominant Strategy Equilibrium Extensive-Form Games

• Dominant strategies • Game trees
• often do not exist • Normal-form representation
• Backward induction

51/51