3

Eliminating the Impossible:

Solving a Game When Rationality
Is Common Knowledge

1. Derive the strategic form of the Mugging game in Figure 2.9 of Chapter 2 (page 30),
and determine whether any strategies are either strictly dominated or weakly
dominated.

ANSWER: The strategic form game is shown in the figure below. As the mugger
has one information set (the initial node) and three actions, he has three strat-
egies. Simon has two information sets; one is associated with the mugger show-
ing a gun (which is a singleton) and one with the mugger not showing a gun
(which comprises two nodes; one in which the mugger has a gun and one in
which he does not). As Simon has two feasible actions at each information set,
he has four feasible strategies. Let a strategy be represented by x/y, where
Simon chooses action x at the information set in which a gun is shown
and action y when a gun is not shown. x and y can either be resist (R) or do not
resist (DNR).

Simon
R/R R/DNR DNR/R DNR/DNR
Use gun and show 3,2 3,2 4,5 4,5
Mugger Use gun and hide 3,2 5,4 3,2 5,4
Do not use gun 2,6 6,3 2,6 6,3

For the mugger, none of his strategies is either strictly or weakly dominated.
For Simon, none of his strategies is strictly dominated, but he does have some
weakly dominated strategies: DNR/R weakly dominates R/R, and DNR/DNR weakly
dominates R/DNR. If the mugger chooses to use a gun and show it, then Simon
should definitely choose DNR. Weak dominance eliminates those strategies, which
has him resist when he sees a gun.

2. In the Dr. Seuss story “The Zax,” a North-Going Zax and a South-Going Zax on their
treks soon find themselves facing each other. Each Zax must decide whether to continue
in their current direction or move to the side so that the other may pass. As the story
reveals, neither of them moves and that stalemate perpetuates for many years. Write
down a strategic form game of this situation.

ANSWER: Each Zax has two strategies: remain (in its present position) and move
(to the side). For each, remain is a dominant strategy, so one possible strategic
form is that shown in the figure below.

South-Going Zax
Remain Move
Remain 1,1 3,0
North-Going Zax
Move 0,3 2,2

3-1
3-2 SOLUTIONS MANUAL CHAPTER 3: ELIMINATING THE IMPOSSIBLE: SOLVING A GAME WHEN RATIONALITY IS COMMON KNOWLEDGE

3. For the Team-project game, suppose a jock is matched up with a sorority girl, as shown.

Sorority girl
Low Moderate High
Low 3,0 4,1 5,2
Jock Moderate 2,2 3,4 4,3
High 1,6 2,5 3,4

a. Assume that both are rational and that the jock knows that the sorority girl is ratio-
nal. What happens?

ANSWER: For the jock, both low and moderate strictly dominate high, and low
strictly dominates moderate. None of the sorority girl’s strategies is strictly domi-
nated, however. After eliminating the strictly dominated strategies, the reduced
game is as shown in the figure below. As we don’t know what the sorority girl
believes about the jock, we cannot go any further. The answer is then that the jock
chooses low and the sorority girl chooses low, moderate, or high.

Sorority girl
Low Moderate High
Jock Low 3,0 4,1 5,2

b. Assume that both are rational and that the sorority girl knows that the jock is ratio-
nal. What happens?

ANSWER: With the game shown in the figure above, the sorority girl now knows
the jock is rational and thus will play low. Hence, she should choose high as it
strictly dominates both low and moderate. Hence, the jock chooses low effort and
the sorority girl chooses high effort.

4. Consider the strategic form game shown.

Player 2
x y z
a 1,3 1,1 0,2
Player 1 b 3,1 2,2 1,0
c 0,2 1,2 3,0

a. Assume that both players are rational. What happens?

ANSWER: For player 1, a is strictly dominated by b. Neither b nor c is strictly

dominated. For player 2, z is strictly dominated by x. Player 1 plays either b or c
and player 2 plays either x or y.

b. Assume that both players are rational and that each believes that the other is rational.
What happens?

ANSWER: By the assumption, we can go two rounds of the iterative deletion of

strictly dominated strategies (IDSDS). After eliminating the strictly dominated
strategies, the game is as shown in the figure below. Now b strictly dominates c for
player 1. Neither of player 2’s strategies is strictly dominated. Thus, player 1
chooses b and player 2 chooses either x or y.

Player 2
x y
b 3,1 2,2
Player 1
c 0,2 1,2
 SOLUTIONS MANUAL CHAPTER 3: ELIMINATING THE IMPOSSIBLE: SOLVING A GAME WHEN RATIONALITY IS COMMON KNOWLEDGE 3-3

c. Find the strategies that survive the IDSDS.

ANSWER: After the first two rounds, the game is as shown in the figure below.
Now y strictly dominates x for player 2. Thus, player 1 chooses b and player 2
chooses y.

Player 2
x y
Player 1 b 3,1 2,2

5. For the strategic form game shown, derive the strategies that survive the IDSDS.

Player 2
x y z
a 5,2 3,4 2,1
b 4,4 3,2 3,3
Player 1
c 3,5 4,4 0,4
d 2,3 1,5 3,0

ANSWER: For player 1, no strategy is strictly dominated. Between a and b, a is

better when player 2 uses x, while b is better when player 2 uses z. Hence, a does
not strictly dominate b, and b does not strictly dominate a. With a and c, a is better
when player 2 uses x, while c is better when player 2 uses y. With a and d, a is bet-
ter when player 2 uses x, while d is better when player 2 uses z. With b and c, b is
better when player 2 uses x, while c is better when player 2 uses y. With b and d, b
is better when player 2 uses x, while d is just as good when player 2 uses z. (Note
that b weakly dominates d but does not strictly dominate it.) Finally, with c and d,
c is better when player 2 uses x, while d is better when player 2 uses z. Hence, none
of player 1’s strategies is strictly dominated.
Turning to player 2, x strictly dominates z as it yields a strictly higher payoff for
all strategies of player 1. We can then eliminate z. With x and y, x is better when
player 1 uses b, but y is better when player 1 uses a. After the first round of the
iterative deletion of strictly dominated strategies (IDSDS), we are left with {a,b,c,d}
and {x,y}. The remaining game is then as shown in the figure below.

Player 2
x y
a 5,2 3,4
b 4,4 3,2
Player 1
c 3,5 4,4
d 2,3 1,5

Now d is strictly dominated by a (as well as by b and c). One can show that no
other strategies are strictly dominated. The surviving game is then as shown in the
figure below.

Player 2
x y
a 5,2 3,4
Player 1 b 4,4 3,2
c 3,5 4,4

No further strategies can be eliminated. All we can conclude is that player 1 will
play either a, b, or c and player 2 will play either x or y.
3-4 SOLUTIONS MANUAL CHAPTER 3: ELIMINATING THE IMPOSSIBLE: SOLVING A GAME WHEN RATIONALITY IS COMMON KNOWLEDGE

6. Two Celtic clans—the Garbh Clan and the Conchubhair Clan—are set to battle.
(Pronounce them as you’d like; I don’t speak Gaelic.) According to tradition, the
leader of each clan selects one warrior and the two warriors chosen engage in a fight
to the death, the winner determining which will be the dominant clan. The three top
warriors for Garbh are Bevan (which is Gaelic for “youthful warrior”), Cathal (strong
in battle), and Duer (heroic). For Conchubhair, it is Fagan (fiery one), Guy (sensible),
and Neal (champion). The leaders of the two clans know the following information
about their warriors, and each knows that the other leader knows it, and furthermore,
each leader knows that the other leader knows that the other leader knows it, and so
forth (in other words, the game is common knowledge): Bevan is superior to Cathal
against Guy and Neal, but Cathal is superior to Bevan against Fagan. Cathal is
superior to Duer against Fagan, Guy, and Neale. Against Bevan, Guy is best. Against
Cathal, Neal is best. Against Duer, Fagan is best. Against Bevan, Fagan is better than
Neal. Against Cathal, Guy is better than Fagan. Against Duer, Guy and Neal are com-
parable. Assuming that each leader cares only about winning the battle, what can you
say about who will be chosen to fight?

ANSWER: Let us use the iterative deletion of strictly dominated strategies to

determine who will battle. Since Cathal is superior to Duer against all three of
the warriors from the Conchubhair Clan, we can then eliminate Duer. Next,
note that Bevan is best against Guy and Neal (since Bevan is superior to Cathal
and Cathal is superior to Duer) and Cathal is best against Fagan. Therefore,
neither Bevan nor Cathal is strictly dominated. Thus, in the first round, we can
delete only Duer from the Garbh Clan. Since each of the warriors in the
Conchubhair Clan is best against someone, none of them is strictly dominated.
After one round of the IDSDS, we are left with Bevan and Cathal for the Garbh
Clan and Fagan, Guy, and Neal for the Conchubhair Clan. Against Bevan and
Cathal, Guy is superior to Fagan, so Fagan can be eliminated. Since Guy is bet-
ter than Neal against Bevan and Neal is better than Guy against Cathal, neither
Guy nor Neal can be eliminated. Note that we cannot eliminate anyone from
the Garbh Clan in round 2 because the Conchubhair Clan has the same three
warriors as in round 1. At the end of the second round, we are left with Bevan
and Cathal for the Garbh Clan and Guy and Neal for the Conchubhair Clan.
Against Guy and Neal, Bevan is superior to Cathal, so we can delete Cathal.
Since the warriors for the Garbh Clan in round 3 are the same as in round 2,
we cannot eliminate anyone from the Conchubhair Clan. After three rounds,
we are left with Bevan for the Garbh Clan and Guy and Neal for the Conchubhair
Clan. Since Guy is superior to Neal against Bevan, Neal is eliminated. We con-
clude that the leader of the Garbh Clan will choose Bevan and the leader of the
Conchubhair Clan will select Guy. As an alternative approach to finding this
pair, we could have used the information given to construct a strategic form
game. One such game, which is compatible with the information given, is shown
in the figure below.

Conchubhair Clan
Fagan Guy Neal
Bevan 2,1 1,2 2,0
Garbh Clan Cathal 3,0 0,1 1,2
Duer 1,2 –1,0 0,0

7. Consider the two-player strategic form game depicted.

Player 2
w x y z
a 1,2 0,5 2,2 4,0
b 1,3 5,2 5,3 2,0
Player 1
c 2,3 4,0 3,3 6,2
d 3,4 2,1 4,0 7,5
 SOLUTIONS MANUAL CHAPTER 3: ELIMINATING THE IMPOSSIBLE: SOLVING A GAME WHEN RATIONALITY IS COMMON KNOWLEDGE 3-5

a. Derive the strategies that survive the IDSDS.

ANSWER: Examining player 1’s strategies, first note that d is optimal for player 1
when player 2 is expected to use w. Thus, d cannot be strictly dominated since to
be strictly dominated requires that there is another strategy that yields a higher
payoff for all strategies of the other player. Since b is best for player 1 when 2 uses
x, then b is not strictly dominated either. c is not strictly dominated since it yields
a higher payoff than a and b when player 2 uses w and a higher payoff than d when
player 2 uses x. a is strictly dominated by c (and also by d) in that c yields a higher
payoff than a for any strategy of player 2. We then find that the set of strategies for
player 1 that survive the first round of the iterative deletion of strictly dominated
strategies (IDSDS) is {b,c,d}.
Turning to player 2’s strategies, x is best for player 2 when player 1 uses a, w
and y are both optimal when player 1 uses b, and z is best when player 1 uses d.
Thus, none of player 2’s strategies is strictly dominated. After one round of IDSDS,
the game is as shown in the figure below.

Player 2
w x y z
b 1,3 5,2 5,3 2,0
Player 1 c 2,3 4,0 3,3 6,2
d 3,4 2,1 4,0 7,5

Since we failed to eliminate any of player 2’s strategies in the first round, we are
unable to eliminate any of player 1’s strategies in the second round (if you are
unconvinced by this statement, check for yourself). Turning to player 2, w and y
are best when player 1 uses b, and z is best when player 1 uses d. Thus, w, y, and z
are not strictly dominated. However, x is strictly dominated by w.
After two rounds of IDSDS, the game is as shown in the figure below.

Player 2
w y z
b 1,3 5,3 2,0
Player 1 c 2,3 3,3 6,2
d 3,4 4,0 7,5

For player 1, d is best when player 2 uses z and b is best when player 2 uses y.
However, c is strictly dominated by d. Since none of player 1’s strategies was
eliminated in the second round, none of player 2’s strategies can be eliminated in
the third round.
After three rounds of IDSDS, the game is as shown in the figure below.

Player 2
w y z
b 1,3 5,3 2,0
Player 1
d 3,4 4,0 7,5

Since none of player 2’s strategies was eliminated in the third round, none of
player 1’s strategies can be eliminated in the fourth round. Since w and y are both
optimal when player 1 uses b, and z is optimal when player 1 uses d, then we cannot
eliminate any of player 2’s strategies. Given that no strategies are eliminated in this
round, no strategies can be eliminated in any further rounds. We conclude that
{b,d} for player 1 and {w,y,z} for player 2 are the strategies that survive the IDSDS.

b. Derive the strategies that survive the iterative deletion of weakly dominated strate-
gies. (The procedure works the same as the IDSDS, except that you eliminate all
weakly dominated strategies at each stage.)
3-6 SOLUTIONS MANUAL CHAPTER 3: ELIMINATING THE IMPOSSIBLE: SOLVING A GAME WHEN RATIONALITY IS COMMON KNOWLEDGE

ANSWER: Examining player 1’s strategies, first note that d is the unique optimal
strategy for player 1 when player 2 is expected to use w. Thus, d cannot be weakly
dominated since to be weakly dominated requires that there is another strategy
that yields at least as high a payoff for all strategies of the other player and a
strictly higher payoff for some strategies of the other player. Since b is the unique
optimal strategy for player 1 when player 2 uses x, then b is not weakly dominated
either. c is not weakly dominated since it yields a strictly higher payoff than a and
b when player 2 uses w and a strictly higher payoff than d when player 2 uses x. a
is weakly (and strictly) dominated by c (and also by d). We then find that the set
of strategies for player 1 which survive the first round of the iterative deletion of
weakly dominated strategies (IDSDS) is {b,c,d}.
Turning to player 2’s strategies, x is the unique optimal strategy for player 2 when
player 1 uses a, and z is the unique optimal strategy when player 1 uses d. w weakly
dominates y since it yields an identical payoff when player 1 uses a, b, or c and a
strictly higher payoff when player 1 uses d. w is not weakly dominated by x since it
yields a strictly higher payoff when player 1 uses b, and it is not weakly dominated
by z since it yields a strictly higher payoff when player 1 uses a. Therefore, the set of
strategies for player 1 that survive the first round of the IDSDS is {w,x,z}.
After one round of IDSDS, the game is as shown in the figure below.
Player 2
w x z
b 1,3 5,2 2,0
Player 1 c 2,3 4,0 6,2
d 3,4 2,1 7,5

d is not weakly dominated since it is the unique optimal strategy when player 2
uses w, and b is not weakly dominated since it is the unique optimal strategy when
player 2 uses x. c is not weakly dominated by b since it yields a strictly higher
payoff when player 2 uses w, and it is not weakly dominated by d since it yields a
strictly higher payoff when player 2 uses x. None of player 1’s strategies is elimi-
nated in the second round of IDSDS.
Turning to player 2, w is the unique optimal strategy when player 1 uses b and
z is the unique optimal strategy when player 1 uses d. x is strictly and thus weakly
dominated by w.
After two rounds of IDSDS, the game is as shown in the figure below.
Player 2
w z
b 1,3 2,0
Player 1 c 2,3 6,2
d 3,4 7,5

For player 1, d strictly and therefore weakly dominates both b and c. Since none
of player 1’s strategies was eliminated in the second round, none of player 2’s
strategies can be eliminated in the third round.
After three rounds of IDSDS, the game is as shown in the figure below.
Player 2
w z
Player 1 d 3,4 7,5

There is nothing left for player 1 to do. Since z yields a strictly higher payoff
than w when player 1 uses d, then z strictly and therefore weakly dominates w.

8. Consider the three-player game shown. Player 1 selects a row, either a1, b1 or c1. Player 2
selects a column, either a2 or b2. Player 3 selects a matrix, either a3 or b3. The first
number in a cell is player 1’s payoff, the second number is player 2’s payoff, and the
last number is player 3’s payoff. Derive the strategies that survive the IDSDS.
 SOLUTIONS MANUAL CHAPTER 3: ELIMINATING THE IMPOSSIBLE: SOLVING A GAME WHEN RATIONALITY IS COMMON KNOWLEDGE 3-7

a3 b3

a2 b2 a2 b2
a1 3,1,0 2,3,1 a1 3,1,1 1,3,2
b1 0,3,1 1,1,0 b1 2,0,2 2,2,1
c1 1,0,2 1,2,1 c1 1,1,1 0,2,0

ANSWER: To begin, consider player 1. Neither a1 nor b1 is strictly dominated, as

a1 yields the highest payoff for player 1 when players 2 and 3 choose (a2,a3), while
b1 is best when players 2 and 3 choose (b2,b3). However, a1 strictly dominates c1.
Thus, the surviving strategies for player 1 are a1 and b1. Turning to player 2, nei-
ther of her strategies is strictly dominated since a2 is best when players 1 and
3 choose (b1,a3) and b2 is best when players 1 and 3 choose (a1, a3). Finally, neither
of player 3’s strategies is strictly dominated as a3 is best when players 1 and
2 choose (c1,a2) and b3 is best when players 1 and 2 choose (a1,a2). After the first
round, the reduced game is as shown in the figure below.

a3 b3
a2 b2 a2 b2
a1 3,1,0 2,3,1 a1 3,1,1 1,3,2
b1 0,3,1 1,1,0 b1 2,0,2 2,2,1

Since no strategies of players 2 and 3 were eliminated in the first round, no strat-
egies of player 1 can be eliminated in the second round. Neither of player 2’s strate-
gies is strictly dominated, as a2 is best when players 1 and 3 choose (b1, a3) and b2
is best when players 1 and 3 choose (a1, a3). For player 3, b3 strictly dominates a3.
After the second round, the reduced game is as shown in the figure below.

b3
a2 b2
a1 3,1,1 1,3,2
b1 2,0,2 2,2,1

In round 3, neither of player 1’s strategies is strictly dominated, but for player 2,
b2 strictly dominates a2. After the third round, the reduced game is as shown in the
figure below.

b3
b2
a1 1,3,2
b1 2,2,1

In the fourth round, b1 strictly dominates a1 and this solves the game since each
player has only one strategy remaining. Thus, by the iterative deletion of strictly
dominated strategies, we conclude that the strategy profile that will be played is
(b1,b2 ,b3).

9. A gang controls the drug trade along North Avenue between Maryland Avenue and
Barclay Street. The city grid is shown below.

Maryland Charles St. Paul Calvert Barclay

Avenue Street Street Street Street

North
Avenue
3-8 SOLUTIONS MANUAL CHAPTER 3: ELIMINATING THE IMPOSSIBLE: SOLVING A GAME WHEN RATIONALITY IS COMMON KNOWLEDGE

The gang leader sets the price of the drug being sold and assigns two gang members to
place themselves along North Avenue. He tells each of them that they’ll be paid 20% of
the money they collect. The only decision that each of the drug dealers has is whether
to locate at the corner of North Avenue and either Maryland Avenue, Charles Street,
St. Paul Street, Calvert Street, or Barclay Street. The strategy set of each drug dealer is
then composed of the latter five streets. Since the price is fixed by the leader and the
gang members care only about money, each member wants to locate so as to maximize
the number of units he sells.
For simplicity, assume that the five streets are equidistant from each other. Drug
customers live only along North Avenue and are evenly distributed between Maryland
Avenue and Barclay Street (so no customers live to the left of Maryland Avenue or to the
right of Barclay Street). Customers know that the two dealers set the same price, so they
buy from the dealer that is closest to them. The total number of units sold on North
Avenue is fixed. The only issue is whether a customer buys from drug dealer 1 or drug
dealer 2. This means that a drug dealer will want to locate so as to maximize his share
of customers. We can then think about a drug dealer’s payoff as being his customer
share. The figure below shows the customer shares or payoffs.

Drug Dealers’ Payoffs Based on Location

Dealer 2's location

Maryland Charles St. Paul Calvert Barclay
1 1 1 7 1 3 3 5 1 1
Maryland ,
2 2
,
8 8
,
4 4
,
8 8
,
2 2
7 1 1 1 3 5 1 1 5 3
Charles ,
8 8
,
2 2
,
8 8
,
2 2
,
8 8
Dealer 1's 3 1
, 5 3
, 1 1
, 5 3
, 3 1
,
St. Paul
location 4 4
5 3
8 8
1 1
2 2
3 5
8 8
1 1
4 4
7 1
Calvert ,
8 8
,
2 2
,
8 8
,
2 2
,
8 8
1 1 3 5 1 3 1 7 1 1
Barclay ,
2 2
,
8 8
,
4 4
,
8 8
,
2 2

Let us go through a few so that you understand how they were derived. For example,
suppose dealer 1 locates at the corner of Maryland and North and dealer 2 parks his
wares at the corner of St. Paul and North. All customers who live between Maryland and
Charles buy from dealer 1, as he is the closest to them, while the customers who live to the
right of Charles buy from dealer 2. Hence, dealer 1 gets 25% of the market and dealer 2
1 3
gets 75%. Thus, we see that (4 , 4 ) are the payoffs for strategy pair (Maryland,
St. Paul). Now, suppose instead that dealer 2 locates at Charles and dealer 1 at Maryland.
The customer who lies exactly between Maryland and Charles will be indifferent as to
whom to buy from. All those customers to his left will prefer the dealer at Maryland, and
1 7
they make up one-eighth of the street. Thus, the payoffs are (8 , 8 ) for the strategy pair
(Maryland, Charles). If two dealers locate at the same street corner, we’ll suppose that
1 1
customers divide themselves equally between the two dealers, so the payoffs are (2 , 2 ).
Using the IDSDS, find where the drug dealers locate.

ANSWER: Consider what locations are optimal for dealer 1 for some location of
dealer 2. If dealer 2 chooses to locate at Maryland, then dealer 1’s optimal location
is at Charles. At this location, he picks up all of the customers to the right of Charles
and half of those between Maryland and Charles, which means getting an impres-
7
sive market share of 8 . Thus, Charles cannot be strictly dominated as it is the best
location for some location of dealer 2. If instead dealer 2 locates at Charles, then
dealer 1’s optimal location is to park it just to the right of dealer 2 at St. Paul and
5
pick up 8 of the market (all those customers to the right of St. Paul and those to the
left of St. Paul up to the mid-point between Charles and St. Paul). Thus, St. Paul is
not strictly dominated either. If dealer 2 locates in the middle of North Avenue at
St. Paul, then dealer 1 wants to do the same. Hence, St. Paul is not strictly domi-
nated. If dealer 2 locates at Calvert, then dealer 1 wants to locate just to his left at
St. Paul, which we already know is not dominated. Finally, if dealer 2 locates at
7
Barclay, then dealer 1 sets up shop at Calvert to grab 8 of the market. Thus, locating
at Calvert is not strictly dominated. In sum, Charles, St. Paul, and Calvert are not
strictly dominated locations because each is optimal for some location of dealer 2
 SOLUTIONS MANUAL CHAPTER 3: ELIMINATING THE IMPOSSIBLE: SOLVING A GAME WHEN RATIONALITY IS COMMON KNOWLEDGE 3-9

(and thus there is not any other location that produces a strictly higher payoff for
every location of dealer 2). This leaves only Maryland and Barclay as candidates for
elimination. In fact, Maryland is strictly dominated by Charles, while Barclay is
strictly dominated by Calvert. It is then strictly dominated for dealer 1 to locate at
either end of North Avenue. As the game is symmetric, the same answer is derived
for dealer 2; the only strictly dominated locations are Maryland and Barclay. After
one round of the IDSDS, the reduced game is as shown in the figure below.

Dealer 2’s location

Charles St. Paul Calvert
1 1 3 5 1 1
Charles ,
2 2
,
8 8
,
2 2
Dealers 1’s 5 3
, 1 1
, 5 3
,
St. Paul
location 8 8 2 2 8 8
1 1 3 5 1 1
Calvert ,
2 2
,
8 8
,
2 2

Inspection reveals that St. Paul dominates both Charles and Calvert for dealer 1,
and similarly for dealer 2. Thus, after two rounds of the IDSDS, we are left with a
single strategy for each dealer, which is to locate at St. Paul. If gang members are
rational and each gang member believes the other is rational, then they’ll both
locate at St. Paul and each sell to half of the customers there.

10. Two students are to take an exam, and the professor has instructed them that the student
with the higher score will receive a grade of A and the one with the lower score will
receive a B. Student 1’s score equals x1  1.5, where x1 is the amount of effort she invests
in studying. (That is, I assume that the greater the effort, the higher is the score.)
Student 2’s score equals x2, where x2 is the amount of effort he exerts. It is implicitly
assumed that student 1 is the smarter of the two, in that, if the amount of effort is held
fixed, student 1 has a higher score by an amount of 1.5. Assume that x1 and x2 can take
any value in {0,1,2,3,4,5}. The payoff to student i is 10  xi if she gets an A, and 8  xi if
she gets a B, i  1, 2.
a. Derive the strategies that survive the IDSDS.

ANSWER: Let us first show that for either player, zero effort strictly dominates
effort levels of 3, 4, and 5. If xi  0 then i’s payoff is at least 8, which occurs
when she gets a B. By choosing effort xi, the highest possible payoff is 10  xi,
which occurs when she gets an A. Since 8  10  xi when xi  2, then zero effort
strictly dominates effort of 3, 4, or 5. Now consider player 1. We know that
strategies 3, 4, and 5 are strictly dominated, but what about strategies 0, 1, and
2? A useful point to note is that if there is some strategy for player 2 such that
a particular strategy for player 1 is best (that is, it yields the highest payoff for
player 1), then that strategy for 1 is not strictly dominated since strict domi-
nance means there is another strategy that yields a strictly higher payoff for all
strategies of the other player. In considering strategy 0, note that x1  0 is the
uniquely best strategy for player 1 when x2 is 0 or 1, as in both cases player 1
receives a payoff of 10, whereas when x1  0 his payoff is 10  x1, which is
lower. Thus, there is no strategy that strictly dominates zero effort. A similar
argument shows that x1  1 is not strictly dominated. If x2  2, then x1  1
yields a payoff of 9, while x1  0 yields a payoff of 8 (as player 2’s score is higher
in that case) and the payoff is 10  x1 when x1  1, which is lower than 9.
Finally, note that x1  2 is optimal when x2  3, as the resulting payoff is 8,
whereas the payoff is 8 from x1  0, 7 from x1  1, and 10  x1 when x1  2.
Though x1  2 generates the same payoff as x1  0, we can still conclude that
there is no strategy that strictly dominates x1  2. It is concluded that the strat-
egies for player 1 that survive the first round of deletion of strictly dominated
strategies are {0,1,2}. Now consider player 2. x2  0 is not strictly dominated, as
it is the optimal strategy for player 2 when x1  4, as in that case player 1 gets
a B regardless of x2. Since his payoff is then 8  x2, it is clearly maximized at
x2  0. Now consider x2  1. I want to show that this strategy is strictly dominated
by x2  0. Note that regardless of x1, player 2 has the lower score whether he
chooses x2  0 or x2  1. Since x2  0 gives a payoff of 8 and x2  1 gives a
3-10 SOLUTIONS MANUAL CHAPTER 3: ELIMINATING THE IMPOSSIBLE: SOLVING A GAME WHEN RATIONALITY IS COMMON KNOWLEDGE

payoff of 7, zero effort strictly dominates effort of player 1. Now consider the
remaining strategy of x2  2. Suppose x1  0. In that case, the payoff to player
2 from x2  1 is 8  x2 (since he gets a B) and from x2  2 is 10  x2 (since he
gets an A). His payoff is maximized at x2  2. Given there is some strategy for
player 1 such that x2  2 yields the maximum payoff, then x2  2 cannot be
strictly dominated. It is concluded that the strategies for player 2 that survive
the first round of the IDSDS are {0,2}. Now move to round 2 of this iterative
process. The payoff matrix in the figure below shows the surviving strategies.
The first number in a cell is player 1’s payoff.

x2
0 2
0 10,8 8,8
x1 1 9,8 9,6
2 8,8 8,6

x1  1 strictly dominates x1  2. Neither of the other two strategies for player

1 is dominated. None of player 2’s strategies is strictly dominated since x2  0
yields a higher payoff than does x2  2 when x1  1 or 2 and x2  2 yields the
same payoff as does x2  0 when x1  0. It is concluded that the strategies that
survive the second round of the IDSDS for player 1 are {0,1} and for player 2 are
{0,2}. Moving to the third round, the resulting payoff matrix is shown in the fig-
ure below.

x2
0 2
0 10,8 8,8
x1
1 9,8 9,6

Inspection reveals that no strategies are strictly dominated. Thus, the IDSDS
predicts that player 1 will exert effort of 0 or 1 and player 2 will exert effort of
0 or 2.

b. Derive the strategies that survive the iterative deletion of weakly dominated strat-
egies. (The procedure works the same as the iterative deletion of strictly domi-
nated strategies, except that you eliminate all weakly dominated strategies at
each stage.)

ANSWER: Now let us repeat the analysis when we are instead iteratively deleting
weakly dominated strategies. From part (a), we know that effort of 3, 4, and 5 are
all strictly dominated and therefore weakly dominated. For player 1, efforts of 0
and 1 are not weakly dominated for the same reasons given in part (a): there is a
strategy for player 2 such that each yields the strictly highest payoff. While x1  2
is not strictly dominated, it is weakly dominated by x1  0. The latter yields a pay-
off of at least 8 regardless what player 2 does, while x1  2 yields a payoff of at
most 8. Furthermore, zero effort does strictly better when player 2 chooses effort
of 0 or 1. The strategies for player 1 that survive the first round of the IDSDS are
{0,1}. With respect to player 2, it was previously shown that strategies 1, 3, 4, and
5 are strictly dominated and, therefore, weakly dominated. Zero effort is not
weakly dominated by the same argument as in part (a). x2  2 is weakly dominated
for reasons analogous to those given for player 1. The lone strategy for player 2
that survives the first round of the IDSDS is x1  0.
Moving to the second round, x1  0 weakly dominates x1  1 since both result
in player 1’s getting an A (as player 2 is choosing 0 effort). So the former yields a
payoff of 10 and the latter a payoff of 9. The IDSDS predicts that both players will
exert zero effort.

11. Groucho Marx once said, “I’ll never join any club that would have me for a member.”
Well, Groucho is not interested in joining your investment club, but Julie is. Your club
 SOLUTIONS MANUAL CHAPTER 3: ELIMINATING THE IMPOSSIBLE: SOLVING A GAME WHEN RATIONALITY IS COMMON KNOWLEDGE 3-11

has 10 members, and the procedure for admitting a new member is simple: Each person
receives a ballot that has two options: (1) admit Julie and (2) do not admit Julie. Each
person can check one of those two options or abstain by not submitting a ballot. For
Julie to be admitted, she must receive at least six votes in favor of admittance. Letting
m be the number of ballots submitted with option 1 checked, assume that your payoff
function is

r 1 if m  6, 7, 8, 9, 10
0 if m  0, 1, 2, 3, 4, 5
a. Prove that checking option 1 (admit Julie) is not a dominant strategy.

ANSWER: If all the other players check option 2, then Julie is not admitted regard-
less whether you check option 1, check option 2, or abstain. Since all three strate-
gies yield the highest payoff in that case, none of them is strictly dominated and
thus there is no dominant strategy.

b. Prove that abstaining is a weakly dominated strategy.

ANSWER: If at least six of the other members submit ballots with option 1
checked, then Julie is admitted regardless of what you do; your payoff is 1 with all
strategies. If four or fewer of the other members vote in favor of option 1, then
Julie is denied admittance regardless of what you do; your payoff is 0 with all
strategies. This leaves only the case when five of the other members submit ballots
in favor of admitting Julie. Abstaining results in a payoff of 0, as Julie ends up with
only five supporting votes. Voting and checking option 1 results in her admittance
and thus a payoff of 1. Hence, abstaining is weakly dominated by voting in favor
of Julie.

c. Now suppose you’re tired at the end of the day, so that it is costly for you to attend
the evening’s meeting to vote. By not showing up, you abstain from the vote. This is
reflected in your payoff function having the form

1 if m  6, 7, 8, 9, 10 and you abstained

1
u 2 if m  6, 7, 8, 9, 10 and you voted
0 if m  0, 1, 2, 3, 4, 5 and you abstained
1
2 if m  0, 1, 2, 3, 4, 5 and you voted

Prove that abstaining is not a weakly dominated strategy.

ANSWER: If the other members vote so that m  6, then the payoff from showing
1
up and voting is 2 , while it’s 1 from not showing up and thus abstaining. Since
there are strategies for the other players whereby abstention is the unique optimal
strategy, then abstention cannot be weakly dominated.

12. Derive all of the rationalizable strategies for the game shown.

Player 2
x y z
a 0,4 1,1 2,3
Player 1 b 1,1 2,2 0,0
c 3,2 0,0 1,4

ANSWER: All strategies are rationalizable. The following cycle of beliefs serves
to show strategies a and c for player 1 and x and z for player 2 are rationalizable.
(1) Player 1 using c is optimal if player 1 believes player 2 will use x. Hence, player 1
using c is consistent with player 1 being rational. (2) x is optimal for player 2 if
player 2 believes player 1 will use a. Hence, player 1 believing that player 2 will use
x is consistent with player 1 believing that player 2 is rational. (This “rationalizes”
3-12 SOLUTIONS MANUAL CHAPTER 3: ELIMINATING THE IMPOSSIBLE: SOLVING A GAME WHEN RATIONALITY IS COMMON KNOWLEDGE

player 1’s belief in step 1 that player 2 will use x.) (3) a is optimal for player 1 if
player 1 believes player 2 will use z. Hence, player 1 believing that player 2 believes
that player 1 will use a is consistent with player 1 believing that player 2 believes
that player 1 is rational. (4) z is optimal for player 2 if player 2 believes player 1
will use c. Hence, player 1 believing that player 2 believes that player 1 believes
that player 2 will use z is consistent with player 1 believing that player 2 believes
that player 1 believes that player 2 is rational. (5) c is optimal for player 1 if player 1
believes player 2 will use x. Hence, player 1 believing that player 2 believes that
player 1 believes that player 2 believes that player 1 will use c is consistent with
player 1 believing that player 2 believes that player 1 believes that player 2 believes
that player 1 is rational. One can then use the argument in step 2 to rationalize
player 1’s belief about player 2 using x. We then have a cycle in which there is
always another layer of beliefs that rationalizes the current layer. All the strategies
in this cycle are then rationalizable, which means a and c for player 1 and x and z
for player 2. This doesn’t necessarily mean that b and y are not rationalizable, as
there may be beliefs, consistent with rationality being common knowledge, that
support them. In fact, there are—in which case it turns out that all strategies in
this game are rationalizable.
Here is a cycle that rationalizes b and y. (1) Player 1 using b is optimal if player 1
believes player 2 will use y. Hence, player 1 using b is consistent with player 1
being rational. (2) y is optimal for player 2 if player 2 believes player 1 will use b.
Hence, player 1 believing that player 2 will use y is consistent with player 1 believ-
ing that player 2 is rational. (3) b is optimal for player 1 if player 1 believes player 1
will use y. Hence, player 1 believing that player 2 believes that player 1 will use
b is consistent with player 1 believing that player 2 believes that player 1 is ratio-
nal. (4) y is optimal for player 2 if player 2 believes player 1 will use b. Hence,
player 1 believing that player 2 believes that player 1 believes that player 2 will use
y is consistent with player 1 believing that player 2 believes that player 1 believes
that player 2 is rational. We have a cycle!

13. Consider the two-player game:

Player 2
x y z
a 5,1 4,2 0,1
Player 1 b 1,2 0,4 6,3
c 2,3 1,2 2,1

a. Find the strategies that are consistent with both players being rational and each
player believing the other player is rational.

ANSWER: Round 1: y strictly dominates z for player 2.

Player 2
x y
a 5,1 4,2
Player 1 b 1,2 0,4
c 2,3 1,2

Round 2: a strictly dominates b and c for player 1. Neither strategy strictly domi-
nates for player 2.

1 plays a and 2 plays x or y. As 2 does not know whether 1 knows that 2 is rational,
then 2 does not know whether 1 knows that 2 does not play z.
 SOLUTIONS MANUAL CHAPTER 3: ELIMINATING THE IMPOSSIBLE: SOLVING A GAME WHEN RATIONALITY IS COMMON KNOWLEDGE 3-13

b. In addition to that assumed in part (a), assume that player 2 knows player 1 knows player
2 is rational. Find strategies consistent with these beliefs.

ANSWER: Round 3: y strictly dominates x for player 2.

Player 2
x y
Player 1 a 5,1 4,2

Note that the information available to player 1 has not changed from part (a),
therefore, Player 1 still plays a. However, now player 2 knows player 1 knows
player 2 is rational.
Therefore, player 2 knows player 1 knows that player 2 will not play z from which
it follows that player 2 knows player 1 plays a (because a strictly dominates b and c in
the absence of z). Player 2 then plays y. Hence, player 1 plays a and player 2 plays y.

14. Len and Melanie are deciding what to do Saturday night. The options are to see
Mozart’s opera Don Giovanni or go to the local arena to watch Ultimate Fighter. Len
prefers Ultimate Fighter, while Melanie prefers Don Giovanni. As a possible compro-
mise, a friend suggests that they attend “Rocky: The Ballet,” which is a newly produced
ballet about Rocky Balboa, the down-and-out boxer from the streets of Philadelphia who
gets a shot at the title. Each would like to go to their most preferred performance, but
each also cares about attending with the other person. Also, Len may feel guilty about
spending a lot of money for a ticket to Ultimate Fighter when Melanie is not with him;
Rocky: The Ballet is cheaper. Don Giovanni and Ultimate Fighter are both expensive
tickets, but Melanie would not feel guilty about attending her first choice alone and
spending a lot of money. Both Len and Melanie are flying back into town Saturday after-
noon and each must independently decide which to attend. The strategic form of the
game is shown below. Using the IDSDS, what will they do?

Melanie
Don Ultimate Rocky:
Giovanni Fighter The Ballet
Don Giovanni 1,5 0,0 0,2
Len Ultimate Fighter 3,3 6,1 3,2
Rocky: The Ballet 4,3 2,0 5,4

ANSWER: Round 1: Ultimate Fighter (UF) and Rocky: The Ballet (R) strictly
dominate Don Giovanni (DG) for Len. For Melanie, DG and R strictly dominate
UF. After eliminating these strategies, the game is now:

Melanie
Don Giovanni Rocky: The Ballet
Ultimate Fighter 3,3 3,2
Len
Rocky: The Ballet 4,3 5,4

Round 2: R strictly dominates UF for Len, while nothing is strictly dominated for
Melanie. The game is now:

Melanie
Don Giovanni Rocky: The Ballet
Len Rocky: The Ballet 4,3 5,4

Round 3: R strictly dominates DG for Melanie. Thus, they both go to see Rocky:
The Ballet.
3-14 SOLUTIONS MANUAL CHAPTER 3: ELIMINATING THE IMPOSSIBLE: SOLVING A GAME WHEN RATIONALITY IS COMMON KNOWLEDGE

15. A total of 10 players are each choosing a number from {0,1,2,3,4,5,6,7,8}. If a player’s num-
ber equals exactly half of the average of the numbers submitted by the other nine players,
then she is paid $100; otherwise, she is paid 0. Solve for the strategies that survive the IDSDS.

ANSWER: Round 1: The average is no higher than 8 so half the average is no higher
than 4. Thus, 4 weakly dominates any number above 4. Compare 4 to a number
above 4, call it x. If the average is 8 then half of the average is 4 in which case 4 pays
off while x does not pay off. If the average is less than 8 then neither 4 nor x pays
off. Hence, 4 weakly dominates 5,6,7, and 8. Next note that 0,1,2,3, and 4 are not
weakly dominated. If all others choose 8 then 4 is uniquely best. If all others choose
6 then 3 uniquely is best. If all others choose 4 then 2 is uniquely best. If all others
choose 2 then 1 is uniquely best. If all others choose 0 then 0 is uniquely best.
Round 2: The remaining strategies are 0,1,2,3, and 4. By the same argument as used
in Round 1, 2 weakly dominates 3 and 4. 2 is not weakly dominated because it is
uniquely best when all others choose 4; 1 is not weakly dominated because it is
uniquely best when all others choose 2; 0 is not weakly dominated because it is
uniquely best when all others choose 0.
Round 3: The remaining strategies are 0,1, and 2. By the same argument as used in
Round 1, 1 weakly dominates 2. 1 is not weakly dominated because it is uniquely
best when all others choose 2; 0 is not weakly dominated because it is uniquely best
when all others choose 0.
Round 4: The remaining strategies are 0 and 1. 0 weakly dominates 1. If all others
choose 0 then, because half of 0 is 0, it pays off, and 1 does not. If anything else is
chosen then the average lies between 0 and 1 in which case neither 0 nor 1 pays off.

16. Monica and Isabel are roommates who, on this particular Saturday morning, are trying
to decide what scarf to wear. Each has a Burberry scarf (which we’ll denote B), a tan scarf
(denoted T), and a mauve scarf (denoted M). They care about the scarf but also about
whether they end up wearing the same or different scarves. The preference ordering
(from best to least preferred outcome) for Monica is: (1) she wears B and Isabel wears T
or M; (2) she wears T and Isabel wears B or M; (3) she wears B and Isabel wears B; (4) she
wears T and Isabel wears T: (5) she wears M and Isabel wears M; and (6) she wears M
and Isabel wears B or T. Isabel’s preference ordering is: (1) she wears T and Monica wears
B or M; (2) she wears M and Monica wears B or T; (3) she wears T and Monica wears T;
(4) she wears M and Monica wears M; (5) she wears B and Monica wears B; and (6) she
wears B and Monica wears T or M. Applying the IDSDS, which scarves will be worn?

ANSWER: The preferences of Isabel and Monica can be represented in the follow-
ing strategic form game:
Isabel
B T M
B 3,1 5,5 5,4
Monica T 4,0 2,3 4,4
M 0,0 0,5 1,2

Round 1: If Monica expects Isabel to wear the Burberry (denoted B) or mauve

(denoted M) scarf then Monica wants to wear the tan scarf (denoted T). If Monica
expects Isabel to wear B or M then Monica wants to wear B. Thus, neither B nor
T are strictly dominated. However, M is strictly dominated by both B and T for
Monica. If Isabel expects Monica to wear B or T then Isabel wears M. If Isabel
expects Monica to wear B or M then she wears T. Thus, neither M nor T are strictly
dominated for Isabel. However, B is strictly dominated by T and M for Isabel.
Round 2: After Round 1, B and T are the surviving strategies for Monica, and T
and M are the surviving strategies for Isabel. Let us consider Monica. If Isabel
wears T then Monica prefers B and if Isabel wears M then Monica prefers B. B
is Monica’s favorite scarf and she knows that it is a strictly dominated strategy
for Isabel to wear it. Turning to Isabel, if Monica wears B then Isabel prefers T;
and if Monica wears T then Isabel prefers M. Thus, both scarves are not strictly
dominated for Isabel.
 SOLUTIONS MANUAL CHAPTER 3: ELIMINATING THE IMPOSSIBLE: SOLVING A GAME WHEN RATIONALITY IS COMMON KNOWLEDGE 3-15

Round 3: After Round 2, B survives for Monica, and T and M survive for Isabel.
For Isabel, T strictly dominates M given Monica is choosing B.
The answer is that Monica wears the Burberry scarf and Isabel wears the tan scarf.

17. Consider the following game.

Player 2
w x y z
a 1,3 4,4 2,2 6,1
b 0,4 3,2 0,0 5,5
Player 1
c 1,2 5,3 2,2 1,6
d 2,3 2,4 4,2 6,2

a. Find the strategies that survive the IDSDS.

ANSWER: The strategies that survive after round 1 of the iterative deletion of
strictly dominated strategies are:

Player 2
w x z
a 1,3 4,4 6,1
Player 1 c 1,2 5,3 1,6
d 2,3 2,4 6,2

The strategies that survive after round 2 of the iterative deletion of strictly domi-
nated strategies are:

Player 2
x z
a 4,4 6,1
Player 1 c 5,3 1,6
d 2,4 6,2

In round 3, no strategies are strictly dominated. Therefore, strategies a, c, and d

survive for player 1, and strategies x and z survive for player 2.

b. Find the rationalizable strategies.

ANSWER: We know that the strategies eliminated by the iterative deletion of

strictly dominated strategies are not rationalizable. Thus, it follows from part (a)
that we can focus on this reduced game in answering part (b):

Player 2
x z
a 4,4 6,1
Player 1 c 5,3 1,6
d 2,4 6,2

Strategy a is optimal for player 1 if she believes player 2 will choose z; z is optimal
for 2 if he believes 1 will choose c; c is optimal for 1 if she believes 2 will choose x;
and x is optimal for 2 if he believes 1 will choose a. We then have a cycle which
means strategies a and c for player 1 and strategies x and z for player 2 are ratio-
nalizable. Next consider: d is optimal for 1 if she believes 2 will choose z, and we’ve
already shown that z is rationalizable. Thus, a, c, d are rationalizable for player 1,
and x and z are rationalizable for player 2.
3-16 SOLUTIONS MANUAL CHAPTER 3: ELIMINATING THE IMPOSSIBLE: SOLVING A GAME WHEN RATIONALITY IS COMMON KNOWLEDGE

18. Consider the three-player game below. Player 1 selects a row, either a1, b1, or c1.
Player 2 selects a column, either a2, b2, or c2. Player 3 selects a matrix, either a3 or b3
or c3. The first number in a cell is player 1’s payoff, the second number is player 2’s
payoff, and the last number is player 3’s payoff. Derive the strategies that survive
the IDSDS.

a3 b3 c3

a2 b2 c2 a2 b2 c2 a2 b2 c2
a1 3,1,4 2,2,2 3,1,4 a1 1,1,2 3,3,1 2,2,2 a1 4,0,1 3,1,1 3,5,2
b1 2,4,1 5,3,3 1,2,2 b1 2,2,0 1,1,0 3,0,3 b1 2,5,0 2,4,2 3,2,1
c1 5,4,5 4,1,6 5,0,1 c1 1,3,3 0,4,1 3,2,2 c1 2,6,3 6,1,3 0,0,0

ANSWER:
Round 1
Player 1: No strategies are strictly dominated because a1 is a best reply to (a2, c3),
b1 is a best reply to (b2, a3), and c1 is a best reply to (a2, a3)
Player 2: No strategies are strictly dominated because a2 is a best reply to (b1, a3),
b2 is a best reply to (c1, b3), and c2 is a best reply to (a1, c3)
Player 3: a3 strictly dominates c3. a3 is a best reply to (a1, a2) and b3 is a best
reply to (b1, c2).
After eliminating the strictly dominated strategies, we have:

Player 3: a3 Player 3: b3
Player 2 Player 2
a2 b2 c2 a2 b2 c2
a1 3,1,4 2,2,2 3,1,4 a1 1,1,2 3,3,1 2,2,2
Player 1 b1 2,4,1 5,3,3 1,2,2 Player 1 b1 2,2,0 1,1,0 3,0,3
c1 5,4,5 4,1,6 5,0,1 c1 1,3,3 0,4,1 3,2,2

Round 2
Player 1: No strategies are strictly dominated.
Player 2: b2 strictly dominates c2.
Player 3: No strategies are strictly dominated.
After eliminating the strictly dominated strategies, we have:

Player 3: a3 Player 3: b3
Player 2 Player 2
a2 b2 a2 b2
a1 3,1,4 2,2,2 a1 1,1,2 3,3,1
Player 1 b1 2,4,1 5,3,3 Player 1 b1 2,2,0 1,1,0
c1 5,4,5 4,1,6 c1 1,3,3 0,4,1
 SOLUTIONS MANUAL CHAPTER 3: ELIMINATING THE IMPOSSIBLE: SOLVING A GAME WHEN RATIONALITY IS COMMON KNOWLEDGE 3-17

Round 3
Player 1: No strategies are strictly dominated.
Player 2: No strategies are strictly dominated.
Player 3: a3 strictly dominates b3.
After eliminating the strictly dominated strategies, we have:

Player 3: a3
Player 2
a2 b2
a1 3,1,4 2,2,2
Player 1 b1 2,4,1 5,3,3
c1 5,4,5 4,1,6

Round 4
Player 1: c1 strictly dominates a1.
Player 2: No strategies are strictly dominated.
Player 3: No strategies are strictly dominated.
After eliminating the strictly dominated strategies, we have:

Player 3: a3
Player 2
a2 b2
b1 2,4,1 5,3,3
Player 1
c1 5,4,5 4,1,6

Round 5
Player 1: No strategies are strictly dominated.
Player 2: a2 strictly dominates b2.
Player 3: No strategies are strictly dominated.
After eliminating the strictly dominated strategies, we have:

Player 3: a3
Player 2
a2
b1 2,4,1
Player 1
c1 5,4,5

Round 6
Player 1: c1 strictly dominates b1.
Player 2: No strategies are strictly dominated.
Player 3: No strategies are strictly dominated.
The answer is (c1,a2,a3).
3-18 SOLUTIONS MANUAL CHAPTER 3: ELIMINATING THE IMPOSSIBLE: SOLVING A GAME WHEN RATIONALITY IS COMMON KNOWLEDGE

19. Consider a four-player game in which each player chooses between two strategies: a and
b. Their payoffs are shown in the accompanying table for the 16 possible strategy pro-
files. Find the strategies that survive the IDSDS.

Strategy Profiles Payoffs

Player 1 Player 2 Player 3 Player 4 Player 1 Player 2 Player 3 Player 4
a a a a 3 1 2 1
a a a b 2 5 3 3
a a b a 4 2 4 4
a a b b 3 2 5 2
a b a a 2 3 1 0
a b a b 4 4 0 3
a b b a 3 5 2 6
a b b b 2 0 3 5
b a a a 1 5 3 3
b a a b 5 2 1 2
b a b a 1 6 4 5
b a b b 1 3 5 1
b b a a 2 3 2 4
b b a b 2 3 1 0
b b b a 2 7 4 3
b b b b 4 5 3 1

ANSWER:
Round 1
Player 1: No strategy is strictly dominated as a is a best reply to (a,a,a) and b
is a best reply to (b,b,b)
Player 2: No strategy is strictly dominated as a is a best reply to (b,b,b) and b
is a best reply to (b,a,b)
Player 3: b strictly dominates a.
Player 4: No strategy is strictly dominated as a is a best reply to (a,a,b) and b
is a best reply to (a,a,a)
We now have:

Players’ Strategies Players’ Payoffs

Player 1 Player 2 Player 3 Player 4 Player 1 Player 2 Player 3 Player 4
a a b a 4 2 4 4
a a b b 3 2 5 2
a b b a 3 5 2 6
a b b b 2 0 3 5
b a b a 1 6 4 5
b a b b 1 3 5 1
b b b a 2 3 4 3
b b b b 4 5 3 1
 SOLUTIONS MANUAL CHAPTER 3: ELIMINATING THE IMPOSSIBLE: SOLVING A GAME WHEN RATIONALITY IS COMMON KNOWLEDGE 3-19

Round 2
Player 1: No strategy is strictly dominated as a is a best reply to (a,b,a) and b
is a best reply to (b,b,a).
Player 2: No strategy is strictly dominated as a is a best reply to (b,b,a) and b
is a best reply to (b,b,a)
Player 3: No strategy is strictly dominated.
Player 4: a strictly dominates b.

Players’ Strategies Players’ Payoffs

Player 1 Player 2 Player 3 Player 4 Player 1 Player 2 Player 3 Player 4
a a b a 4 2 4 4
a b b a 3 5 2 6
b a b a 1 6 4 5
b b b a 2 3 4 3

Round 3
Player 1: a strictly dominates b.
Player 2: No strategy is strictly dominated as a is a best reply to (b,b,a) and b
is a best reply to (a,b,a).
Player 3: No strategy is strictly dominated.
Player 4: No strategy is strictly dominated.

Players’ Strategies Players’ Payoffs

Player 1 Player 2 Player 3 Player 4 Player 1 Player 2 Player 3 Player 4
a a b a 4 2 4 4
a b b a 3 5 2 6

Round 4: For player 2, b strictly dominates a.

The answer is (a,b,b,a).

20. For the game below, find the strategies that survive the IDSDS when mixed strategies
can be used to eliminate a pure strategy as being strictly dominated.

Player 2
x y z
a 2,4 3,0 0,1
Player 1
b 0,0 1,5 4,2

ANSWER:

Round 1: For player 1, strategy a is best when player 2 chooses x or y; strategy b is

best when 2 chooses z. Thus, none of player 1’s strategies are strictly dominated.
For player 2, strategies x and y are not strictly dominated because they are the best
strategy when 1’s strategy is a and b, respectively. While strategy z is not strictly
dominated by either x or y, it is strictly dominated by a mixed strategy composed
3-20 SOLUTIONS MANUAL CHAPTER 3: ELIMINATING THE IMPOSSIBLE: SOLVING A GAME WHEN RATIONALITY IS COMMON KNOWLEDGE

of x and y. Suppose x is chosen with probability p and y with probability 1  p.

This mixed strategy strictly dominates pure strategy z when:
1
Player 1 chooses a: p  4  (1  p)  0  1 or p  4
3
Player 1 chooses b: p  0  (1  p)  5  2 or 5
 p
Thus, if .25 p .6 then choosing x with probability p and y with probability
1  p yields a strictly higher payoff than pure strategy z.
Round 2: The game is now

Player 2
x y
a 2,4 3,0
Player 1
b 0,0 1,5

For player 1, strategy a strictly dominates b. For player 2, neither x nor y are
strictly dominated.
Round 3: Given that player 1’s only surviving strategy is a, x strictly dominates y
for player 2. Therefore, the strategy pair (a,x) is the lone survivor of the iterative
deletion of strictly dominated strategies. Note that if mixed strategies are not
allowed then all strategy pairs survive the iterative deletion of strictly dominated
strategies.