WIRTSCHAFTSUNIVERSITÄT WIEN

DEPARTMENT OF STRATEGY AND INNOVATION

INSTITUTE FOR MARKETS AND STRATEGY

SBWL Decision Sciences

Game Theory I: Strategic Behavior

Lecture I-7 - Summary

© WU IMS
 TWO GAME THEORISTS

▪ The risks of knowing game theory

© WU IMS 2
 ▪ The value of knowing
what rational agents
would do

© WU IMS 3
 Organization

▪ Exam
▪ Game theory in a nutshell
• Recap of 6 lectures
• Very little time … can only be an overview
▪ Prize lottery
▪ Tutorial

© WU IMS 4
 Organization

▪ Date and time of exam:

• Wednesday 24 April 2024
• 4:00pm-6:00pm
• D5.0.001

▪ Exam will be 90 minutes.

▪ Be on time!

▪ Only (non-programmable) calculator allowed.

© WU IMS 5
 Outline
▪ Basics
• Definitions and terminology, notation
▪ Equilibria
• Nash equilibrium
• Subgame perfect Nash equilibrium
• Mixed strategy equilibria
• Multiple equilibria and coordination
▪ Ways to find equilibria
• Iterated elimination of dominated strategies
• Inspection of normal form → NE
• Rollback in extensive form → SPNE
• Best response functions
• Indifference condition for mixed equilibria
© WU IMS 6
 Outline
▪ Repeated games
• Finitely repeated games
• Infinitely repeated games
▪ Specific games:
• Dilemmas, Bargaining, Chicken, Battle of the Sexes,
Matching Pennies …
▪ Behavioral aspects of strategic behavior
• Social preferences
• Limited depth of reasoning

© WU IMS 7
 Dixit/Skeath Dixit/Skeath/Reiley
References 2nd edition 3rd edition
Chapter:Section Chapter:Section
Terminology: Strategies, Payoffs, etc. 2:3 2:3
Extensive form / normal form games 3:1, 4:1 3:1, 4:1
Nash Equilibrium 4:2, 4:4 4:2, 4:4
Rollback 3:2 3:2
Finitely repeated games 11:2A 11:2A
Infinitely repeated games 11:2B,C,D 11:2B,C,D
Mixed Strategies 7 7
Competition 5:1 5:1
Commitment 10:3, 10:4 10:3, 10:4
Bargaining 17:3 18:3
Dilemmas/Cooperation 11, 12 11, 12
Coordination 8 8

© WU IMS 8
 Basics

▪ Basics
• What is a game?
• What is a strategy?
• What kind of games are there?
• How to write down games?

© WU IMS 9
 Basics
▪ A Game is defined by
• the players
• their potential strategies (possible courses of action)
• utility payoffs associated with combinations of strategies
▪ In math:
• Players i = 1, ..., n
• Strategies si  Si
• Payoffs ui (si , s-i )
▪ Si – set of potential strategies for player i
▪ si – strategy of player i
▪ s-i – set of strategies chosen by all other players j  i
▪ ui (si , s-i ) – payoff (utility) of player i if player i chooses si
and other players have chosen their strategies s-i
© WU IMS 10
 Basics
▪ Strategy:
• a complete plan of action for the game, conditional on
time, nature, information, actions of other players
• covers the whole game

▪ A strategy of a player basically tells us what the player will do

at each possible point in the game where he might be asked
to make a choice.

▪ In dynamic games, strategies must be conditioned on

• reaching certain nodes in the course of the game, and
• interact with information sets (what players know at each
point of time)
© WU IMS 11
 Basics
▪ Games can be:
• Simultaneous (players decide at the same time) or sequential (some players
observe other player’s decisions)
• One-shot (game played once) or repeated (game played repeatedly)
• Finite (game ends at a foreseeable point of time) or infinite (unknown or
probabilistic end or no end at all)
• Symmetric (all players have same available strategies and same utility payoffs)
or asymmetric (players differ in strategies or payoffs)
• Zero/constant-sum (one player looses what the other wins) or non-
zero/constant-sum (both players can win/loose simultaneously)
• Discrete (players have limited set of strategies) or continuous (players have
infinite number of available strategies)
• With complete (everything is known about all players’ strategy sets and
corresponding utility payoffs) or incomplete information (not)вся необхідна інфа, як от правила,
опції куди рухатися, які пейоффи
тощо
• With perfect (no uncertainty about other players’ moves until this point) or
imperfect information (not) якщо ти ти знаєш, яка роль у твого суперника, на якому він вузлі тощо
• Cooperative (players can make binding commitments/contracts before the
game) vs. non-cooperative (not: like all games studied)
© WU IMS 12
 Basics
▪ Normal form representation
Players
Player 2
Strategies
left right
1 0
up 2 0
Player 1
0 2
down
0 1

Payoffs of Player 1
Payoffs of Player 2
© WU IMS 13
 Basics
▪ Extensive form representation
Decision nodes left (2,1)
Root 2
up
right (0,0)
1 Outcomes

left (0,0)
down
2
right (1,2)
Branches

© WU IMS 14
 Basics
▪ Extensive form – Information sets
left (2,1)
2
up
right (0,0)
1
left (0,0)
down
2
right (1,2)
Information set
Player 2 doesn’t know at which
of the the two nodes she is
© WU IMS 15
 Equilibria

▪ Equilibria
• Nash Equilibrium
• Subgame Perfect Nash Equilibrium
• Mixed Strategy Nash Equilibrium

© WU IMS 16
 Equilibria – Nash Equilibrium
▪ Nash Equilibrium
• If the combination of strategies leads to an outcome of a
game in which – given the strategies of the other players – no
player would want to deviate by choosing a different
strategy, then we call it an “equilibrium”.
• Any other outcome is an implausible way of rationally playing
the game, because at least one player could improve by
selecting another strategy.
▪ In math:
• Players i = 1, ..., n, Strategies si  Si, Payoffs ui (si , s-i )
• The strategy profile s* = (s1*, s2*, s3*, …, sn*)  S constitutes a
Nash equilibrium if for each player i the strategy si*  Si is a
best response to the strategies of the other players,
si*  Si*, i.e. if ui (si*, s-i*)  ui (si , s-i*)  si  Si , i = 1, ..., n.
© WU IMS 17
 Equilibria - Subgame perfection
▪ Subgame perfection
• A Nash equilibrium of a game is subgame perfect if it describes
a Nash equilibrium in each subgame of the original game
• A SPNE is a Nash equilibrium refinement
Nash Equilibria of a game here at least in one sub-game you are

non-creadiblre threat
One or more of them are пояснення є в блокноті

Subgame Perfect Nash Equilibria

▪ Every dynamic game with a finite number of players, n < , a

finite set of strategies Si for each player i, and complete
information has at least one subgame perfect Nash equilibrium
in pure or mixed strategies.
▪ Concept of Subgame Perfection implicitly excludes not-credible
threats in games
© WU IMS 18
 Equilibria – Mixed Strategy Nash Eq.
▪ In a mixed strategy, a player chooses a probability
distribution over her pure strategies. That means her
strategy is to play each of her pure strategies with a
certain probability.
▪ Thus, a mixed strategy equilibrium specifies a
combination of all players’ mixed strategies such that no
player has an incentive to play a different probability
distribution.

© WU IMS 19
 Equilibria – Multiple equilibria
▪ There can be multiple Nash equilibria.
• Equilibrium refinements
– Are all equilibria plausible? Do all players’ strategies
prescribe rational behavior at all nodes – even the ones
which are not reached in equilibrium? (subgame perfection)
– Are beliefs correct/rationally derived? (sequential
equilibrium, perfect Bayesian NE, …)
– Are equilibria robust? (trembling hand perfection,
evolutionary stable NE)
• Coordination (equilibrium selection problems)
– Focal points
– Payoff dominance and risk dominance
– Communication, commitment, strategic moves
© WU IMS 20
 Equilibria – Multiple equilibria
▪ Focal points
• Schelling (1960) argued that with multiple equilibria, people
look for a „focal point“, an equilibrium which differs in some
aspects from the others, is more natural, salient, special, etc.
• This way, joint experiences, cultural similarities, other out-of-
the-game histories serve as coordination devices
▪ Out of equilibrium dynamics
• If people have problems to coordinate, an inspection of the
incentives players face around the equilibria might give
indication about the direction behavior will go, i.e. which
equilibrium will be eventually reached.
• This often implies path-dependency, that means the out-of-
equilibrium starting point of game play determines in which
equilibrium players will end up.
© WU IMS 21
 Finding equilibria

▪ Finding equilibria
• Dominance and iterated elimination
• Inspection of normal form → Nash equilibria
• Rollback in extensive form → Subgame perfect NEs
• Continuous strategy spaces and best response functions
• Mixed strategy equilibria and the indifference condition

© WU IMS 22
 Finding Equilibria – Elimination
▪ Dominant strategies: > strictly better than all other strategies,
independent of what the other players are doing,
→ will always be played
▪ Dominated strategies: < strictly worse than some other strategy,
for whatever the other players are doing
→ will never be played
▪ Weakly dominant strategies: >= than all other strategies
▪ Weakly dominated strategies: <= than some other strategy
▪ When analyzing a game, we can reduce the game by deleting a
dominated strategy, and then further analyze the remaining game
→ if used iteratively, sometimes only one strategy for each player
remains = Nash equilibrium
▪ If there is a dominant strategy for one player → all other strategies
are dominated → player will only choose this dominant strategy
© WU IMS 23
 Finding Equilibria – Inspection
▪ In normal form of game, inspect:
• For each column player 2 might choose, find best
responses (best rows) of player 1, mark payoff with *
• For each row player 1 might choose, find best responses
(best columns) of player 2, mark with *
• All cells in which all players Player B
play best responses are
– per definition – left right
Nash equilibria 2* 1
up 1
* 8*
▪ up and left are
dominant Player A
strategies: whatever 6* 5
the other does, they down
are the best reply
0 6
© WU IMS 24
 Finding Equilibria – Inspection
▪ In normal form of game, inspect:
• For each column player 2 might choose, find best
responses (best rows) of player 1, mark payoff with *
• For each row player 1 might choose, find best responses
(best columns) of player 2, mark with *
• All cells in which all players Player B
play best responses are
– per definition – left right
Nash equilibria 4* 4*
up 5
* 4*
Player A
1 5*
down
1 4*
© WU IMS 25
 Finding Equilibria – Inspection
Extensive ( 0, 4)
up
form 2 down
up ( 30 , 5 )
1 up
( 2 , 20 )
down 2 down ( 50 , 19 )

The same game 2

in Normal form
u:up- u:up- u:down- u:down-
d:up d:down d:up d:down
up 0,4 0,4 30*, 5* 30, 5*
1
down 2*, 20* 50*, 19 2 , 20* 50*, 19
© WU IMS 26
 Finding Equilibria – Rollback
3. 1.
up ( 0, 4) Crucial difference in
2 down rollback: P1 assumes
up ( 30 , 5 )
that P2 behaves optimal
1 2. up
( 2 , 20 ) at each possible node
down 2 down ( 50 , 19 )
(SGs 1 & 2)

▪ Find Subgame Perfect Nash equilibria by rollback:

• For the last decision nodes, find the optimal choice.
• Then, given these choices, find the optimal choice at the
nodes before. Continue until root of the game tree.
▪ In subgame 1, 2 should choose “down”.
▪ In subgame 2, 2 should choose “up”.
▪ In subgame 3 (=full game), given behavior in subgames 1 and 2,
1 should choose “up”.
© WU IMS 27
 Finding Equilibria – Inspection and Rollback
▪ To find Nash equilibria:
• Derive all possible strategies and corresponding payoffs
• Write down payoff matrix
• Inspection of payoff matrix

▪ To find subgame perfect Nash equilibria:

• Write down extensive form
• Rollback: inspection of each node (each subgame),
beginning from the end of the game tree

▪ → In simultaneous move games (all players make only one

decision, all at the same time):
Nash equilibria = subgame perfect Nash equilibria
© WU IMS 28
 Finding Equilibria – Continuous strategies
▪ Example: 2 firms set quantities, marginal costs=0, demand
function P=1000-Q
▪ Best response functions
• Profit i = P * qi = (1000-(qi+qk))qi = 1000 qi –qi2+ qiqk
• To get BR: max i → 0 = 1000 - 2qi - qk
• Best response function: qi* = 500 – 0.5 qk
• Nash equilibrium = best response functions intersect
• → qi* = 500 – 0.5 (500 – 0.5 qi*) → qi* = 333
▪ Sometimes, arguments are enough (iterated elimination):
• 2 firms set prices, constant marginal costs=0, all demand
goes to lower price, split half if equal prices
• For each positive price of other firm → underbid
• Prices < 0 yield losses → always better p=0 than p<0
• → only stable situation = mutual best response: price=0
© WU IMS 29
 Finding Equilibria – Mixed strategies
Player B
▪ 2 pure strategy Nash equilibria Left q Right1-q
▪ (up, left) p 3 2
Up
3 15
▪ (down, right) Player A
1-p 4 16
▪ Indifference condition: Down
1 16
▪ Each player plays a mixed strategy, such that player A
chooses up with probability p and down with 1-p, and
player B chooses left with probability q and right with (1-q).
▪ In mixed equilibrium, players choose p and q such that the
other player is indifferent between his pure strategies.
▪ Why? If not, then the other would choose a pure strategy.
▪ Player A: p·3+(1-p)·4=p·2+(1-p)·16 → p=12/13
▪ Player B: q·3+(1-q)·15=q·1+(1-q)·16 → q=1/3
▪ The mixed strategy equilibrium is ((12/13,1/13);(1/3,2/3))
© WU IMS 30
 Finding Equilibria – Mixed strategies
Player B
Player B’s best response function Left q Right 1-q
to Player A’s p choice p 3 2
Up
Player A’s best response function 3 15
to Player B’s q choice Player A
1-p 4 16
Player B’s q Down
1 16
1
Left

▪ The situations where both play

should be able to draw this in th e exam
best response is where the best
reply functions intersect: at
0.5
• p=0 and q=0
• p=1 and q=1
• p=12/13 and q=1/3
Right
0
Down 0.5 Up 1
Player A’s p
© WU IMS 31
 To state an equilibrium, you have to name the
full strategies of all players!
(1,1)
▪If you just say “The equilibrium is Lager
Light beer 2
(Light beer, White)”, then there is
important information missing. White (5,5)
1 Stout
▪Player 1’s choice of “Light beer” (0 , 7 )
only makes sense when one knows
that Player 2 would choose “Stout” Strong beer 2
if Player 1 would choose “Strong”.
Pale Ale ( 7 , 0 )
▪That part of Player 2’s strategy (that he would choose “Stout” if Player 1 chooses
“Strong”) is essential to understand this equilibrium.

▪So the fully-stated (subgame-perfect) equilibrium is:

»Player 1 chooses Light beer,
»Player 2 chooses “White” if P1 chooses “Light”,
and “Stout” if P1 chooses ”Strong”.
32
 Repeated games

▪ Repeated games
• Finitely repeated games
• Infinitely repeated games

© WU IMS 33
 Repeated games – Finitely repeated

▪ If there is only one subgame perfect Nash equilibrium in the

one-shot game, then
• the only subgame perfect equilibrium of the finitely
repeated game is to play this one-shot subgame-perfect
Nash equilibrium again and again ...

▪ If there is only one Nash equilibrium in a one-shot game

• then this NE must be subgame-perfect →
• the only subgame perfect equilibrium of the finitely
repeated game is to play this one-shot Nash equilibrium
again and again ...

© WU IMS 34
 Repeated games – Finitely repeated
ship (+50, +50)
▪ To derive the subgame perfect Nash buy
S not ship
(+0, +70)
B not buy
equilibrium of the repeated game, we (+35, +35)

use rollback and start at the last round. ship (+50, +50)
▪ There, the seller should not ship when buy
S not ship
(+0, +70)
B not buy
the buyer buys: there is no future. (+35, +35)

▪ Thus, the buyer should not buy. ship (+50, +50)

▪ Thus, the second last round has no buy
S not ship
(+0, +70)
B not buy
influence on behavior in the last round. (+35, +35)
Players behave as if there would be no ship (+50, +50)
future. buy
S not ship
(+0, +70)
B
▪ Etc. not buy
(+35, +35)

© WU IMS 35
 Repeated games – infinitely repeated
Player B ▪ Possible trigger strategies:
C D a) “Start with cooperation. If defection
Player A

5 6 observed once, then defect forever.”

C
5 1 b) “Start with cooperation. Then do
1 2 whatever the other before.”
D
6 2
δ ▪ Test: for a) all-time defection profitable?
▪ forb) one- or all-time defection profitable?
▪ All-time defection: 6 +   t  2  5 +   t  5 →   1/ 4 

t =1 t =1
t = t =1 1− 
 
▪ One-time defection: 6 +  1 +   t  5  5 +  5 +   t  5 →   1/ 4
t =2 t =2

▪ If future is important enough (δ sufficiently large), then given all

players play a trigger strategy, deviation is not profitable, so
those trigger strategies are an equilibrium.
© WU IMS 36
 Repeated games – infinitely repeated
Player B
▪ Folk theorem: Each payoff vector which C D
• can be reached by a combination plan of

Player A
5 6
C
strategies (=feasible) 5 1
• has payoffs better than one-shot D
1 2
equilibrium payoffs for all players 6 2
▪ can be implemented as average by
(1,6)
• a) playing the combination 6
plan over available strategies (5,5)
• b) threatening to always play the one-
shot equilibrium if some-

Player 2
body deviates from this plan
▪ The threat is credible, as it is a (2,2)
stage game equilibrium, and worse than (6,1)
any payoff in the green area.
0 Player 1 6
© WU IMS 37
 Specific games

▪ Specific games
• Dilemma games
• Bargaining: Dictator and Ultimatum
• Battle of the Sexes
• Chicken
• Hawk and Dove
• Matching Pennies

© WU IMS 38
 Specific games – Dilemma games
▪ Dilemma game is a game where “individual rationality” and
“group rationality” suggest different ways of behavior
▪ Example
• Price competition (Exps 2&3 souvenirs → 5-person
dilemmas, Exp 13 bolt producers → 2-person dilemma)
• Group: all high price; Ind: undercut

▪ Ways to overcome social dilemmas

▪ Communication, commitment, thresholds
▪ Reputation in repeated interaction
▪ Long-term horizon (“infinity”)

© WU IMS 39
 Specific games – Dilemma games
▪ The prototype: WIFE
The Prisoner’s dilemma
Confess Deny

10 years 25 years
Confess
10 years 1 year
HUSBAND
1 years 3 years
Deny
25 years 3 years

▪ Husband and wife are both jailed, get minor penalties of both
deny, bigger penalties if both confess, but get incentives to
rant out the other. Cannot communicate.
▪ Unique Nash equilibrium in dominant strategies which sets
both players worse off
▪ If the other confesses, I should confess
▪ If the other denies, I should confess
© WU IMS 40
 Bargaining – Dictator and Ultimatum
▪ Dictator Game ▪ Ultimatum Game
100
(100-x, x)
accept

A Offer x B
A Offer x (100-x, x)
reject
(0, 0)
0.1
0.1

▪ One NE: ▪ Many NE, but

A: x=0.1 only one SPNE:
A: x=0.1,
B: accept any x

© WU IMS 41
 Specific games – Battle of the Sexes
▪ 2 pure strategy Nash equilibria: Football fan goes to

• (Opera, Opera) q Opera Football 1-q

p 20 10
• (Football, Football) Opera fan
Opera
30 10
▪ But one NE favors one goes to 1-p 0 30
Football
player, the other NE the 0 20
other player.
▪ → Coordination problem → you would like to move first …
▪ Plus mixed strategy equilibrium
▪ Opera fan tries to set Football fan indifferent, such that
▪ p·20+(1-p)·0=p·10+(1-p)·30 → p= ¾
▪ Football fan tries to set Opera fan indifferent such that
q·30+(1-q)·10=q·0+(1-q)·20 → q= ¼
▪ In mixed NE, both choose their favorite in 75% of the cases.
© WU IMS 42
 Specific games – Chicken
Driver 2
▪ Again, 2 pure strategy Nash Equilibria q Swerve Straight 1-q
• (Straight, Swerve) Swerve
p 0 20
0 -20
• (Swerve, Straight) Driver 1
1-p
▪ Main difference to PD Straight
20
-20
-100
-100
-10
• In PD, when the -10
other deviates from “cooperation”, then I have Prisoners’
an incentive to deviate, too. In Chicken not. dilemma
• In PD dominant strategies, in Chicken not.
▪ Strategic moves: brinkmanship. Disable the steering wheel.
▪ Plus mixed strategy equilibrium
▪ The other is indifferent if p·0 + (1-p)·-20 = p·20 + (1-p)·-100
▪ → p=8/10 → because of symmetry q=8/10
▪ → each driver swerves with a probability of 80%
© WU IMS 43
 Specific games – Hawk and Dove
▪ Structure like Chicken, with 2 pure strategy Bird 2
Nash equilibria q Dove 1-qHawk
(Dove,Hawk), (Hawk,Dove) p Dove V/2 V
V/2 0
▪ Interpretation Bird 1
1-p Hawk 0 (V-C)/2
• What bird mutation (hawk V (V-C)/2
or dove) is successful given a certain population?
• Do stable Hawk/Dove bird populations exist such that
mutations are not successful? → evolutionary stability
▪ Mixed equilibrium = evolutionary stable equilibrium
• p*V/2 + (1-p)*0 = p*V + (1-p)*(V-C)/2 → p = 1-V/C
• The evolutionary stable proportion of Hawks and Doves is V/C
to 1-V/C. That is, the higher the value of resource V and the
lower fighting costs, the more Hawks. If V>C, all will be Hawks.

© WU IMS 44
 Specific Games: Matching Pennies
▪ No pure strategy Nash equilibrium Keeper
Left q Right 1-q
▪ No situation is stable, one p 20 -20
player has always an Left
-20 20
incentive to deviate. Striker
1-p -20 20
▪ You would like to move 2nd. Right
20 -20
Striker’s p
Left
1
▪ Mixed equilibrium: the only
situation where both play
best response is where the
best reply functions intersect
0.5 In math:
p*20+(1-p)*-20 = p*-20+(1-p)*20
→ p=0.5
Righ Keeper’s best q response function to p
t 0 Striker’s best p response function to q
Right 0.5 1
Keeper’s qLeft
© WU IMS 45
 Strategic Moves
▪ A strategic move is an action that changes the game to be played
▪ Try to change
• Set of strategies
• Payoffs
▪ Such that the set of equilibria change to your favor. E.g.
• Disable your steering wheel in Chicken game
• Commit to build a new factory in quantity competition
• Close your eyes as 2nd mover when there is a 1st-mover-
advantage
▪ To truly change a game, strategic moves must be:
• public, irreversible, easy to understand

© WU IMS 46
 Behavioral aspects

▪ Behavioral aspects of strategic behavior

• Social preferences
• Limited depth of reasoning

© WU IMS 47
 Behavioral aspects – Social preferences

y
+

− +
− x

▪ x = how important is my wellbeing to me?

▪ y = how important is other’s wellbeing to me?
© WU IMS 48
 Behavioral aspects – Social preferences
▪ My utility depends on my own income + my relative standing
→ ui=f(yi, øy-i) with øy-i = (ji yj)/(n-1)
▪ I don’t like to be worse off than others and I don’t like to be
better off than others → inequality aversion
▪ Might be asymmetric: inequality to one’s disadvantage hurts
more than inequality to one’s advantage
▪ Example: Ultimatum game A B (80,20)
( 0, 0) UB ?
▪ Egoism: B UA ?
(50,50)
UB ?
B always accepts, A chooses up ( 0, 0)
▪ With inequality aversion: For B, accepting yields
higher payoff than rejecting, but the involved inequality might
be so “painful” that the overall utility is lower than getting zero
and having equality
▪ A anticipates that some B might reject
▪ Thus, some A choose down, and some B reject unfair offers
© WU IMS 49
 Behavioral aspects – Social preferences
▪ My utility depends on my own income and how nice I think
the other was to me, and how nice I think I was to the other
guy → i likes to reciprocate, if not too costly
▪ Beliefs: what does i think about j’s intentions?

▪ Example: Ultimatum game B (80,20)

▪ For both actions of A, if A ( 0, 0) UB ?
B UA ?
B accepts, this is kind to A, (50,50)
( 0, 0)
UB ?
if B rejects, this is unkind to A
▪ If A chooses up this is unkind to B, if A chooses down, this is
kind to B
▪ Thus, if reciprocity is strong, there might be two equilibria:
up/reject (unkind/unkind), and down/accept (kind/kind)

© WU IMS 50
 Behavioral aspects – Social preferences
▪ In all these models of social preferences:
• Utility maximization (=rationality)
• Tradeoff between social motivations (aversion to
inequality, desire to behave reciprocally) and associated
costs (i.e., lower own income)
▪ If game payoffs are properly expressed in utility and players
are assumed to be rational (able to maximize), then we can
apply standard game theory tools to analyze these social
interactions

© WU IMS 51
 Behavioral aspects – Depth of reasoning
▪ People might not fulfill the assumption of rationality
• Specifically, they might only have a limited depth of
reasoning (only one, two, ... cycles of thinking)
• This might actually be rational if thinking is costly →
tradeoff between task optimisation and thinking costs
• Against such players, it might be not the best strategy to
play as if they are rational
▪ k-level thinking models
• Players have different levels of rationality, which can be
expressed by a level k.
• Level-0 players play some simple strategy which is not
dependent on other players‘ behavior.
• Level-k>0-players play a best response to the distribution
of players of levels 0 to k-1.
© WU IMS 52
 Behavioral aspects – Depth of reasoning
▪ Example
1 2 3 4 5 6 7 8 9 10
A: 96
A B A B A B A B A B B: 32

A: 3 A: 1.5 A: 6 A: 3 A: 12 A: 5.5 A: 24 A: 11.5 A: 48 A: 22.5

B: 1 B: 4 B: 2 B: 8.5 B: 4 B: 17 B: 8 B: 34 B: 16 B: 68

▪ Level 0: go down at last move, i.e. 9/10

▪ Level 1: best response (BR) against level 0: go down at 7/8
▪ Level 2: BR against 0 and 1: mix of down at 7/8 or 5/6
▪ Level 3: BR against 0, 1, 2: mix of down at 7/8, 5/6, 3/4
▪ Level 4: BR to 0, 1, 2, 3: mix of down at 7/8, 5/6, 3/4, 1/2
▪ … more levels …
© WU IMS 53
 What we covered …
▪ Basics
• Definitions and terminology, notation
▪ Equilibria
• Nash equilibrium
• Subgame perfect Nash equilibrium
• Mixed strategy equilibria
• Multiple equilibria and coordination
▪ Ways to find equilibria
• Iterated elimination of dominated strategies
• Inspection of normal form → NE
• Rollback in extensive form → SPNE
• Best response functions
• Indifference condition for mixed equilibria
© WU IMS 54
 What we covered …
▪ Repeated games
• Finitely repeated games
• Infinitely repeated games
▪ Specific games:
• Dilemmas, Bargaining, Chicken, Battle of the Sexes,
Matching Pennies …
▪ Behavioral aspects of strategic behavior
• Social preferences
• Limited depth of reasoning

© WU IMS 55
 Course evaluation: C1

© WU IMS 56
 Course evaluation: C2

© WU IMS 57
 Done

▪ Good luck.

© WU IMS 58