Game Theory

Part 1: Static Games with Complete

Information

1
What is it all about?

 We focus on situations where:

 All decision-makers (agents, players) are rational.
 Each player has a well defined set of available choices (strategies).
 The outcome (payoffs) depends on the strategies chosen by all
players – there is “strategic interaction”.

 Example: Six people go to a restaurant.

 Each person pays his own meal – a simple decision problem.
 Before the meal, it is agreed that the total bill will be split evenly
among them – a game.

2
What is game theory?

3 short answers:

 Game theory is another word for "multi-person decision theory".

 Game theory is a tool for the analysis of strategic interaction.
 Game theory is a formal method for the analysis of strategic behavior
in a group of rational players.

 Game theory has applications in

 Economics and sociology
 Politics and law
 even in biology!

3
Static games with complete information:
What is a game?
A static game with complete information is a set G = {I, S, U}, where
 I is a nonempty, finite set of n players
•i  I denotes a player
 S = S1 × S2 × … × Sn = Si × S-i
• Si is a nonempty set called player i's strategy set
• S-i = S1 × ... × Si-1 × Si+1 × … × Sn is the product of strategy sets of
i's opponents
• s = (s1, ... , sn) = (si, s-i)  S is a strategy profile of all players
• si  Si denotes a strategy of player i
• s-i = (s1, ... , si-1, si+1, ... , sn)  S-i is the strategy profile of i's
opponents
 U = (u1, u2, ... , un ) is the vector of payoff functions
• ui: S → R denotes player i's payoff function
• (R is the set of real numbers)

4
Non-cooperative games: Players cannot sign
binding contracts. Three basic assumptions:
 One-shot, simultaneous-move game
Each player chooses a strategy without knowledge of the other
players’ choices. Players receive their payoffs and the game ends.
 Complete information
 The players’ strategy sets and payoff functions are "common
knowledge" (CK) among the players.
• A fact T is CK, if every player knows T, and every player
knows that everyone knows T, and every player knows that
everyone knows that everyone knows T, and ... (ad infinitum).
 Rationality
 Players are rational, i.e., each player maximizes his payoff, given
his belief about what his opponents will play.
 Rationality of players is CK.

5
 Bi-matrix games

 A static game is called finite, if all strategy sets are finite.

 A finite 2-player game can be represented in bi-matrix form, it is
called a bi-matrix game. If player 1 has v strategies and player 2 has
w strategies, then the game is called a v×w bi-matrix game.
 As a convention, the first player chooses from the rows of the bi-
matrix, the second player chooses from the columns. In each cell, the
payoffs are listed in the order u1, u2.
 Example: A 3×2 bi-matrix game with I = {Player 1, Player 2} and
S1 = {a, b, c}, S2 = {d, e}
Player 2
d e
a u1(a,d), u2(a,d) u1(a,e), u2(a,e)
Player 1 b u1(b,d), u2(b,d) u1(b,e), u2(b,e)
c u1(c,d), u2(c,d) u1(c,e), u2(c,e)
6
 Example: Prisoners’ Dilemma

 Two prisoners are convicted of a joint crime. In separate cells, each is

told: "If you both stay mum, we can imprison you for only 1 year. If both
of you confess, you get 6 years each. But if you confess and your
accomplice stays mum, you go free, and he gets 9 years of jail."
 Set of players I = {Prisoner 1, Prisoner 2}, strategy sets S1 = S2 =
{Mum, Confess}, payoff functions:
u1(M, M) = -1, u1(M, C) = -9, u1(C, M) = 0, u1(C, C) = -6
u2(M, M) = -1, u2(M, C) = 0, u2(C, M) = -9, u2(C, C) = -6
Prisoner 2
Players
Strategies Mum Confess
Mum -1 , -1 -9 , 0
Prisoner 1
Confess 0 , -9 -6 , -6

Payoffs
7
Example: Dating Game

Pat
Opera Boxing
Opera 2, 1 0, 0
Chris
Boxing 0, 0 1, 2

 Chris and Pat want to spend the evening together. Chris prefers going
to the opera and Pat prefers the boxing event. Being separated is
equally unpleasant, wherever they go.
 Set of players:I = {Chris, Pat}
 Strategy sets: S1 = S2 = {Opera, Boxing}
 Payoff functions:
u1(O, O) = 2, u1(O, B) = 0, u1(B, O) = 0, u1(B, B) = 1
u2(O, O) = 1, u2(O, B) = 0, u2(B, O) = 0, u2(B, B) = 2

8
 Example: Matching Pennies

Player 2
Head Tail
Head -1 , 1 1 , -1
Player 1
Tail 1 , -1 -1 , 1

 Two players simultaneously show one side of a coin. If the faces of the
coins are different, player 1 gets player 2's coin. If the faces match,
player 2 gets player 1's coin.
 Set of players: I = {Player 1, Player 2}
 Strategy sets: S1 = S2 = { Head, Tail }
 Payoff functions:
u1(H, H) = -1, u1(H, T) = 1, u1(T, H) = 1, u1(T, T) = -1
u2(H, H) = 1, u2(H, T) = -1, u2(T, H) = -1, u2(T, T) = 1
9
Definition: strictly dominated strategy

 Let si', si"  Si. Strategy si' is strictly dominated by strategy si" if
ui(si', s-i) < ui(si", s-i) for all s-i  S-i.

 This means that, no matter what the opponents choose,

strategy si" gives a strictly higher payoff to player i than strategy si'.

 We say that strategy si" strictly dominates strategy si'.

 Think about it:

Can a rational player play a strictly dominated strategy?
Can you construct a game where every strategy is strictly dominated?

10
Definition: weakly dominated strategy

 Let si', si"  Si. Strategy si' is weakly dominated by strategy si" if
ui(si', s-i) ≤ ui(si", s-i) for all s-i  S-i and
ui(si', s-i) < ui(si", s-i) for some s-i  S-i.

 This means that, no matter what the opponents choose,

strategy si" gives at least as much payoff to player i as strategy si', and
a strictly higher payoff for some choice of the opponents.

 We say that strategy si" weakly dominates strategy si'.

 Think about it:
Can a rational player play a weakly dominated strategy?
Is a strictly dominated strategy always weakly dominated?

11
Definition: dominant strategies

 Strategy si is strictly dominant if

ui(si, s-i) > ui(si', s-i) for all s-i  S-i and for all si' ≠ si  Si.

 Strategy si is weakly dominant if

for all si' ≠ si  Si we have ui(si, s-i) ≥ ui(si', s-i) for all s-i  S-i and
ui(si, s-i) > ui(si', s-i) for some s-i  S-i.

 A strictly dominant strategy strictly dominates all other strategies of

the player. A weakly dominant strategy weakly dominates all other
strategies of the player.

 Think about it:

Can you construct a game with a strictly dominant, but without a
strictly dominated strategy?

12
Definition: best responses

 Strategy si is a best response of player i to s-i if

ui(si, s-i) ≥ ui(si', s-i) for all si'  Si.

 Strategy si is a better response to s-i than si' if

ui(si, s-i) > ui(si', s-i).

 We denote the set of player i's best responses to s-i by Bi(s-i). If player i
holds the belief s-i, then (since he is rational) he will play some element
of this set, i.e. some best response to s-i.

 A strategy of player i that is not a best response to any belief s-i is

called a never-best response.

13
Questions about best responses

 Think about it: Can you show that ... ?

 a dominant strategy of player i is a best response to any s-i
 strategy si is a better response than si' to any s-i  S-i if and only if
si strictly dominates si'
 a strictly dominated strategy is a never-best response
 there are games where Bi(s-i) is empty for all i and all s-i
 there are games where Bi(s-i) = Si for all i and all s-i

14
IESDS – Iterated elimination of strictly dominated
strategies
 A rational player will never play a strictly dominated strategy. Hence,
a strictly dominated strategy can be eliminated. Here is an algorithm
for the IESDS:
1. Check if the game has a strictly dominated strategy. If no, you are
done. If yes, eliminate it from the game.
2. Now the game has been reduced, it is "smaller" and less complex.
Go to step 1.
 You end up with a game without strictly dominated strategies. The
remaining strategies are said to survive IESDS. If for each player
only a single strategy survives, then the game is said to be
dominance-solvable.
 Since rationality is CK, a player will never choose a strategy which
is deleted during the IESDS (an iteratively strictly dominated
strategy). Hence, if a game is dominance-solvable, each player will
play his surviving strategy.
15
IESDS – an example

Player 2
Left Middle Right
Up 1, 0 1, 2 0, 1
Player 1
Down 0, 3 0, 1 2, 0

16
IESDS – an example

Player 2
Left Middle Right
Up 1, 0 1, 2 0, 1
Player 1
Down 0, 3 0, 1 2, 0

17
IESDS – an example

Player 2
Left Middle Right
Up 1, 0 1, 2 0, 1
Player 1
Down 0, 3 0, 1 2, 0

18
IESDS – an example

Player 2
Left Middle Right
Up 1, 0 1, 2 0, 1
Player 1
Down 0, 3 0, 1 2, 0

19
Example: Tourists & Natives

 Only two bars (bar 1, bar 2) in a city

 Can charge price of €2, €4, or €5 for a drink
 6,000 tourists pick a bar randomly and order one drink
 4,000 natives select the lowest price bar (if indifferent, each with
probability ½) and order one drink

 Example 1: Both charge €2

each gets (in expectation) 5,000 customers and €10,000 revenue

 Example 2: Bar 1 charges €4, Bar 2 charges €5
 Bar 1 gets 3,000 + 4,000 = 7,000 customers and €28,000 revenue
 Bar 2 gets 3,000 customers and €15,000 revenue

20
Solving "Tourists & Natives" by IESDS

Bar 2
€2 €4 €5
€2 , , ,
Bar 1 €4 , , ,
€5 , , ,

Calculate the payoffs (in thousands of Euros), enter

them in the bimatrix and solve by IESDS!
Yes, do it NOW!

21
Rationalizability and IENBR (iterated elimination
of never-best responses)
 A rational player always plays a best response to his belief. Hence, a
never-best response can be eliminated. Here is an algorithm for the
IENBR:
1. Check if the game has a never-best response. If no, you are done. If
yes, eliminate it from the game.
2. Now the game has been reduced, it is "smaller" and less complex.
Go to step 1.

 You end up with a game where every strategy is a best response to

some belief. The surviving strategies are said to be rationalizable in
the original game.

22
More on rationalizability

 Remember that a rational player only plays strategies that survive

IESDS. But it’s not true that any such strategy can always be played
by a rational player!

 A rational player only plays rationalizable strategies, AND any such

strategy can be played by a rational player!

 In 2-player games, the concepts of IESDS and IENBR are identical:

A strategy survives IESDS if and only if it is rationalizable.
 But in 3-or-more-player games, rationalizability is a stronger concept
than survival of IESDS.

23
Definition: Nash equilibrium

 We know that a player always plays a best response to his belief, and
that this belief is consistent with CK of rationality. Nevertheless, after
the game the belief can turn out to have been wrong.
 If it happens to be the case that all beliefs are correct, then each player
indeed plays a best response to the actual choice of his opponents. In
this case the strategy profile is called a Nash equilibrium.

 A strategy profile s* is a Nash equilibrium if

si*  Bi(s-i*) for all i  I.

 In words, a strategy profile is a Nash equilibrium, if every player plays

a best response to his opponents' choice.

24
Why Nash equilibrium (NE) is an important concept

1. In a NE, the players' beliefs are correct.

2. In a NE, each player plays optimally, given the choices of the other
players.
3. If a NE is played, then after the game no player regrets his choice.
4. If pre-play communication is possible (but remember: no binding
contracts!) and players agree on some strategy profile to be played,
then this strategy profile must be a NE.
5. If there is a "plausible solution" to a game, then it must be a NE. (On
the other hand, not every NE is necessarily a plausible solution!)
6. No player can increase his payoff by unilaterally deviating from a NE,
so in a NE no player has an incentive to deviate from his choice. This
is why NE are called strategically stable or self-enforcing.

25
 Pareto optimality and more on Nash equilibria

 NE need not exist, i.e. not every game has a NE.

 NE, if they exist, need not be unique, i.e. a game may have several NE.
 A NE, if it exists, need not be Pareto-optimal (efficient).

 Repetition: Pareto-optimality
 Strategy profile s is Pareto-better than strategy profile s' if
ui(s) ≥ ui(s') for all i  I, and ui(s) > ui(s') for some i  I.
 In this case s is called a Pareto-improvement over s'.
 A strategy profile s is Pareto-optimal, if there is no Pareto-
improvement over s, i.e. if no other strategy profile is Pareto-better
than s.

Which strategy profiles are Pareto-optimal in the Prisoner’s Dilemma?

26
 How to find NE in bi-matrix games

 In general, finding a NE in a given game can be difficult. However, for

bi-matrix games there is an easy method:

 In each column of the bi-matrix, find the highest payoff(s) of player

1 (left side!) and underline it (them).
 In each row, find the highest payoff(s) of player 2 (right side!) and
underline it (them).
 If in a cell of the bi-matrix both payoffs are underlined, then the
corresponding strategy profile is a Nash equilibrium.

 Think about it: Can you ... ?

construct a bi-matrix game without a NE
construct a bi-matrix game where every strategy profile is a NE

27
A Note on Payoffs

 In examples, payoffs are often amounts of money, minus years in jail,

"points", or similar things. This is mainly for the ease of exposition,
but keep in mind that:
 Payoffs are meant to be von Neumann-Morgenstern utilities. This
means that attitudes toward risk are already incorporated in them.
Hence, if risk (chance) is involved, rational players simply maximize
their expected payoff.

 Think about it:

Assume you are Chris in the Dating Game, and you know that Pat rolls
a dice to determine where to go. If the outcome is "1" or "2", she goes
to the boxing event, otherwise to the opera. What is your best response
to this "mixed" strategy?

28
 Definition: Mixed strategies

A mixed strategy of player i is a probability distribution on Si.

 Strategies in Si are called pure strategies to avoid confusion with
mixed strategies.
 The interpretation of a mixed strategy is that the player does not
choose a pure strategy, but chooses a probability distribution on
his set of pure strategies, and lets chance do the actual choice
(called the realization of the mixed strategy).
 A mixed strategy of player i is denoted by σi, and Σi is the set of
mixed strategies of player i. σ-i and Σ-i are defined as usual.
 If Si is finite, e.g. Si = {si1, si2, ... , sik}, and σi places probabilities
(also called weights) pi1, ... , pik on these pure strategies, then we
conveniently write: σi = pi1 si1 + pi2 si2 + ... + pik sik.

29
More on mixed strategies

 By using expected payoffs, we are actually extending each

player's utility function ui to a function on the set of mixed
strategy profiles Σ = Σi × Σ-i = Σ1 ×...×Σn.
 All the previous definitions involving pure strategies can be
extended to mixed strategies by replacing
si, Si, s-i, S-i, s, S by σi, Σi, σ-i, Σ-i, σ, Σ, respectively.
 A pure strategy can be identified with the mixed strategy placing
weigth 1 on it and weight 0 on all other pure strategies, i.e. we
can identify sij ≡ 0si1 + ... + 0sij-1 + 1sij + 0sij+1 + ... + 0sik.

 Think about it:

Can a pure strategy be strictly dominated by a mixed strategy
without being strictly dominated by a pure strategy?
30
Mixed strategies as best responses

 The support of a mixed strategy is the set of pure strategies

given positive weight by it:
supp(σi) = {sij  Si: pij > 0}.

 σi is a best response to σ-i if and only if all pure strategies in

supp(σi) are best responses to σ-i. (*)
 Intuitively, if a pure strategy in the support were not a best
response, the player could do better by shifting the weight from
this pure strategy to some pure best response.

 Think about it:

Can you prove theorem (*)? (Try it with s-i instead of σ-i first.)

31
Nash's Theorem

 Remember: A strategy profile σ* is a Nash equilibrium (NE) if

σi*  Bi(σ-i*) for all i  I.

In 1950, John F. Nash proved:

 Every finite game has at least one Nash equilibrium.

 Think about it:

Find all NE of the Matching Pennies game!
Find all NE of the asymmetric variant of the Dating game where the
lower-right payoffs are replaced by 2, 3!
Can you construct a game without a NE?

32
Mixed NE as steady states of a learning process

 Imagine a large population of players which are repeatedly randomly

matched in pairs to play a (symmetric) 2-person-game, with each
player using some pure strategy from S = {s1, s2, ... , sk}. The vector of
proportions (p1, p2, ... , pk) of pure strategies used in the population at
time t is called the population state at time t; it is therefore a mixed
strategy σ(t).
 Every now and then a player may change his strategy and switch to
another one which gives a higher payoff. The rules guiding this
switching define a social learning process.
 A learning process is said to converge, if σ(t) converges to some
steady state (or fixed point) σ*, as t goes to infinity. A learning
process need not converge, but if it does, then σ* must be a Nash
equilibrium.
 Why?
 Give a real-world example of this!
33
Game Theory

Part 2: Dynamic Games with Complete

Information

34
Example: The Mafia’s prey

 7 Mafia gangsters have captured a bag containing € 1 Mio. in € 10 bills.

Now these bills have to be distributed among them.
 There is an old allocation algorithm every Mafia member strictly
follows: First, the oldest gangster suggests a distribution. Then all
gangsters vote on this distribution. The majority decides, and in case of a
tie the oldest gangster has the decisive vote.
 If the suggested distribution gets a majority of votes, it is implemented.
If it is rejected, the oldest gangster is killed and the process starts again
with the second oldest gangster suggesting a distribution.
 Each gangster wants to survive. Given survival, he wants to maximize
the number of coins he receives. Given the number of coins he receives,
he prefers his comrades to be killed. All this is CK in the Mafia.

35
Example: The Mafia’s prey

 7 Mafia gangsters have captured a bag containing € 1 Mio. in € 10 bills.

Now these bills have to be distributed among them.
 There is an old allocation algorithm every Mafia member strictly
follows: First, the oldest gangster suggests a distribution. Then all
gangsters vote on this distribution. The majority decides, and in case of a
tie the oldest gangster has the decisive vote.
 If the suggested distribution gets a majority of votes, it is implemented.
If it is rejected, the oldest gangster is killed and the process starts again
with the second oldest gangster suggesting a distribution.
 Each gangster wants to survive. Given survival, he wants to maximize
the number of coins he receives. Given the number of coins he receives,
he prefers his comrades to be killed. All this is CK in the Mafia.

 Who gets how much?

36
Dynamic games with complete information:
Trees, nodes, and actions
 A dynamic game with complete information is a game
proceeding over several stages, so there is a time-structure. Its
exact definition is rather complicated, but here is a description:
 A dynamic game is described by a game tree. A game tree
specifies which player moves at which stage, what information he
has, which actions are available to him, and what payoffs are
associated with different sequences of actions.
 A game tree starts at a node called the root. Some player starts
with one of several available actions. Each action leads to a
different node of the tree. At each node either one of the players
moves (a decision node) or "Nature" makes a chance move
according to some probability distribution, or the game ends (in a
terminal node). Each terminal node is associated with a vector
of payoffs for the players.
37
Information sets and perfect information

 For each player, the set of decision nodes where he has to move is
partitioned into information sets. An information set contains all
nodes the moving player cannot distinguish between. (He knows
at which information set he has arrived, but not at which node
within this information set. Necessarily, the set of possible actions
has to be the same for all nodes in an information set.) An
information set is depicted by a dashed line connecting the
respective nodes in the game tree.
 If an information set contains only a single node, it is called a
singleton. A dynamic game where all information sets are
singletons and there are no chance moves is said to have perfect
information, otherwise there is imperfect information.

38
Plays and strategies in dynamic games

 A sequence of actions leading from the root to some terminal

node is called a play of the game.
 If a dynamic game contains only finitely many nodes, we call it
finite. If a dynamic game is infinite, it either contains nodes with
infinitely many possible actions, or it allows for plays of infinite
lengths. If all possible plays have finite length, we say the game
has a finite horizon.

 A pure strategy of a player in a dynamic game is a complete

plan of action. This means that it has to specify one of the
available actions at each information set of this player.

39
The strategic form of a dynamic game

 Given a pure strategy for each player (i.e. a pure strategy profile),
the play of the game is uniquely determined (up to chance
moves), and so is the vector of payoffs.
 Hence we can translate a dynamic game into a static game just
by constructing each player's pure strategies and payoff function.
The static game constructed in this way is called the strategic
form of the dynamic game.
 Having done this, we can apply all the definitions, including
domination, mixed strategies, etc., from static games to dynamic
games as well. Most importantly:

 A strategy profile σ* is called a Nash equilibrium of a dynamic

game if it is a Nash equilibrium of its strategic form.

40
 An example

1
 Think about it:
a b
N
2 What are the pure
i [0.8]
strategies of the
h [0.2] players?
c
d
1 e
2 What is the strategic
-2
f f k form of this game?
g g j -5

0 1 Find all pure-strategy

2 4 5 0 3
3 3 NE of this game!
0 1 0 0 10

41
 Example: Entry Deterrence

 Firm Y is a monopolist in the X

market, but is threatened by entry of
X. If X decides to stay out of the in out
market, it makes no profit, and Y Y
keeps the monopoly profit of 8. But
if X enters, Y can either share the 0
share fight
8
market, yielding a profit of 3 to each
firm, or start a price war, which 3 -1
would result in losses of 1 to both. 3 -1

 Solve this game! What will the

firms eventually do?

42
 Analyzing Entry Deterrence

 Each firm has only one decision

X
node, so X's and Y's pure strategy
sets are {in, out} and {share, fight}. in out
 The strategic form of the game is:
Y
Y
share fight 0
share fight
in 3,3 -1 , -1 8
X
out 0,8 0,8 3 -1
3 -1
 The NE are easy to find here, they are
(in, share) and (out, fight).
 Do you think that (out, fight) is a
plausible solution?

43
 Analyzing Entry Deterrence (ctd.)

 If the play reached Y's decision

X
node, i.e. if X chose "in", then Y
would not fight. Hence the NE (out, in out
fight) rests on the non-credible
threat of Y to fight in case of X's Y
entry. fight
0
share 8
 Being rational, X should enter, and
Y should then share the market. The 3 -1
NE (in, share) is a plausible solution 3 -1
of this game.
 The NE (out, fight) is implausible,
because it prescribes an irrational a subgame
move of Y in the subgame starting
at Y's decision node.
44
 Subgames

 A subgame of a dynamic game is a 1

part of the game tree that starts at a
singleton information set and N
contains all subsequent nodes (and 2
only those).
 Note that by this definition the
complete game is always a subgame 1 2
of itself. All other subgames are
called proper subgames.
 A game need not have a proper
subgame. Give an example!
 A strategy profile in the complete
game naturally induces a (subgame-)
strategy profile in each subgame. The proper subgames of the
example from p. 41.

45
 Subgame perfect equilibrium (SPE)

 A Nash equilibrium of a dynamic X

game is said to be subgame in out
perfect, if it induces a Nash
equilibrium in every subgame. Y
0
share fight
 Subgame perfection excludes all NE 8
which rely on non-credible threats. 3 -1
3 -1
 It can be shown that every finite
dynamic game with complete This subgame is a 1-player game with NE
information has a subgame perfect (share). The NE (out, fight) of the
Nash equilibrium. complete game induces (fight) in this
subgame, hence it is not subgame perfect.
Only the NE (in, share) is a SPE.
46
 Finding SPEs in perfect information games:
Backwards induction
 In a perfect information game with X
finite horizon, there is an easy way in out
of finding SPEs, called backwards
induction: Y
 Start at the end of the game tree (at 0
share fight
one of the last decision nodes) and 8
determine the moving player's -1
3
optimal action(s). Move up to the 3 -1
next-to-last decision nodes and
determine the moving players' If Y gets to move, its optimal action is
optimal actions, taking the optimal "share". Given this, X effectively
actions determined before as given. decides between the payoff vectors (3,3)
Proceed in this way until you reach and (0,8). Hence its optimal action is
"in". Backwards induction thus delivers
the root.
the SPE (in, share).
47
 Example: Centipede game

 You and another guy are sitting in two separated rooms in front of a
slot-machine. The two slot-machines are connected, and each has a
coin slot and a Stop button. Whenever one of you throws € 1 into the
slot machine, the other one receives € 2. However, if someone presses
the Stop button, the game is over. You move alternatingly for at most
10 rounds. You start.

 How would you play this game in reality?

 How would rational money-maximizing players play this game?

48
 Example: Finitely repeated PD

2
Cooperate Defect
Cooperate 3,3 0,5
1
Defect 5,0 1,1

 You (player 1) and another guy (player 2) are playing this variant of
the Prisoners' Dilemma game. Obviously, in the one-shot game,
D(efect) is strictly dominant. However, you play this game repeatedly
for 5 rounds. In the end you receive the sum of the payoffs of each
round.

 How would you play the repeated game in reality?

 How would rational players play the repeated game?

49
 Example: Indefinitely repeated PD (rPD)

 The rPD: You are playing the PD 2

C D
repeatedly. However, the number of
C 3,3 0,5
rounds is not fixed. After each round, the 1
game ends with a probability of p and is D 5,0 1,1
continued with a probability of 1-p.
 Show that both players playing "Always Defect" is a NE for the rPD.
 The following strategy T is called a trigger strategy for the rPD:
 Start with C. In every round thereafter, if the outcomes until now have
all been (C, C), play C, otherwise play D.
 Show that the profile (T, T) induces cooperation in every round.
 Show that, for small values of p, (T, T) is a Nash equilibrium of the
repeated PD.
 Show that (T, T) is even subgame perfect.

50