Professional Documents
Culture Documents
Definition 1.1. Game theory is a systematic study of strategic interactions among rational
individuals.
1.1 Assumptions
1. Rationality The most important, and maybe one of the most controversial, assumption
of game theory which brings about this discipline is that individuals are rational. We
assume that individuals are rational.
Definition 1.2. An individual is rational if she has well-defined objectives (or prefer-
ences) over the set of possible outcomes and she implements the best available strategy
to pursue them.
Rationality implies that individuals know the strategies available to each individual, have
complete and consistent preferences over possible outcomes, and they are aware of those
preferences. Furthermore, they can determine the best strategy for themselves and flaw-
lessly implement it. Also we can say that, an individual is rational if he/she has well-
defined preferences over the set of possible outcomes and he/she implements flawlessly
the best available strategy to pursue them to achieve the maximum outcome.
1
late. Now, if one of the students openly says that “ instructor told me that he (instructor)
will be late ” then each student knows that each student knows that instructor will be
late, then each student knows that each student knows that each student knows that the
instructor will be late, and so on, ad infinitum. The announcement made the mutual
known fact common knowledge among the students.
Definition 1.3. A fact “A” is said to be a common knowledge among players of a game
if all the players know “A”, all the players know that all the players know “A”, all the
players know that all the players know that all players know “A” and so on.
3. No outside communication.
4. Full knowledge of the game: Pay-offs and actions are observable and known by all.
• In a simultaneous game, each player has only one move, and all moves are made simul-
taneously.
• In a sequential game, no two players move at the same time, and players may have to
move several times.
• There are games that are neither simultaneous nor sequential. Does the game have ran-
dom moves? Games may contain random events that influence its outcome. They are
called random moves.
• Do players have complete information? This means that all players know the structure
of the game—the order in which the players move, all possible moves in each position,
and the payoffs for all outcomes.
• Real-world games usually do not have complete information. In our games we assume
complete information in most cases, since games of incomplete information are more
difficult to analyze.
• Is the game zero-sum? Zero-sum games have the property that the sum of the payoffs to
the players equals zero. A player can have a positive payoff only if another has a negative
payoff. Poker and chess are examples of zero-sum games. Real-world games are rarely
zero-sum.
2
• A cooperative game is one where the results of the negotiations can be put into a contract
and be enforced. There must also be a way of distributing the payoff among the members
of the coalition.
We will mainly study -
• Static games of Complete Information (Simultaneous-move games)
– Normal-form games and Nash equilibrium.
– Pure and mixed strategy equilibrium.
– Cournot Model of duopoly.
– Bertrand model of duopoly.
– Hotteling’s electoral competition model.
• Dynamic games of Complete Information (Sequential-move games)
– Complete and perfect information
– Backward Induction method.
– Stackelberg model of duopoly.
– Repeated Games
– two-stage repeated games.
* Extensive-form games.
* Subgame perfect Nash equilibrium(SPNE).
Information
Complete Incomplete
Strategic-form game Bayesian Game
Simultaneous
Moves (Static game) (Static game)
Extensive-form game Extensive-form game
Sequential
(Dynamic game) (Dynamic game)
3
2 Static games of complete Information
Here first the players simultaneously choose actions, then the players receive payoffs that de-
pend on the combination of actions just chosen. Here within the class of such static (simultaneous-
move) games, we will be focusing on the complete information. That is each player’s payoff
function is common knowledge among all the players.
1) The players in the game. Let there are ‘n’ players i.e. i = 1, 2, ...n.
3) The payoff received by each player for each combination of strategies that could be chosen
by the players. We generally write these payoffs in a matrix form, which is called Payoff
Matrix, where in each cell we write the payoffs in a co-ordinate form i.e. (x,y) form and x
- the first entity is the row player’s payoff from his/ her corresponding strategies and y- the
later entity is the payoff is the column player’s payoff from his/ her corresponding strategies.
Definition 2.1. A Nash equilibrium is a set of strategies, one for each player, such that no
player has incentive to change his or her strategy given what the other players are doing.
4
2.4 Examples
2.4.1 Strict Dominance
Consider a hypothetical game, which has the following payoff matrix:
P2
Launch Don’t launch
P1 Launch (40,40) (100,-50)
Don’t launch (-50,100) (-40,-40)
Let’s see how players play-
P2’s decision
P2
Launch Don’t launch
P1 Launch (,40) >(,-50)
Don’t launch (,100) >(,-40)
We can see from the above pay-of matrix that, if P2 chooses “Launch” then P2 will get
either 40 or 100 and if P2 chooses “Don’t launch” then P2 will get either -50 or 40. So at any
circumstances-irrespective of what P1 is choosing P2 will always prefer “Launch” as (40 > −50
and 100 > −40).
As a rational player choosing strategy “Launch” is always preferable to P2 ⇒ “Launch” is
strictly dominating strategy. That is “Don’t launch” is strictly dominated by “Launch”.
P1’s decision
P2
Launch Don’t launch
P1 Launch (40,) (100,)
Don’t launch (-50,) (-40,)
We can see from the above pay-of matrix that, if P1 chooses “Launch” then P1 will get
either 40 or 100 and if P1 chooses “Don’t launch” then P1 will get either -50 or 40. So at any
circumstances-irrespective of what P2 is choosing P1 will always prefer “Launch” as (40 > −50
and 100 > −40).
As a rational player choosing strategy “Launch” is always preferable to P1 ⇒ “Launch” is
strictly dominating strategy. That is “Don’t launch” is strictly dominated by “Launch”.
One way we can see that, being P1 and P2 are rational, they will never play “Don’t launch”
(as it is strictly dominated by “Launch) irrespective of what other players are doing. And there-
fore the Nash equilibrium is (Launch, Launch)=(40,40).
5
2.4.2 Weak Dominance
Consider the following game
P2
L R
P1 U (7,2) (5.4)
D (7,2) (2,2)
• For P1 neither “U” nor “D” is strictly dominated as: “U” is better than “D” if P2 plays
“L”(1 > 0) but “D” is better than “U” if P2 plays “R” (2 > 0).
• For P2 “R” is strictly dominated by “M” (2 > 1 and 1 > 0) =⇒ P2 as a rational player
will never play “R”.
• P1 knows that P2 is rational then P1 can eliminate “R” from P2’s strategy space. That is,
if P1 knows that P2 is rational then P1 can play the following game:
P2
L M
P1 U (1,0) (1,2)
D (0,3) (0,1)
• In this truncated game, “D” is now strictly dominated by “U” (1 > 0, 1 > 0) for P1.
• Thus, if P2 knows that P1 is rational and P2 knows that P1 knows that P2 is rational,
then P2 can eliminate “D” from P1’s strategy space. Then we have the following game
P2
L M
P1 U (1,0) (1,2)
• P2 will never play as a rational player =⇒ we have only (U,M), which is the outcome
of the game.
6
2.4.4 Prisoner’s Dilemma- A classic game
Two suspects are arrested and charged with a crime. The police department lacks sufficient ev-
idence to convict them, unless at least one confesses. the police hold them in separate cells and
explain the consequences that will follow from the action they could take. If neither confess
then both will be convicted of a minor offense and sentenced to 1 year in jail. If both confess
then both will be sentenced to 6 years in jail. If one confesses and other doesn’t then the con-
fessor will be released immediately but other will go to jail for 9 years.
P2’s decision
(So, for P2 “NC” is strictly dominated by “C”.[one can verify using strict dominance method])
P1’s decision
(So, for P1 “NC” is strictly dominated by “C”.[one can verify using strict dominance method])
To find the NE, check for which strategies players are mutually agreeing to each other that is
find the Mirror Image. In this example we can see that, when P1 is choosing “C” P2 chooses
“C” and when P2 is choosing “C” P1 chooses “C”. That is players are mutually agreeing for
“C” and “C”. We write the NE as - (C, C)=(-6,-6).
7
In terms of mirror image we can see that, P1C → P2C has a mirror image - P2C → P1C and
vice-versa. But, P1NC → P2C has no mirror image (i.e., P2C → P1NC doesn’t exist). Simi-
larly, P2NC → P1C has no mirror image (i.e., P1C → P2NC doesn’t exist).
We can see an interesting fact from the result is that, if both of them would have chosen “Not
Confess”, then they would have ended up in better position (instead of going to jail for 6 years,
they would go to jail for 1 year. which is a Pareto Optimal situation). But, it didn’t happen.
why? (will come later !!).
Girl’s decision
‘ Boy’s decision
8
• If Girl chooses “Movie” then-
Using the concept of mutually agreement or mirror image we get two Nash equilibria- (Foot-
ball, Football)=(F,F)=(2,1) and (Movie, Movie)=(M,M)=(1,2). This is a classic example
where we can see that more than one Nash equilibrium exists.
– If both are “Heads”(H,H) then, Leonardo wins 1 penny from Kate i.e. +1 to Leonardo
and -1 to Kate (as Kate is losing 1 penny).
– If both are “Tails”(T,T) then we have same scenario for Leonardo.
– If one is “head” and other is “tail” (H,T) then, Kate wins 1 penny from Leonardo
i.e. +1 to Kate and -1 to Leonardo (as Leonardo losing 1 penny).
– Same scenario will appear to Kate in case of (T,H) option.
– Look, here the pay-off matrix is constructed based on the players gaining or loosing
the number of penny not how many pennies all total they are having.
– So, the payoff matrix is-
Kate
H T
Leonardo H (+1,-1) (-1,+1)
T (-1,+1) (+1,-1)
9
• If Kate chooses “H” then-
We can see that, there is no situation where they are mutually agreeing to each other also there
DO NOT EXIST any mirror image for any set of strategies i.e. there is no Nash equilibrium. So,
can we say that there is no Nash equilibrium ? - NO!!. Whatever we have seen in sections 2.4.1,
2.4.2, 2.4.3, 2.4.4 and 2.4.5 respective strategies are called pure strategies and those Nash
equilibria are called “Pure-Strategy Nash Equilibrium” as those equilibria can be written in
terms of set of strategies i.e. (C,C), (M,M), (Launch, Launch) etc.. And here comes the concept
of Mixed-Strategy Nash equilibrium. So, in this Matching Pennies game we can not say that
there is no Nash equilibrium; we will say that there is no Pure-strategy Nash equilibrium. But
a game can have both Pure and Mixed strategy equilibria.
Definition 2.2. Best response is the strategy which produces the most favorable outcome for a
player, taking other player’s strategies are given
• P2 believes that P1 will play “T” with the probability ‘p’. Then automatically P1 will
play “B” with probability ‘1-p’ (as the total probability is 1)
• similarly P1 believes that, P2 will play “L” with probability ‘q’ and “R” with probability
‘1-q’.
• Using these assigned probabilities, both the players will calculate their expected payoff
for their respective strategies.
P1’s expected payoff: P1 will consider the probabilities he/she assigned (from his/ her belief)
to P2’s strategies and his/her own payoffs getting from his/her respective strategies to calculate
the expected payoff from his/ her strategies.
10
• Expected payoff for P1 from playing B is : EP1 (B) = −5 ∗ q + (−3) ∗ (1 − q)
P2’s expected payoff: P2 will also consider the probabilities he/she assigned (from his/ her
belief) to P1’s strategies and his/her own payoffs getting from his/her respective strategies to
calculate the expected payoff from his/ her strategies.
Now we can see that, if EP1 (T ) > EP1 (B), then as a rational player P1 will obviously choose
“T” and if EP1 (T ) < EP1 (B), then P1 will choose “B”, which are nothing but their pure strate-
gies. Now to be indifferent between these two strategies (or you can consider that, as they don’t
have any pure strategy equilibrium) EP1 (T ) = EP1 (B) must hold. And EP1 (T ) = EP1 (B)q = 49 .
Using the similar argument and P2’s expected payoffs, we will get p = 83 .
Therefore, the Best response function for P1 given P2’s probability ‘q’, BR1 (q) is:
p = 0
if q > 83
BR1 (q) = p = 1 if q < 38
p ∈ [0, 1] if q = 38
11
Therefore, the Best response function for P2 given P1’s probability ‘p’, BR2 (p) is:
q = 0
if p < 94
BR2 (p) = q = 1 if q > 94
q ∈ [0, 1] if p = 49
BR1 (q)
3
8
0 1 p
BR2 (p)
3
8
0 4 1 p
9
If we plot these two best response curves together then the intersection point is the Mixed-
strategy Nash equilibrium. (What we got mathematically earlier)
12
q
1 BR2 (p)
NE=( 49 , 38 )
q∗ = 3
8
BR1 (q)
0 p
p∗ = 4
9
1
So, the Mixed strategy Nash equilibrium is (p∗ , q∗ ) = ( 94 , 38 ). And this particular game has no
pure strategy equilibrium. You can find mixed strategy equilibrium for a game even if it has
pure strategy equilibrium.
NOTE: We have to solve the “Matching Pennies game” in this way. Try to find the mixed
strategy equilibrium for the matching pennies game(see section 2.4.6) and also for the battle of
sexes game (see section 2.4.5).
13