GAME THEORY
ANSWERSTOEXERCISES—PARTI
(FROMTHETEXTBOOKOFC.HOLT)
GabrieleCamera
EconomicScienceInstitute,ChapmanUniversity
1
PROBLEM 1
What is the Nash Equilibrium for the Traveler’s Dilemma game where choices
must be in the set A = {−50, −49, · · · , −1, 0, 1, · · · , 49, 50} and there is a
penalty/reward rate R = 5?
2
ANSWER TO PROBLEM 1
Action set: A1 = A2 = {−50, −49, · · · , −1, 0, 1, · · · , 49, 50}
Payoff matrix:
-50 -49 -48 ... 49 50
-50 (-50,-50) (-45,-55) (-45,-55) ... (-45,-55) (-45,-55)
-49 (-55,-45) (-49,-49) (-44,-54) ... (-44,-54) (-44,-54)
-48 (-55,-45) (-54,-44) (-48,-48) ... (-43,-53) (-43,-53)
.. .. .. .. ... .. ..
49 (-55,-45) (-54,-44) (-53,-43) ... (49,49) (54,44)
50 (-55,-45) (-54,-44) (-53,-43) ... (44,54) (50,50)
(−50, −50) is the unique Nash Equilibrium
3
NASH EQUILIBRIUM: A PROOF
Step 1: Existence (direct proof)
Conjecture(-50,-50) is a NE. This conjecture follows from what we saw in class.
π1(−50, −50) = −50 > π1(a1, −50) = −55 for all a1 > −50.
π2(−50, −50) = −50 > π2(−50, a2) = −55 for all a2 > −50.
Hence, (−50, −50) is a NE
4
NASH EQUILIBRIUM: A PROOF
Step 2: Uniqueness (indirect proof)
By contradiction, suppose that pâ1, â2q ‰ p´50, ´50q is a NE
Case 1: â1 ă â2 (the case â2 ă â1 is equivalent)
π1pâ1, â2q “ â1 ` R
π2pâ1, â2q “ aˆ1 ´ R
But player 2 can do better, choosing a2 “ â1:
π2 pâ1, â1q “ â1 ą π2pâ1, â2q “ â1 ´ R (contradiction)
5
NASH EQUILIBRIUM: A PROOF
Case 2: â1 = â2
Player 1’s payoff is
π1(â1, â1) = â1
But player 1 can do better, by cutting â1 by 1:
π1(â1 − 1, â1) = â1 − 1 + R = â1 + 4 > â1 = π(â1, â1) (contradiction)
6
PROBLEM 2
Consider a Traveler’s Dilemma with choices in A = {80, 81, . . . 200}, and R = 10.
Two players play four rounds with claims (80, 195), (98, 117), (96, 117), (130, 100)
respectively. Calculate the earnings of the first player. What would his earnings
be if the first player chose 200 in each round?
7
ANSWER TO PROBLEM 2
Action set for each player: A = {80, 81, . . . , 199, 200}
Penalty/reward rate: R = 10
Player 1’s payoff in each round: π1(a1, a2) = min(a1, a2) + θ1(a1, a2) × R, where
⎧
⎪
⎨ 1 if a1 < a2
θ1(a1, a2) := 0 if a1 = a2
⎪
⎩ −1 if a > a
1 2
So, the payoff of player 1 is
π1 = (80 + 10) + (98 + 10) + (96 + 10) + (100 − 10) = 394
8
TAKING MORE RISK
Suppose player 1 chose 200 in each round. His earnings would be
π̄1 “ p195 ´ 10q ` p117 ´ 10q ` p117 ´ 10q ` p100 ´ 10q “ 489
which is more profitable than π1
9
PROBLEM 3
Consider a Traveler’s Dilemma with R=$5 and N players and choices in the set
A = {80, 81, . . . 200}. If all choices are not equal, there is a $5 penalty if one’s
choice is not the lowest and a $5 reward rate for the person with the lowest choice.
Speculate on what the effect on average claims of increasing N from 2 to 4 players
is in a setup with repeated random matchings
10
ANSWER
Case N “ 2: unique NE “ p80, 80q
Since R “ 5 is small compared with any claim in t80, 81, . . . , 199, 200u, individuals
might be tempted to choose high claims
In a repeated game, players may learn what to expect.
Consequently, they might coordinate on higher claims over time as in the experiment
by Capra et al.
11
CHOICE FREQUENCIES: (light bars=HIGH, dark bars=LOW)
12
MULTIPLE PLAYERS
Case N “ 4: Players are matched together
Action profile a “ pa1, a2, a3, a4q with ai P t80, 81, . . . , 200u
Player i’s payoff: πipaq “ minpaq ` θipaq ˆ R
$
’
&1 if ai “ minpaq ă aj for some j ‰ i
θipaq :“
’
0 if a1 “ a2 “ a3 “ a4
%
´1 if ai ą minpaq
Note: rewards/penalties are not zero-sum (more than one player may pay R, and
more than one player may receive R)
13
NASH EQUILIBRIUM
Example:
Suppose that a1 is the lowest choice with ai > a1 for each i = 2, 3, 4, then
• Player 1 gets a1 + R
• Players 2, 3 and 4 get a1 − R
By symmetry, (80,80,80,80) is the unique NE (see Problem 2)
Again, individuals may increase their earnings from round to round by coordinating
on higher choices
14
THE IMPACT OF MORE THAN 2 PLAYERS
Anticipating others’ intentions is more difficult since there is more heterogeneity
On average, one might expect:
• Choices might be smaller than in 2-player case (more competition)
• More difficult to coordinate on upper bound choices (more heterogeneity)
• If there is convergence to NE, then faster convergence (more competition)
15
PROBLEM 4
Consider a Matching Pennies Game with the following payoff matrix
Left Right
Top 30, 40 40, 80
Bottom 40, 80 80, 40
(i) Does there exist a Nash Equilibrium in pure strategies?
(ii) Does there exist a Nash Equilibrium in mixed strategies?
16
ANSWER TO PROBLEM 4 (QUESTION I): CASE 1
Row player chooses Top and Column player chooses Right
Left Right
Top 30, 40 40, 80
Bottom 40, 80 80, 40
{Top,Right} is not a Pure NE, since Row player can do better by choosing Bottom
17
CASE 2
Row player chooses Bottom and Column player chooses Right
Left Right
Top 30, 40 40, 80
Bottom 40, 80 80, 40
{Bottom,Right} is not a Pure NE, since Column player can do better by choosing
Left
18
CASE 3
Row player chooses Top and Column player chooses Left
Left Right
Top 30, 40 40, 80
Bottom 40, 80 80, 40
{Top,Left} is not a Pure NE, since either Column player can do better by choosing
Right or Row player can do better by choosing Bottom
19
CASE 4
Row player chooses Bottom and Column player chooses Left
Left Right
Top 30, 40 40, 80
Bottom 40, 80 80, 40
{Bottom,Left} is a Pure NE, since both players cannot do better by changing
action
20
ANSWER TO PROBLEM 4 (QUESTION II)
Let {q, 1 − q} be the set of probability distribution over {Top, Bottom}, and
{p, 1 − p} be the set of probability distribution over {Left, Right}
Row player: choose Top with probability q to make column player indifferent
column player chooses Left column player chooses Right
1
q40 + (1 − q)80 = q80 + (1 − q)40 ⇒ q= ,
2
Column player: choose Left with probability p to make row player indifferent:
row player chooses Top row player chooses Bottom
4
30p + 40(1 − p) = 40p + 80(1 − p) ⇒ p= >1
3
21
MIXED STRATEGY NASH EQUILIBRIUM
Every finite strategic game has a mixed strategy Nash Equilibrium
Here, the mixed strategy Nash Equilibrium is the profile ({0, 1}, {1, 0})
22
