Professional Documents
Culture Documents
Game Theory (18Cs7F2) 7 Semester, CSE Department Faculty: Anitha Sandeep UNIT-3
Game Theory (18Cs7F2) 7 Semester, CSE Department Faculty: Anitha Sandeep UNIT-3
GAME THEORY(18CS7F2)
7th Semester,
CSE department
Faculty: Anitha Sandeep
UNIT-3
Go, change the world
Contents
• Extensive Games
Extensive games with perfect information
Strategies and outcomes
Nash equilibrium
Subgame perfect equilibrium
Finding subgame perfect equilibria of finite horizon games:
Backward induction
Illustrations:The ultimatum game, Stackelberg’s model of duopoly
Go, change the world
with ordinal with vNM with perfect info. With imperfect info.
preferences preferences
Entry Game
• Players – Challenger(player1), Incumbent(player2)
• Terminal histories – (Out), (In, Fight), (In, Accommodate)
• Player function is defined for all proper sub histories.
P(ɸ) = Challenger/player1
P(In) = Incumbent/player2
• Preferences-
For player1: U1(In, Accommodate) > U1(Out) > U1(In, fight)
For player2: U2(Out)> U2(In, Accommodate) > U2(In, fight)
Go, change the world
(2,1) (0,0)
Action set of a player i: Ai(h) = {a | (h, a) is a history}
Finite horizon game: If length of longest terminal history is finite, then
the game is called finite horizon game.
Finite game: If the game has a finite horizon and finitely many terminal
histories, we say the game is finite.
Go, change the world
2. For the following game, write set of players, terminal histories, player
function and preferences:
1
2 C D
1 E F (2,0)
G H (3,1)
(1,2) (0,0)
Ans: Players – 1 and 2
Terminal histories – (D), (C,F), (C,E,G) and (C,E,H)
Player function: P(ɸ) = 1, P(C)=2 , P(C,E) =1
Preferences: Player1 prefers (C,F) to (D) to (C,E,G) to (C,E,H)
Player2 prefers (C,E,G) to (C,F) to (D) = (C,E,H)
Go, change the world
(2,1) (0,0)
Here player2 plays last if player1 enters In. For player2, it is better to
select accommodate rather than fight since 1>0.
Nor for player1, if he enters In, he knows that he will get payoff of 2
and if he goes out he will get a payoff of 1. Therefore player1 selects
In.
Go, change the world
Now player1 knows that player2 will play up whether he play left or
right. So, he has to decide whether he has to play left or right. If he
plays left, he will get payoff of 1 and if he plays right he will get payoff
of 4. So, player1 plays right. So, NE is player1 plays right and player2
plays up. (Right, Up) is solution.
Go, change the world
Strategy:
A player strategy specifies the action the player chooses for every
history after which it is her/his turn to move.
Definition: A strategy of a player i in an extensive game with perfect
information is a function that assigns to each history h after which it is
player i’s turn to move an action in Ai(h){action set after history h, h is
proper sub history of some terminal history}
Ex: 1. For entry game, strategy of player1 are {in, out} and for player2
strategies are {accommodate, fight}
Go, change the world
Ex-2 1 Strategies
x1 x2 Player1: S1= {x1,x2}
2 2 Player2: S2={y1a1, y1a2, y2a1, y2a2}
y1 y2 a1 a2
Ex-3 1 Strategies
x1 x2 Player1: S1={x1z1,x1z2,x2z1,x2z2}
2 2 Player2: S2={y1a1, y1a2, y2a1, y2a2}
y1 y2 a1 a2
1
z1 z2
Go, change the world
Exercise: 156.2 (c) page 157
Represent the following as an extensive form game tree.
Description: Politicians Rosa and Ernesto have to choose either
Berlin(B) or Havana(H) as the location for party. They choose
sequentially. A third person Karl, determines who chooses first. Both
Rosa and Ernesto care only about their actions they choose, not about
who chooses first. Rosa prefers the outcome in which both she and
Ernesto choose B to that in which they both choose H, and prefers this
outcome to either of the ones in which she and Ernesto choose
different actions; she is indifferent between these last two outcomes.
Ernesto’s preferences differ from Rosa’s in that the role of B and H are
reversed. Karl preferences are same as Ernesto’s. Model this as an
extensive game with perfect information.
Go, change the world
Strategy Profile
It is like an action profile. It is a collection of different strategies of
different players.
Strategy profile s=(s1,s2,……sn) where si is the strategy of ith player.
A strategy profile s traces a terminal history. This is called outcome of
strategy profile, represented as O(s).
Ui(O(s)) is the payoff of ith player to the outcome of a strategy profile.
Go, change the world
Nash Equilibrium
A strategy profile s* is a NE if for every player i and every strategy si of
player i, the terminal history O(s*) generated by s* is at least as good
according to i’s preferences as the terminal history O(si, s*-i) generated
by the strategy profile (si, s*-i) in which player i choses si while every
other player j choses sj*.
ie, ui(O(s*)) ≥ ui(O(si, s*-i)) for every strategy si of player i.
Go, change the world
Ex:1
For entry game, 1 S1={In,Out}
In Out S2={Accom,Fight}
2 (1,2)
Accom Fight
(2,1) (0,0)
Players – 1,2
Actions – A1=S1={In, Out}, A2 = S2 ={Accom, Fight} 1/2 Accom Fight
Preferences – are given by game table as below: In (2*, 1'') (0, 0)
• Ex-4
Find NE by converting the game into
Strategic game.
Ans: NE:{ (L,ad), (R,ac) ,(R,bc)}
Robust NE is (R,ac)
P1/P2 ac ad bc bd
If player1 deviates then fight is best action and also player is not
observing fight.
But in (In, Accomm), there is no such problem.
Go, change the world
SPNE/SPE Definition
1.For every player i and every history h after which it is the player i’s
turn to move, ui(Oh(s*))≥ ui(Oh(si,s-i*)) for every strategy si of player i.
Here Oh(s) is the terminal history consisting of h followed by sequence
of actions generated by s after h.
2. SPE induces equilibrium in each possible sub game of the entire
game. For every player i, and every non terminal history h after which i
is to make a move, ui(Oh(s*))≥ ui(Oh(si,s-i*)) for every strategy si of player
i.
NOTE: for any strategy profile s, OΦ(s) = O(s)
Go, change the world
Ans: (OutOB, B)
⇒ a2(-1)+(c+a1-a2).1=-2a2+c+a1=0 (u1,u2)
∴ a2=½ (c +a1)
Now consider sub game of length 2
u1= a1( c+a2-a1) = a1 ( c+½ (c +a1)-a1) = ½a1(3c-a1)
Here player1 plays and he maximizes his profit.
Go, change the world
(½a1(3c-a1))=0
⇒ ½a1(-1) +(3c-a1) ½=0
⇒ a1*=c
∴ a2=½ (c +a1) becomes a2*=c
So a1*=c and a2*=c are equilibrium outcomes.
Go, change the world
2. Firm-Union Bargaining
Description: A union and a firm are bargaining. First, the union presents
a wage demand w ≥ 0. The firm can accept or reject this w. If the firm
chooses accepts, then the firm chooses an amount L ≥ 0 of labor to
hire. The firm’s output is L(100 − L) when it uses L ≤ 50 units of labor,
and 2500 if L > 50. The price of output is 1. If firm rejects demand, no
production takes place(L=0). The firm’s preferences are represented by
its profits. The union’s preferences are represented by the total wage
bill, wL.
a) Model this as an extensive game with perfect information.
b) Find SPE of game.
c) Find payoff to firm and union at SPE
Go, change the world
Firm-Union Bargaining
Soln:
• Players: The union and the firm.
• Terminal histories: The set of sequences (w, L),
where w is the wage demand offered by union
reject
and L is units of labor the firm uses.
• Players’ function: P(∅) is the union,
P(w) is the firm
• Preferences:
Go, change the world
Firm-Union Bargaining
For every w ≥ 0, there is a subgame where
the firm’s payoff depends on w.
Profit has a quadratic part (if L ≤ 50) and
a linear part (if L > 50), and is continuous at L = 50.
reject
We want to find the profit-maximizing choice of L.
The linear part is decreasing in L, so we can ignore
it (its maximum is at L = 50).
Quadratic part is maximized at L = .
Firm’s profit is: (100 − ) − w =
Go, change the world
3. Ultimatum game
Description: Two players want to divide some amount of money c. In
the first round, player1 proposes a division of money x to be given to
player2. In the next round player2 has 2 choices either to accept(Y) or
to reject(N). If player2 accepts, both get amount. If he rejects both get
0.
Model:
Here player1 has ∞ number of actions. And player2 has a weapon to
threaten by rejecting offer.
• Players – 1and 2
• Terminal history- sequences of the form (x,z) where 0 ≤ x ≤ c and
Z=Y/N
• Player function- P(ɸ)=1, P(x)=2
Go, change the world
• Preferences- are represented by payoffs that players are getting at the end
of the game.
U1(x, Y) = c-x U2(x, Y) = x U1(x, N) = 0 = U2(x, N)
The game tree is as below.
To find SPE
Consider sub game of length=1
Here player2 plays and he chooses Y if x>0
If x=0, player2 is indifferent between Y ,N
So, player2 has 2 optimal strategies.
a) Accept any offer with x≥ 0
b) Accept any offer with x>0 and reject if x=0
Go, change the world
Centipede game
• The centipede game is an extensive-form game in game theory in
which two players alternately get a chance to take the larger share of
a slowly increasing money stash. ... “Centipede game” is so-called
because its original version consisted of a 100-move sequence.
• There are 2 players-player1 and player2. Each player starts with ₹1.
They alternate saying ‘stop’ or ‘continue’ starting with player1. when
a player says ‘continue’, ₹1 is taken by refree from his pile and ₹2 are
put in other players pile. As soon as either player says ‘stop’ game is
terminated and each player receives money currently in her pile.
Alternatively the game stops if both players piles reach ₹100. Model
game and find SPE.
Go, change the world
Soln :
The game tree is as below:
SPE : Unique SPE saying S at each sub game for both players.
Terminal history reached by SPE is (S). Equilibrium payoff is (1,1)
Go, change the world