Price Theory B – Game Theory – Static games of

complete information

Alex Gershkov

The Hebrew University of Jerusalem

May 2018

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 1 / 50
 Game Theory

Snyder and Nicholson, twelfth ed. Ch 8.

Static games of complete information.
Normal form representation
Dominant strategy and dominant strategy equilibrium
Iterative elimination of strictly dominated strategies
Nash equilibrium in pure strategies
Nash equilibrium in mixed strategies
Topic 2: Dynamic games of complete information
Extensive form representation
Nash equilibrium
Subgame perfect equilibrium
Backward induction

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 2 / 50
 What is game theory

"Game theory concerns the behaviour of decision makers whose

decisions a¤ect each other. Its analysis is from a rational rather than a
psychological or sociological viewpoint. It is a sort of umbrella theory for
the rational side of social science, where ‘social’is interpreted broadly, to
include human as well as non-human players (computers, animals,
plants). Its methodologies apply in principle to all interactive situations,
especially in economics, political science, evolutionary biology, and
computer science." (The New Palgrave Dictionary of Economics)

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 3 / 50
 Elements of a game

Players. Are the individuals who make decisions.

Actions. An action by player i is a choice (s)he can make. Player i’s action
set is the entire set of actions available to i.
Payo¤s. By player i’s payo¤, we mean the utility i receives after all players
have executed their strategies and the game has been played out.
Information. Speci…es what player i knows at di¤erent points in time.
Usually modeled as information sets.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 4 / 50
 The Prisoners’Dilemma

Two suspects are charged with a crime. The police lack su¢ cient evidence to
convict the suspects, unless at least one confesses.
If neither confesses: each is sentenced to 1 month of prison.
If both confess: each is sentenced to 6 months of prison.
If only one confesses: he/she is released immediately but the other is
sentenced to 9 months of prison.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 5 / 50
 The Prisoner’s Dilemma as a game

This is a game with the following features:

two players
actions are equivalent to strategies in a static or ‘one-shot’game.
the strategy set for each player is Si = fSilence, Confess g, i = 1, 2 for
players 1 and 2; or fS, C g for short. Each player chooses a strategy (or
action) si 2 Si , i.e., si = S or C giving 4 possible outcomes (C , C ), (C , S ),
(S, C ), (S, S ) for (s1 , s2 ).
the payo¤s (π 1 (s1 , s2 ), π 2 (s1 , s2 )) associated with these outcomes:
(π 1 (C , C ), π 2 (C , C )), (π 1 (S, C ), π 2 (S, C )), (π 1 (C , S ), π 2 (C , S )) and
(π 1 (S, S ), π 2 (S, S )) are shown in the payo¤ matrix.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 6 / 50
 The Prisoner’s Dilemma as a game

We can construct the following payo¤ matrix:

take a suspect’s utility to be -sentence in months
then the payo¤ matrix takes the form:
S C
S 1, 1 9, 0
C 0, 9 6, 6

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 7 / 50
 The normal-form representation

The normal-form representation of a game speci…es:

1 the players in the game,
2 the strategies available to each player,
3 the payo¤ received by each player for each combination of strategies that
could be chosen by the players.

De…nition
The normal-form representation of a game speci…es the set of players N, the
players’strategy spaces S1 , ..., Sn and their payo¤ functions π 1 ,...,π n . We denote
this game by G = fN, S1 , ..., Sn , π 1 , ..., π n g.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 8 / 50
 Best response

For any vector (list) y = (y1 , ..., yn ), denote by y i the vector

(y1 , ..., yi 1 , yi +1 , ..., yn ), which is the portion of y not associated with player i.

De…nition
Player i’s best response or best reply to the strategies s i chosen by the other
players is the strategy si that yields him/her the greatest payo¤; that is

π i ( si , s i ) π i si0 , s i for any si0 6= si .

The best response is called a strong (or strict) best response if no other
strategies are equally good, and weakly best otherwise.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 9 / 50
 Dominant strategy equilibrium

De…nition
1. The strategy si is called a dominant strategy if it is a player’s strictly best
response to any strategies the other players might pick, in the sense that whatever
strategies they pick, his payo¤ is highest with si . Mathematically,

π i (si , s i ) > π i si0 , s i for any si0 6= si and any s i .

2. Strategy si0 is a dominated strategy, if there exists si such that for any possible
strategies of the other players, s i , i’s payo¤ from si0 is strictly less than from si .
Mathematically, there exists si such that

π i (si , s i ) > π i si0 , s i any s i .

De…nition
A dominant strategy equilibrium is a strategy combination consisting of each
player’s dominant strategy.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 10 / 50
 Joint project

You are working with a friend on a joint project.

Each of you can either work hard or shirk.
If your friend works hard then you prefer to shirk.
You prefer the outcome of your both working hard to the outcome of your
both shirking.
The worst outcome for you is that you work hard and your friend shirks.

work hard shirk

work hard 2, 2 0, 3
shirk 3, 0 1, 1

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 11 / 50
 Duopoly

Two …rms.
Each …rm charges either a high or a low price.
If both choose High, each earns 1000.
If one chooses High and the other chooses Low, the …rst one loses 200, the
second one gets 1200.
If both choose Low, each earns 600.
Each …rm cares only about its pro…t.

High Low
High 1000, 1000 200, 1200
Low 1200, 200 600, 600

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 12 / 50
 Other applications

An arms race

Assume that each of two countries can build an arsenal of nuclear bombs, or
can refrain from doing this. Each country’s favorite outcome is that it has
bombs and the other country does not; the next best outcome is that neither
country has any bombs; the next best outcome is that both countries have
bombs; the worst outcome is that only the other country has bombs.

Common property

Two farmers decide to which extent they allow their sheep to graze on the
village common. Each farmer prefers that her sheep graze a lot to a little,
regardless of the other farmer’s action, but prefers that both farmers’sheep
graze a little to both farmers’sheep grazing a lot.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 13 / 50
 Rationality

If we assume that the players are rational, then we should expect them to choose
whatever they deem to be best for them.
Claim A rational player will never play a dominated strategy.
Knowledge of the game implies that a player should recognize dominated
strategies, and rationality implies that these strategies will be avoided.

This idea is called dominant strategy equilibrium (DSE).

Remember: The solution to a game should always be described as the set of the
players’strategies, not their payo¤s!

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 14 / 50
 Evaluating Dominant Strategy equilibrium

1. If DSE exists, this solution concept guarantees uniqueness.

If the game G = N, fSi gni=1 , fπ i gni=1 has a dominant strategy equilibrium s ,
then s is the unique dominant strategy equilibrium.
2. How about the existence? Consider the Battle of the Sexes game

Chris
O F
Alice O 2, 1 0, 0
F 0, 0 1, 2

In many games, DSE will fail to predict the players’choices.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 15 / 50
 Evaluating Dominant Strategy equilibrium
3. Pareto criterion.
The outcome prescribed by this solution concept may not be Pareto optimal (see
the Prisoner’s Dilemma).
The failure of Pareto optimality is not a failure of the solution concept. The
assumption that the players are rational causes them to confess in the Prisoner’s
Dilemma. The failure of Pareto optimality implies that the players would bene…t
from modifying the environment and creating some enforcement mechanism that
will punish those who confess ("ma…a").
Assume that "punishment" from confessing generates disutility of z, then the
modi…ed Prisoner’s Dilemma is represented by

Player 2
S C
Player 1 S 1, 1 9, 0 z
C 0 z, 9 6 z, 6 z

If z > 3, this punishment is su¢ cient to ‡ip the predicted equilibrium outcome of
the game, because then S becomes the dominant strategy (and (S, S ) is the
Pareto optimal outcome).
Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 16 / 50
 Rationality

Rationality implies that

1 A rational player will never play a dominated strategy.
2 If a rational player has a dominant strategy then he will play it.
DSE requires only rationality.

Problem
DSE will often fail to exist.

Solution
Alternative approaches that will apply to a variety of games.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 17 / 50
 Iterated Elimination of Strictly Dominated Strategies

Illustration
l m r
1
t 0, 0 4, 1 1, 1
m 1, 4 5, 3 3, 2
b 1, 2 0, 3 4, 1
Player 2 has a dominated strategy r. A rational player 2 should never play r.
If player 1 knows that player 2 is rational, he knows that player 2 never
chooses r (Eliminate r).

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 18 / 50
 Iterated Elimination of Strictly Dominated Strategies

Illustration
l m r
1
t 0, 0 4, 1 1, 1
m 1, 4 5, 3 3, 2
b 1, 2 0, 3 4, 1
Player 2 has a dominated strategy r. A rational player 2 should never play r.
If player 1 knows that player 2 is rational, he knows that player 2 never
chooses r (Eliminate r).
l m
2
t 0, 0 4, 1
m 1, 4 5, 3
b 1, 2 0, 3
Now player 1 has a dominated strategy - b. In this case, player 1 should
never play b. (Eliminate b).

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 18 / 50
 Iterated Elimination of Strictly Dominated Strategies

Illustration (cont.)
l m
3. t 0, 0 4, 1
m 1, 4 5, 3
Player 2 has a dominated strategy m. If P2 is rational and knows that P1 is
rational (and P2 knows that P1 knows that P2 is rational), then P2 shouldn’t
play m. (Eliminate m.)

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 19 / 50
 Iterated Elimination of Strictly Dominated Strategies

Illustration (cont.)
l m
3. t 0, 0 4, 1
m 1, 4 5, 3
Player 2 has a dominated strategy m. If P2 is rational and knows that P1 is
rational (and P2 knows that P1 knows that P2 is rational), then P2 shouldn’t
play m. (Eliminate m.)
l
4. t 0, 0
m 1, 4
Now player 1 has a dominated strategy - m. If P1 is rational and knows that
P2 is rational (and P1 knows that P2 knows that P1 is rational . . . ), then P1
shouldn’t play m. (Eliminate m.)
Hence (t, l) is the only strategy-pro…le surviving IESDS and thus the result of
our analysis.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 19 / 50
 Iterated Elimination of Strictly Dominated Strategies

De…nition
We will call any strategy pro…le s IESDS = s1IESDS , ..., snIESDS that survives the
process of IESDS an iterated-elimination equilibrium (IEE).

Like the concept of DSE, the IEE starts with the premise of rationality. However,
in addition to rationality it requires a lot more.
We will discuss now the additional requirements.
You can …nd explanations about this solution concepts in most of the game theory
books (including introductory). In particular in
1 Myerson R., "Game Theory: Analysis of Con‡ict," Harvard University Press
1991,
2 Fudenberg D., and J Tirole, :Game Theory," The MIT Press1991.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 20 / 50
 Common knowledge

Let Ki X to mean player i knows X .

De…nition
X is called common knowledge among i 2 fA, B g if Ki Kj Ki ...X , (n times) i 6= j,
n = 1, 2, 3, ... for as many rounds as required.

IESDS requires the following assumptions

1 rationality of all players,
2 common knowledge of rationality, and
3 common knowledge of the game structure fN, S, π g.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 21 / 50
 Common knowledge - implication

some people on an island have blue eyes

the rest - green eyes
no one on the island ever knows their own eye color (there are no mirrors)
if a person on the island ever discovers they have blue eyes, that person must
leave the island that day evening
on the island, each person knows every other person’s eye color, but there is
no discussion of eye color
at some point, an outsider comes to the island, calls together all the people,
and makes the following public announcement: "At least one of you has blue
eyes"
it becomes common knowledge that there is at least one islander who has
blue eyes

Question: what is the eventual outcome?

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 22 / 50
 IESDS - a general process

Let Sik denote the strategy set of player i that survives k rounds of IESDS.
Step 1: De…ne Si0 = Si for each i,and set k = 0.
Step 2: Are there players for whom there are strategies si 2 Sik that are
strictly dominated? If yes, go to step 3. If not, go to step 4.
Step 3: For all players i 2 N, remove all strategies si 2 Sik that are strictly
dominated. Set k = k + 1, and de…ne a new game with strategy
sets Sik that do not include the strictly dominated strategies that
have been removed. Go back to step 2.
Step 4: The remaining strategies in Sik survive IESDS.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 23 / 50
 Application: Beauty contest

Consider the following game:

n players
each player chooses an integer between 1 and 100, Si = f1, ..., 100g
the winners are the players who choose an integer number that is closest to
1/2 of the average
That is, if there are three players and they announced numbers 26, 60, 91. The
winner is player 1, since the average is 59, and 12 59 = 29.5

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 24 / 50
 Application: Hotelling location game

Two vendors A and B sell some homogenous good; they have n a discrete o
number of possible locations along some ‘linear city’ [0, 1]: 0, γ1 , γ3 , , γγ
for even γ (thus 1/2 is possible).
Vendors choose their location simultaneously.
Consumers populate the city uniformly [0, 1] and buy their good from
whoever is closer (transport cost).
Pro…t of vendors is their volume of trade (for instance, the price per unit is 1
and marginal cost is zero).

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 25 / 50
 Application: Hotelling location game

First round of elimination: choosing 0 is dominated by γ1 , while choosing γ

γ
γ 1
is dominated by γ .
1 1
1 (0, 0) ! γ,0 : Original utility of A is 0.5, in case of choosing γ it becomes
1
1 2γ > 0.5
1 1 1 1 1
2 0, γ ! γ, γ : Original utility of 2γ , in case of choosing γ it becomes
1
0.5 > 2γ
n 1 n n 1
3 0, γ ! γ, γ with n > 1: Original utility 2γ , in case of choosing γ it
n +1
becomes 2γ
γ 1
Therefore, 0 is dominated by γ1 . Similar argument implies γγ is dominated by γ .
Therefore, both …rms will not choose the extreme locations.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 26 / 50
 Application: Hotelling location game

Second round of elimination: since both …rms eliminated extreme locations

0 and γγ we can show that choosing γ1 is dominated by γ2 , while choosing
γ 1 γ 2
γ is dominated by γ (if γ > 1).

Through repetition of the above elimination, IESDS eliminates all grid points
except for the middle point.
=) IESDS gives the unique recommendation to play (1/2, 1/2).

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 27 / 50
 IESDS - Evaluation

1 Existence. We can apply IESDS to any game (by applying the algorithm). It
does not require the existence of a dominant or dominated strategies.
Existence is a consequence of common knowledge of rationality. We are
giving the players the ability to reason through repeating of the strategic
implications of rationality, while correctly anticipating that other players can
perform similar reasoning. Rationality alone is not su¢ cient.
2 Uniqueness. Consider the Battle of Sexes game

Chris
O F
Alice O 2, 1 0, 0
F 0, 0 1, 2

IESDS cannot restrict the set of strategies. Therefore, "anything can happen".

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 28 / 50
 IESDS - Evaluation

3. Pareto e¢ ciency. We already established (using Prisoner’s Dilemma) that the

unique survivor of IESDS may not be Pareto E¢ cient. Thus common
knowledge of rationality does not mean that players can guarantee the best
outcome for themselves.
4. Relationship netween IESDS and DSE:
If for a game hN, fSi g, π i i s is a dominant strategy equilibrium, then s
uniquely survives IESDS. (See Prisoner’s Dilemma).

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 29 / 50
 Nash equilibrium

Consider the Battle of the Sexes

Chris
O F
Alice O 2, 1 0, 0
F 0, 0 1, 2

Pro…le: A goes to opera (behaves optimally if believes that C goes to opera),

while C goes to football (behaves optimally if believes that A goes to
football).
Additional requirement for Nash equilibrium: in equilibrium the believes are
correct!

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 30 / 50
 Nash equilibrium

In a Nash equilibrium
1 each player is playing a best response to his/her beliefs
2 players have correct beliefs.
Alternatively, each player is choosing a best response to the equilibrium strategies
of all other players.

De…nition
The (pure) strategy pro…le s = (s1 , ..., sn ) is a Nash equilibrium if si is a best
response to s i , for all i. That is,

π i si , s i π i si0 , s i for all si0 2 Si and all i 2 N.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 31 / 50
 Nash equilibrium - example

l m r
t 0, 0 4, 1 1, 1
m 1, 4 5, 3 3, 2
b 1, 2 0, 3 4, 1

(t,l) Nash equilibrium, because BR1 (l) =t and BR2 (t) =l.
In this case, player 1 (correctly) believes that player 2 is playing l and hence player
1 is playing t; player 2 (correctly) believes that player 1 is playing t and hence
player 2 is playing l.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 32 / 50
 Nash equilibrium - Application 1

Stag hunt game

Two hunters (P1 and P2).
Choice between hunting a stag (S) or hare (H).
Hunting a stag requires joint e¤orts (more bene…cial for society).
Hunting a hare is an individual task.

S H
S 5, 5 0, 3
H 3, 0 3, 3

There are two (pure strategy) Nash equilibria: (S,S) and (H,H).
However (S,S) Pareto dominates (H,H).

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 33 / 50
 Nash equilibrium - Application 2

The tragedy of the commons: Hardin (1968)

n players
each choosing how much to produce.
production consumes some of the clean air
K total amount of clean air
each player chooses his own consumption of clean air for production ki
the bene…t of consuming ki for production is ln (ki )
the bene…t of consuming the remainder of the clean air is ln K ∑nj=1 kj
the payo¤ of i is
!
n
π i (ki , k i ) = ln (ki ) + ln K ∑ kj .
j =1

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 34 / 50
 Nash equilibrium - Application 2
If pro…le (k1 , ..., kn ) satis…es that

ki = BRi (k i ) for all i 2 N

then (k1 , ..., kn ) is a Nash equilibrium.

FOC for player i
∂ 1 1
π i ( ki , k i ) = = 0.
∂ki ki K ∑nj=1 kj

Therefore
K ∑ j 6 = i kj
BRi (k i ) = .
2
For the two players case
K k2
k1 (k2 ) = ,
2
K k1
k2 (k1 ) = .
2
Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 35 / 50
 Nash equilibrium - Application 2
Best-response functions (two-player case)

K/2
Nash equilibrium
K/3

K/3 K/2 K

The unique Nash equilibrium is k1 = k2 = K /3

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 36 / 50
 Nash equilibrium - Application 2

Could this two-player society do better?

Choosing production levels to maximize the sum of the agents utilities implies
!
2 n
max
k 1 ,k 2
∑ ln (ki ) + 2 ln K ∑ kj .
i =1 j =1

The FOC are

1 2
= 0,
k1 K k1 k2
1 2
= 0.
k2 K k1 k2

The unique solution is k1 = k2 = K /4, which Pareto dominates (Pareto

improvement on) the Nash equilibrium outcome.
Conclusion. Giving people the freedom to make choices may make them worse
o¤ than if those choices were regulated!

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 37 / 50
 Nash equilibrium

Mixed Strategies
We will now allow players to randomize between di¤erent actions.
Consider the following game (Matching Pennies).
Two players, each player puts a coin (penny) on a table. If both coins come
up the same side (heads or tails) then player 1 gets both; otherwise player 2
does.
H T
H 1, 1 1, 1
T 1, 1 1, 1

There is no pure-strategy Nash equilibrium, however Nash equilibrium exists if we

allow players to choose random actions.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 38 / 50
 Nash equilibrium

Mixed Strategies
Another game — rock-paper-scissors

R P S
R 0, 0 1, 1 1, 1
P 1, 1 0, 0 1, 1
S 1, 1 1, 1 0, 0

Best response for P 1, when he believes that P 2 will play one of his pure strategies
8
< P if s2 = R
s1 ( s 2 ) = S if s2 = P
:
R if s2 = S.

There is no pure strategy equilibrium.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 39 / 50
 Nash equilibrium
Mixed Strategies
Another game — all-pay auction
two players
one prize of value v > 0
players submit their bids simultaneously
the player with the highest bid gets the prize
in case of tie every player gets the prize with probability 1/2
every player pays his/her bids bid
utility of i is 8
< bi if bi < bj
v
π i ( bi , bj ) = b i if bi = bj
: 2
v bi if bi > bj
no pure strategy equilibrium
submitting bid bi > v is dominated by submitting 0
there is no equilibrium in which both players submit bid of v
there is no equilibrium in which both players submit the same bid below v
there is no bid in which agents submit di¤erent bids, and both are below v
Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 40 / 50
 Nash equilibrium

Mixed Strategies
Mixed strategies for …nite strategy sets

De…nition
Let Si = fsi 1 , si 2 , ..., sim g be player i’s …nite set of pure strategies. A mixed
strategy for player i, σi = fσi (si 1 ) , σi (si 2 ) , ..., σi (sim )g is a probability
distribution over Si , where σi (si ) is the probability that i plays si . The set of
distributions over set Si is denoted by ∆Si .

Recall that since σi is a probability distribution,

1 σi (si ) 0 for all si 2 Si
2 ∑ s i 2S i σ i (si ) = 1
Every pure strategy is a mixed strategy with a degenerate distribution that
chooses a single pure strategy with probability 1.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 41 / 50
 Nash equilibrium

Mixed Strategies-Rede…ning Beliefs

If players choosing mixed strategies, we have to enrich and rede…ne the beliefs
that players can have.

De…nition
A belief for player i is given by a probability distribution pi 2 ∆S i over the
strategies of his opponents. We denote by pi (s i ) the probability player i assigns
to his opponents playing s i 2 S i .

For example, in the rock-paper-scissors game, the beliefs of player 1 are

(p1 (r ), p1 (p ), p1 (s ))

with p1 (r ) 0, p1 (p ) 0, p1 (s ) 0 and p1 (r ) + p1 (p ) + p1 (s ) = 1.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 42 / 50
 Nash equilibrium - Expected Payo¤s
Recall the Matching Pennies game.
Assume that player 2 chooses σ2 (H ) = 13 and σ2 (T ) = 23 . If player 1 plays H
with probability 13 he will get 1 while with probability 32 he will get 1. If player 1
plays T with probability 32 he will get 1, while with probability 13 he will get 1.
Player 1 faces lotteries and di¤erent actions imply di¤erent lotteries.
Our players evaluate lotteries based on the expected payo¤ they generate.

De…nition
The expected payo¤ of player i when (s)he chooses the pure strategy si 2 Si and
his/her opponents play the mixed strategy σ i 2 ∆S i is

π i ( si , σ i ) = ∑ σ i ( s i ) π i ( si , s i ) .
s i 2S i

The expected payo¤ of player i when (s)he chooses the mixed strategy σi 2 ∆Si
and his/her opponents play the mixed strategy σ i 2 ∆S i is π i (σi , σ i ) =
!
∑ σ i ( si ) π i ( si , σ i ) = ∑ ∑ σ i ( si ) σ i ( s i ) π i ( si , s i ) .
s i 2S i s i 2S i s i 2S i
Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 43 / 50
 Nash equilibrium - Expected Payo¤s

Consider the rock-paper-scissors game

R P S
R 0, 0 1, 1 1, 1
P 1, 1 0, 0 1, 1
S 1, 1 1, 1 0, 0

Assume that player 2 plays σ2 (R ) = σ2 (P ) = 12 ; σ2 (S ) = 0. The expected

payo¤ to player 1 from his pure strategies
1 1 1
π 1 (R, σ2 ) = 0+ ( 1) + 0 1=
2 2 2
1 1 1
π 1 (P, σ2 ) = 1+ 0+0 ( 1) =
2 2 2
1 1
π 1 (S, σ2 ) = ( 1) + 1+0 0 = 0.
2 2

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 44 / 50
 Mixed-Strategy Nash Equilibrium

De…nition
The mixed-strategy pro…le σ = (σ1 , σ2 , ..., σn ) is a Nash equilibrium if, for each
player, σi is a best response to σ i . That is, for all i 2 N,

π i ( σi , σ i ) π i (σi , σ i ) for any σi 2 ∆Si .

Another interpretation: think of σ i as the belief of player i about his opponents,

pi choices of the pure strategies. The pro…le of mixed strategies σ i captures
beliefs over all of the pure strategies that i’s opponents can play.
Rationality requires that a player plays a best response given their beliefs.
A Nash equilibrium requires that these beliefs are correct.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 45 / 50
 Mixed-Strategy Nash Equilibrium

De…nition
A pure strategy si 2 Si is in the support of σi if σi (si ) > 0.

Suppose that in the Nash equilibrium pro…le σ the support of i’s mixed strategy
σi contains more than one pure strategy, i.e., si and si0 are both in the support of
σi . Since σi is a best response against σ i player i cannot do better than to
randomize between more than one of his pure strategies, si and si0 . When is the
player willing to randomize between pure strategies?
Claim Assume σ is a Nash equilibrium,
1 if si and si0 are both in the support of σi , then

π i (si , σ i ) = π i (si0 , σ i ) = π i (σi , σ i )

2 if si is in the support of σi while si00 is not in the support of

σi , then
π i (si , σ i ) π i (si00 , σ i ).

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 46 / 50
 Mixed-Strategy Nash Equilibrium - Example 1
Consider the following game

L R
U 0, 0 3, 5
D 4, 4 0, 3

There are two pure strategy Nash equilibria: (U, R ) and (D, L). However, there is
also mixed strategy Nash equilibrium.
α: the probability that P1 chooses U (1 α is the probability for D)
β: the probability that P2 chooses L (1 β is the probability for R).
Player 1 will mix if

0 β + 3 (1 β ) = 4 β + 0 (1 β ).

Therefore, 8
< α=1 if β < 37 ,
BR1 ( β) = α 2 [0, 1] if β = 37 ,
:
α=0 if β > 37 .
Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 47 / 50
 Mixed-Strategy Nash Equilibrium - Example 1

Player 2 will mix if

0 α + 4 (1 α ) = 5 α + 3 (1 α ).

Therefore, 8
< β=1 if α < 16 ,
BR2 (α) = β 2 [0, 1] if α = 16 ,
:
β=0 if α > 16 .
To …nd a Nash equilibrium we are looking for a pair of choices (α , β ) for which
the two best-responses cross. That is, α = BR1 ( β ) and β = BR2 (α ). This
yields another equilibrium (σ1 , σ2 ) = 1, 5 , 37 , 47 .
6 6

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 48 / 50
 Mixed-Strategy Nash Equilibrium - Example 1
We plot the best responses

a
BR1(b)
1

BR2(a)
1/6

3/7 1 b

There are three intersections of the best-responses (1, 0), (0, 1) and (1/6, 3/7).
The …rst two correspond to the pure strategy equilibria (D,L) and (U,R), the third
one - mixed strategy equilibrium.
Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 49 / 50
 Summary

We discussed simultaneous move games.

We present them using the normal (strategic) form representation.
Analyses of these games involves applying solution concept in order to predict
the outcome of the interaction/game.
We introduced three central solution concepts: dominant strategy
equilibrium, IESDS and Nash equilibrium.

Alex Gershkov (The Hebrew University of Jerusalem) Game theory May 2018 50 / 50