HMMT November 2019

November 9, 2019
Guts Round
1. [5] A polynomial P with integer coefficients is called tricky if it has 4 as a root.
A polynomial is called teeny if it has degree at most 1 and integer coefficients between −7 and 7,
inclusive.
How many nonzero tricky teeny polynomials are there?
Proposed by: Carl Schildkraut
Answer: 2
If a degree 0 polynomial has 4 as a root, then it must be the constant zero polynomial. Thus, we will
only consider polynomials of degree 1.
If P has degree 1, integer coefficients, and 4 as a root, then it must be of the form P (x) = a(x − 4) =
ax − 4a for some nonzero integer a. Since all integer coefficients are between −7 and 7, inclusive, we
require −7 ≤ 4a ≤ 7, which gives us a = −1, 1. Note that for both values, the coefficient of x is also
between −7 and 7, so there are 2 tricky teeny polynomials.
2. [5] You are trying to cross a 6 foot wide river. You can jump at most 4 feet, but you have one stone
you can throw into the river; after it is placed, you may jump to that stone and, if possible, from
there to the other side of the river. However, you are not very accurate and the stone ends up landing
uniformly at random in the river. What is the probability that you can get across?
Proposed by: Carl Schildkraut and Milan Haiman
1
Answer:
3
To be able to cross, the stone must land between 2 and 4 feet from the river bank you are standing
on. Therefore the probability is 26 = 13 .
3. [5] For how many positive integers a does the polynomial
x2 − ax + a
have an integer root?
Proposed by: Krit Boonsiriseth
Answer: 1
Let r, s be the roots of x2 − ax + a = 0. By Vieta’s, we have r + s = rs = a. Note that if one root is
an integer, then both roots must be integers, as they sum to an integer a. Then,
rs − (r + s) + 1 = 1 =⇒ (r − 1)(s − 1) = 1.
Because we require r, s to be both integers, we have r − 1 = s − 1 = ±1, which yields r = s = 0, 2.
If r = 0 or s = 0, then a = 0, but we want a to be a positive integer. Therefore, our only possibility
is when r = s = 2, which yields a = 4, so there is exactly 1 value of a (namely, a = 4) such that
x2 − ax − a has an integer root.
4. [6] In 2019, a team, including professor Andrew Sutherland of MIT, found three cubes of integers
which sum to 42:
42 = (−8053873881207597 )3 + (80435758145817515)3 + (12602123297335631)3
One of the digits, labeled by an underscore, is missing. What is that digit?
Proposed by: Andrew Gu
Answer: 4
Let the missing digit be x. Then, taking the equation modulo 10, we see that 2 ≡ −x3 + 53 + 13 . This
simplifies to x3 ≡ 4 (mod 10), which gives a unique solution of x = 4.
5. [6] A point P is chosen uniformly at random inside a square of side length 2. If P1 , P2 , P3 , and P4
are the reflections of P over each of the four sides of the square, find the expected value of the area of
quadrilateral P1 P2 P3 P4 .
Proposed by: Carl Schildkraut
Answer: 8
Let ABCD denote the square defined in the problem. We see that if P1 is the reflection of P over AB,
then the area of P1 AB is the same as the area of P AB. Furthermore, if P4 is the reflection of P over
DA, P1 , A, and P4 are collinear, as A is the midpoint of P1 P4 . Repeating this argument for the other
points gives us that the desired area is

[P1 AB]+[P2 BC]+[P3 CD]+[P4 DA]+[ABCD] = [P AB]+[P BC]+[P CD]+[P DA]+[ABCD] = 2[ABCD] = 8.

6. [6] Compute the sum of all positive integers n < 2048 such that n has an even number of 1’s in its
binary representation.
Proposed by: Milan Haiman
Answer: 1048064
Note that the positive integers less than 2047 are those with at most 11 binary digits. Consider the
contribution from any one of those digits. If we set that digit to 1, then the remaining 10 digits can
be set in 29 = 512 ways so that the number of 1’s is even. Therefore the answer is

512(20 + · · · + 210 ) = 512 · 2047 = 1048064.

7. [7] Let S be the set of all nondegenerate triangles formed from the vertices of a regular octagon with
side length 1. Find the ratio of the largest area of any triangle in S to the smallest area of any triangle
in S.
Proposed by: Carl Schildkraut
√
Answer: 3 + 2 2
By a smoothing argument, the largest triangle is that where the sides span 3, 3, and 2 sides of the
octagon respectively (i.e. it has angles 45◦ , 67.5◦ , and 67.5◦ ), and the smallest triangle is that formed
by three adjacent vertices of the octagon. Scaling so that the circumradius of the octagon is 1, our
answer is √
sin (90◦ ) + 2 sin (135◦ ) 1+ 2 √
◦ ◦
=√ = 3 + 2 2,
2 sin (45 ) − sin (90 ) 2−1
where the numerator is derived from splitting the large triangle by the circumradii, and the denominator
is derived from adding the areas of the two triangles formed by the circumradii, then subtracting the
area not in the small triangle.
8. [7] There are 36 students at the Multiples Obfuscation Program, including a singleton, a pair of
identical twins, a set of identical triplets, a set of identical quadruplets, and so on, up to a set of
identical octuplets. Two students look the same if and only if they are from the same identical
multiple. Nithya the teaching assistant encounters a random student in the morning and a random
student in the afternoon (both chosen uniformly and independently), and the two look the same. What
is the probability that they are actually the same person?
Proposed by: Yuan Yao
3
Answer:
17
Let X and Y be the students Nithya encounters during the day. The number of pairs (X, Y ) for which
X and Y look the same is 1 · 1 + 2 · 2 + . . . + 8 · 8 = 204, and these pairs include all the ones in which
X and Y are identical. As X and Y are chosen uniformly and independently, all 204 pairs are equally
likely to be chosen, thus the problem reduces to choosing one of the 36 pairs in 204, the probability
3
for which is 17 .
 9. [7] Let p be a real number between 0 and 1. Jocelin has a coin that lands heads with probability p and
tails with probability 1 − p; she also has a number written on a blackboard. Each minute, she flips the
coin, and if it lands heads, she replaces the number x on the blackboard with 3x + 1; if it lands tails
she replaces it with x/2. Given that there are constants a, b such that the expected value of the value
written on the blackboard after t minutes can be written as at + b for all positive integers t, compute
p.
Proposed by: Carl Schildkraut
1
Answer:
5
If the blackboard has the value x written on it, then the expected value of the value after one flip is

f (x) = p(3x − 1) + (1 − p)x/2.

Because this expression is linear, we can say the same even if we only know the blackboard’s initial
expected value is x. Therefore, if the blackboard value is x0 at time 0, then after t minutes, the
expected blackboard value is f t (x0 ). We are given that x0 , f (x0 ), f 2 (x0 ), . . . is an arithmetic sequence,
so for there to be a constant difference, we must have f (x) = x + c for some c.
This only occurs when 3p + (1 − p)/2 = 1, so p = 1/5.
10. [8] Let ABCD be a square of side length 5, and let E be the midpoint of side AB. Let P and Q be the
feet of perpendiculars from B and D to CE, respectively, and let R be the foot of the perpendicular
from A to DQ. The segments CE, BP, DQ, and AR partition ABCD into five regions. What is the
median of the areas of these five regions?
Proposed by: Carl Schildkraut
Answer: 5
We have DQ ⊥ CE and AR ⊥ DQ, so AR||CE. Thus, we can show that 4ARD ∼ = 4DQC ∼ = 4CP B,
so the median of the areas of the five regions is equal to the area of one of the three triangles listed
above.
BP
√ √
Now, note that 4EBC ∼ 4BP C, so BC = EB
EC =
√1 . This means that BP =
5
5, so CP = 2 5.
Therefore, the area of 4BP C, the median area, is 5.

11. [8] Let a, b, c, d be real numbers such that

min(20x + 19, 19x + 20) = (ax + b) − |cx + d|

for all real numbers x. Find ab + cd.

Proposed by: Yuan Yao
Answer: 380
In general, min(p, q) = p+q
2 −
p−q
2 . Letting p = 20x + 19 and q = 19x + 20 gives a = b = 19.5 and
c = d = ±0.5. Then the answer is 19.52 − 0.52 = 19 · 20 = 380.
12. [8] Four players stand at distinct vertices of a square. They each independently choose a vertex of the
square (which might be the vertex they are standing on). Then, they each, at the same time, begin
running in a straight line to their chosen vertex at 10mph, stopping when they reach the vertex. If
at any time two players, whether moving or not, occupy the same space (whether a vertex or a point
inside the square), they collide and fall over. How many different ways are there for the players to
choose vertices to go to so that none of them fall over?
Proposed by: Carl Schildkraut
Answer: 11
Observe that no two players can choose the same vertex, and no two players can choose each others
vertices. Thus, if two players choose their own vertices, then the remaining two also must choose their
 own vertices (because they cant choose each others vertices), thus all 4 players must choose their own
vertices. There is 1 way to choose the vertices in this case.
Name the players top left, top right, bottom left, and bottom right, based on their initial positions.
Assume exactly one player (without loss of generality, say the top left) chooses their own vertex. Then,
the remaining 3 players have to form a triangle (recall no two player can choose each others vertices).
There are 4 ways to choose which player chooses their own vertex, and 2 ways to choose which direction
the players move in the triangle, thus there are 8 ways to choose the vertices in this case.
Lastly, assume no one chooses their own vertex. We will first prove that no player can choose the
vertex across them. Assume the contrary, without loss of generality, let the top left player chooses the
bottom right vertex. Then, neither of the bottom left and the top right players can choose the others
vertex, because they would meet the top left player at the center of the square. As they cant choose
bottom right (it is chosen by the top left player), and cant choose their own vertex (by assumption),
they both have to choose the top left vertex, which is an immediate contradiction.
Now, the top left player has to choose either the top right vertex or the bottom left. Without loss
of generality, let the player choose the top right vertex. Then, the top right player has to choose the
bottom right vertex (as they can neither go across nor back to top left), the bottom right player has to
choose the bottom left vertex, and the bottom left player has to choose the top left vertex, and all the
choices are determined by the first players choice. There are 2 ways to choose where the first player
goes, thus there are 2 ways to choose the vertices in this case.
In total, there are 1 + 8 + 2 = 11 ways to choose the vertices.
13. [9] In 4ABC, the incircle centered at I touches sides AB and BC at X and Y , respectively. Addition-
ally, the area of quadrilateral BXIY is 25 of the area of ABC. Let p be the smallest possible perimeter
of a 4ABC that meets these conditions and has integer side lengths. Find the smallest possible area
of such a triangle with perimeter p.
Proposed by: Joey Heerens
√
Answer: 2 5
Note that ∠BXI = ∠BY I = 90, which means that AB and BC are tangent to the incircle of ABC
at X and Y respectively. So BX = BY = AB+BC−AC
2 , which means that 25 = [BXIY ] AB+BC−AC
[ABC] = AB+BC+AC .
The√smallest perimeter is achieved when AB = AC = 3 and BC = 4. The area of this triangle ABC
is 2 5.

14. [9] Compute the sum of all positive integers n for which
√ √ √
9 n + 4 n + 2 − 3 n + 16

is an integer.
Proposed by: Milan Haiman
Answer: 18
For the expression to be an integer at least one of n and n + 2 must be a perfect square. We also note
that at most one of n and n + 2 can be a square, so exactly one of them is a square.
√ √
Case 1: n is a perfect square. By our previous observation, it must be that 4 n + 2 = 3 n + 16 ⇒
n = 16.
√ √
Case 2: n + 2 is a perfect square. By our previous observation, it must be that 9 n = 3 n + 16 ⇒
n = 2.
Consequently, the answer is 16 + 2 = 18.
15. [9] Let a, b, c be positive integers such that

a b c
+ + = 1.
77 91 143
 What is the smallest possible value of a + b + c?
Proposed by: James Lin
Answer: 79
We need 13a + 11b + 7c = 1001, which implies 13(a + b + c − 77) = 2b + 6c. Then 2b + 6c must be
divisible by both 2 and 13, so it is minimized at 26 (e.g. with b = 10, c = 1). This gives a + b + c = 79.

16. [10] Equilateral 4ABC has side length 6. Let ω be the circle through A and B such that CA and CB
are both tangent to ω. A point D on ω satisfies CD = 4. Let E be the intersection of line CD with
segment AB. What is the length of segment DE?
Proposed by: Benjamin Qi
20
Answer:
13
Let F be the second intersection of line CD with ω. By power of a point, we have CF = 9, so DF = 5.
This means that [ADB] DE DE
[AF B] = EF = 5−DE . Now, note that triangle CAD is similar to triangle CF A,
FA CA 3 FB CB
so AD = CD = 2 . Likewise, BD = CD = 32 . Also, note that ∠ADB = 180 − ∠DAB − ∠DBA =
180 − ∠CAB = 120, and ∠AF B = 180 − ∠ADB = 60. This means that [ADB] AD·BD·sin 120 4
[AF B] = F A·F B·sin 60 = 9 .
DE
Therefore, we have that 5−DE = 49 . Solving yields DE = 1320
.
17. [10] Kelvin the frog lives in a pond with an infinite number of lily pads, numbered 0, 1, 2, 3, and so
forth. Kelvin starts on lily pad 0 and jumps from pad to pad in the following manner: when on lily
pad i, he will jump to lily pad (i + k) with probability 21k for k > 0. What is the probability that
Kelvin lands on lily pad 2019 at some point in his journey?
Proposed by: Nikhil Reddy
1
Answer:
2
Suppose we combine all of the lily pads with numbers greater than 2019 into one lily pad labeled ∞.
Also, let Kelvin stop once he reaches one of these lily pads.
Now at every leap, Kelvin has an equal chance of landing on 2019 as landing on ∞. Furthermore,
Kelvin is guaranteed to reach 2019 or ∞ within 2020 leaps. Therefore the chance of landing on 2019
is the same as missing it, so our answer is just 21 .

18. [10] The polynomial x3 − 3x2 + 1 has three real roots r1 , r2 , and r3 . Compute
√3
√ √
3r1 − 2 + 3 3r2 − 2 + 3 3r3 − 2.

Proposed by: Milan Haiman

Answer: 0
Let r be a root of the given polynomial. Then
√
r3 − 3r2 + 1 = 0 =⇒ r3 − 3r2 + 3r − 1 = 3r − 2 =⇒ r − 1 = 3
3r − 2.

Now by Vieta the desired value is r1 + r2 + r3 − 3 = 3 − 3 = 0.

19. [11] Let ABC be a triangle with AB = 5, BC = 8, CA = 11. The incircle ω and A-excircle1 Γ are
centered at I1 and I2 , respectively, and are tangent to BC at D1 and D2 , respectively. Find the ratio
of the area of 4AI1 D1 to the area of 4AI2 D2 .
Proposed by: Carl Schildkraut
1
Answer:
9
 Let D10 and D20 be the points diametrically opposite D1 and D2 on the incircle and A-excircle, respec-
tively. As Ix is the midpoint of Dx and Dx0 , we have

[AI1 D1 ] [AD1 D10 ]

= .
[AI2 D2 ] [AD2 D20 ]

Now, 4AD1 D10 and 4AD2 D20 are homothetic with ratio rrA = s−a
s , where r is the inradius, rA is the
A-exradius, and s is the semiperimeter. Our answer is thus
 2  
s−a 4 1
= = .
s 12 9

20. [11] Consider an equilateral triangle T of side length 12. Matthew cuts T into N smaller equilateral
triangles, each of which has side length 1, 3, or 8. Compute the minimum possible value of N .
Proposed by: Matthew Cho
Answer: 16
Matthew can cut T into 16 equilateral triangles with side length 3. If he instead included a triangle
of side 8, then let him include a triangles of side length 3. He must include 122 − 82 − 32 a = 80 − 9a
triangles of side length 1. Thus a ≤ 8, giving that he includes at least

(80 − 9a) + (a) + 1 = 81 − 8a ≥ 17

total triangles, so 16 is minimal.

21. [11] A positive integer n is infallible if it is possible to select n vertices of a regular 100-gon so that
they form a convex, non-self-intersecting n-gon having all equal angles. Find the sum of all infallible
integers n between 3 and 100, inclusive.
Proposed by: Benjamin Qi
Answer: 262
Suppose A1 A2 . . . An is an equiangular n-gon formed from the vertices of a regular 100-gon. Note that
the angle ∠A1 A2 A3 is determined only by the number of vertices of the 100-gon between A1 and A3 .
Thus in order for A1 A2 . . . An to be equiangular, we require exactly that A1 , A3 , . . . are equally spaced
and A2 , A4 , . . . are equally spaced. If n is odd, then all the vertices

must be equally spaced, meaning
n | 100. If n is even, we only need to be able to make a regular n2 -gon from the vertices of a 100-gon,
which we can do if n | 200. Thus the possible values of n are 4, 5, 8, 10, 20, 25, 40, 50, and 100, for a
total of 262.
22. [12] Let f (n) be the number of distinct digits of n when written in base 10. Compute the sum of f (n)
as n ranges over all positive 2019-digit integers.
Proposed by: Milan Haiman
9 102019 − 92019


Answer:
Write
f (n) = f0 (n) + · · · + f9 (n),
where fd (n) = 1 if n contains the digit d and 0 otherwise. The sum of fd (n) over all 2019-digit positive
integers n is just the number of 2019-digit positive integers that contain the digit d. For 1 ≤ d ≤ 9,
X
fd (n) = 9 · 102018 − 8 · 92018 .
n

Also, X
f0 (n) = 9 · 102018 − 92019 .
n
 Summing over all possible values of d, we compute

X 9 X
X
fd (n) = 9(9 · 102018 − 8 · 92018 ) + 9 · 102018 − 92019 = 9 102019 − 92019 .


f (n) =
n d=0 n

23. [12] For a positive integer n, let, τ (n) be the number of positive integer divisors of n. How many
integers 1 ≤ n ≤ 50 are there such that τ (τ (n)) is odd?
Proposed by: Kevin Liu
Answer: 17
Note that τ (n) is odd if and only if n is a perfect square. Thus, it suffices to find the number of integers
n in the given range such that τ (n) = k 2 for some positive integer k.
If k = 1, then we obtain n = 1 as our only solution. If k = 2, we see that n is either in the form pq
or p3 , where p and q are distinct primes. The first subcase gives 8 + 4 + 1 = 13 solutions, while the
second subcase gives 2 solutions. k = 3 implies that n is a perfect square, and it is easy to see that
only 62 = 36 works. Finally, k ≥ 4 implies that n is greater than 50, so we’ve exhausted all possible
cases. Our final answer is 1 + 13 + 2 + 1 = 17.
24. [12] Let P be a point inside regular pentagon ABCDE such that ∠P AB = 48◦ and ∠P DC = 42◦ .
Find ∠BP C, in degrees.
Proposed by: Dylan Liu
Answer: 84◦
Since a regular pentagon has interior angles 108◦ , we can compute ∠P DE = 66◦ , ∠P AE = 60◦ , and
∠AP D = 360◦ −∠AED−∠P DE −∠P AE = 126◦ . Now observe that drawing P E divides quadrilateral
P AED into equilateral triangle P AE and isosceles triangle P ED, where ∠DP E = ∠EDP = 66◦ . That
is, we get P A = P E = s, where s is the side length of the pentagon.
Now triangles P AB and P ED are congruent (with angles 48◦ −66◦ −66◦ ), so P D = P B and ∠P DC =
∠P BC = 42◦ . This means that triangles P DC and P BC are congruent (side-angle-side), so ∠BP C =
∠DP C.
Finally, we compute ∠BP C + ∠DP C = 2∠BP C = 360◦ − ∠AP B −∠EP A − ∠DP E = 168◦ , meaning
∠BP C = 84◦ .
25. [13] In acute 4ABC with centroid G, AB = 22 and AC = 19. Let E and F be the feet of the altitudes
from B and C to AC and AB respectively. Let G0 be the reflection of G over BC. If E, F , G, and G0
lie on a circle, compute BC.
Proposed by: Milan Haiman
Answer: 13
0
Note that B, C, E, F lie on a circle. Moreover, since BC bisects GG , the center of the circle that goes
0 0
through E, F, G, G must lie on BC. Therefore, B, C, E, F, G, G lie on a circle. Specifically, the center
of this circle is M , the midpoint of BC, as M E = M F because M is the center of the circumcircle
of BCEF . So we have GM = BC 3BC
2 , which gives AM = 2 . Then, by Apollonius’s theorem, we have
2 2 2 2 2
AB + AC = 2(AM + BM ). Thus 845 = 5BC and BC = 13.
26. [13] Dan is walking down the left side of a street in New York City and must cross to the right side
at one of 10 crosswalks he will pass. Each time he arrives at a crosswalk, however, he must wait t
seconds, where t is selected uniformly at random from the real interval [0, 60] (t can be different at
different crosswalks). Because the wait time is conveniently displayed on the signal across the street,
Dan employs the following strategy: if the wait time when he arrives at the crosswalk is no more than
k seconds, he crosses. Otherwise, he immediately moves on to the next crosswalk. If he arrives at the
last crosswalk and has not crossed yet, then he crosses regardless of the wait time. Find the value of
k which minimizes his expected wait time.
   19 !
1
Answer: 60 1 −
10

k 9


With probability 1 − 60 ,
Dan reaches the last crosswalk without crossing at any previous site, in
k 9


which case the expected value of his wait time is 30 seconds. Otherwise, with probability 1 − 1 − 60 ,
Dan crosses at an earlier crosswalk, in which case the expected value of his wait time is k2 . We want
to find the k that minimizes
 9  9 !    9 !
k k k k k
30 1 − + 1− 1− = 30 − 30 − 1− 1−
60 2 60 2 60

k
Letting a = 1 − 60 , we can use weighted AM-GM:

1 9 1 9 9
9 10 a 1 − a9 = 9a9 1 − a9


10
10
10
≤
10
1

19 
1

91 
where equality occurs when 9a9 = 1 − a9 , or a = 10 , meaning that k = 60 1 − 10 . Because
our original expression can be written as

30 − 30a(1 − a9 ),

1 
1 9
the minimum occurs at the same value, k = 60 1 − 10 .

27. [13] For a given positive integer n, we define ϕ(n) to be the number of positive integers less than or
equal to n which share no common prime factors with n. Find all positive integers n for which

ϕ(2019n) = ϕ(n2 ).

Proposed by: Carl Schildkraut

Answer: 1346, 2016, 2019
Let p1 , p2 , . . . , pk be the prime divisors of n. Then it is known that ϕ(n) = n · p1p−1
1
. . . pkp−1
k
. As n2 and
2 2 p1 −1 pk −1
n has the same set of prime divisors, it also holds that ϕ(n ) = n · p1 . . . pk . We will examine the
equality in four cases.
• gcd(n, 2019) = 1 In this case, 2019 · n has also 3 and 673 as prime divisors, thus ϕ(2019 · n) =
2019 · n · p1p−1
1
. . . pkp−1
k
· 23 · 673
672
, and the equality implies n = 1342, however gcd(1342, 3) 6= 1,
contradiction. Thus, there is no answer in this case.
• gcd(n, 2019) = 3 In this case, 2019 · n has also 673 as a prime divisor, thus ϕ(2019 · n) =
2019 · n · p1p−1
1
. . . pkp−1
k
· 672
673 , and the equality implies n = 2016, which satisfies the equation. Thus,
the only answer in this case is n = 2016.
• gcd(n, 2019) = 673 In this case, 2019 · n has also 3 as a prime divisor, thus ϕ(2019 · n) =
2019 · n · p1p−1
1
. . . pkp−1
k
· 23 , and the equality implies n = 1346, which satisfies the equation. Thus,
the only answer in this case is n = 1346.
• gcd(n, 2019) = 2019 In this case, 2019 · n has the same set of prime divisors, thus ϕ(2019 · n) =
2019 · n · p1p−1
1
. . . pkp−1
k
, and the equality implies n = 2019, which satisfies the equation. Thus, the
only answer in this case is n = 2019.
Thus, all the answers are n = 1346, 2016, 2019.
28. [15] A palindrome is a string that does not change when its characters are written in reverse order.
Let S be a 40-digit string consisting only of 0’s and 1’s, chosen uniformly at random out of all such
strings. Let E be the expected number of nonempty contiguous substrings of S which are palindromes.
Compute the value of bEc.
 Proposed by: Benjamin Qi
Answer: 113
Note that S has 41 − n contiguous substrings of length n, so we see that the expected number of
palindromic substrings of length n is just (41 − n) · 2−bn/2c . By linearity of expectation, E is just the
sum of this over all n from 1 to 40. However, it is much easier to just compute
∞
X
(41 − n) · 2−bn/2c .
n=1

The only difference here is that we have added some insignificant negative terms in the cases where
n > 41, so E is in fact slightly greater than this value (in fact, the difference between E and this sum
7
is 1048576 ). To make our infinite sum easier to compute, we can remove the floor function by pairing
up consecutive terms. Then our sum becomes
∞
X 81 − 4n
40 + ,
n=1
2n

which is just 40 + 81 − 8 = 113. E is only slightly larger than this value, so our final answer is
bEc = 113 .
√
29. [15] In isosceles 4ABC, AB = AC and P is a point on side BC. If ∠BAP = 2∠CAP , BP = 3,
and CP = 1, compute AP .
Proposed by: Milan Haiman
√
Answer: 2
√ √
3
Let ∠CAP = α, By the Law of Sines, sin 2α = sin1 α which rearranges to cos α = 23 ⇒ α = π6 . This
π
√ √
implies that ∠BAC = 2 . By the Pythagorean Theorem, 2AB 2 = ( 3 + 1)2 , so AB 2 = 2 + 3.
√ √ √ √
Applying Stewart’s Theorem, it follows that AP 2 = ( 3+1)(2+
√
3+1
3)
− 3 ⇒ AP = 2.

30. [15] A function f : Z → Z satisfies: f (0) = 0 and

f ((n + 1)2k ) − f (n2k ) ≤ 1

for all integers k ≥ 0 and n. What is the maximum possible value of f (2019)?
Proposed by: Krit Boonsiriseth
Answer: 4
Consider a graph on Z with an edge between (n + 1)2k and n2k for all integers k ≥ 0 and n. Each
vertex m is given the value f (m). The inequality |f ((n + 1)2k ) − f (n2k )| ≤ 1 means that any two
adjacent vertices of this graph must have values which differ by at most 1. Then it follows that for all
m,
f (m) ≤ number of edges in shortest path from 0 to m
because if we follow a path from 0 to m, along each edge the value increases by at most 1. Conversely,
if we define f (m) to be the number of edges in the shortest path between 0 and m, then this is a valid
function because for any two adjacent vertices, the lengths of their respective shortest paths to 0 differ
by at most 1. Hence it suffices to compute the distance from 0 to 2019 in the graph.
There exists a path with 4 edges, given by

0 → 2048 → 2016 → 2018 → 2019.

Suppose there existed a path with three edges. In each step, the number changes by a power of 2, so
we have 2019 = ±2k1 ± 2k2 ± 2k3 for some nonnegative integers k1 , k2 , k3 and choice of signs. Since
2019 is odd, we must have 20 somewhere. Then we have ±2k1 ± 2k2 ∈ {2018, 2020}. Without loss of
 generality assume that k1 ≥ k2 . Then we can write this as ±2k2 (2k1 k2 ± 1) ∈ {2018, 2020}. It is easy
to check that k1 = k2 is impossible, so the factorization 2k2 (2k1 k2 ± 1) is a product of a power of two
and an odd number. Now compute 2018 = 2 × 1009 and 2020 = 4 × 505. Neither of the odd parts are
of the form 2k1 −k2 ± 1, so there is no path of three steps.
We conclude that the maximum value of f (2019) is 4.
31. [17] James is standing at the point (0, 1) on the coordinate plane and wants to eat a hamburger. For
each integer n ≥ 0, the point (n, 0) has a hamburger with n patties. There is also a wall at y = 2.1
which James cannot cross. In each move, James can go either up, right, or down 1 unit as long as he
does not cross the wall or visit a point he has already visited.
Every second, James chooses a valid move uniformly at random, until he reaches a point with a
hamburger. Then he eats the hamburger and stops moving. Find the expected number of patties that
James eats on his burger.
Proposed by: Joey Heerens
7
Answer:
3
Note that we desire to compute the number of times James moves to the right before moving down to
the line y = 0. Note also that we can describe James’s current state based on whether his y-coordinate
is 0 or 1 and whether or not the other vertically adjacent point has been visited. Let E(1, N ) be the
expected number of times James will go right before stopping if he starts at a point with y-coordinate 1
and the other available point with the same x-coordinate has not been visited. Define E(1, Y ), E(2, N ),
and E(2, Y ) similarly. Then we can construct equations relating the four variables:
1 1
E(1, N ) = E(2, Y ) + (E(1, N ) + 1),
3 3
as James can either go up, right, or down with probability 1/3 each if he starts in the state (1, N ).
Similarly, we have
1 1 1
E(2, N ) = E(1, Y ) + (E(2, N ) + 1), E(1, Y ) = (E(1, N ) + 1),
2 2 2
and E(2, Y ) = E(2, N )+1. Solving these equations, we get E(1, N ) = 37 , which is our answer, as James
starts in that state having gone left 0 times.
32. [17] A sequence of real numbers a0 , a1 , . . . , a9 with a0 = 0, a1 = 1, and a2 > 0 satisfies

an+2 an an−1 = an+2 + an + an−1

for all 1 ≤ n ≤ 7, but cannot be extended to a10 . In other words, no values of a10 ∈ R satisfy

a10 a8 a7 = a10 + a8 + a7 .

Compute the smallest possible value of a2 .

Proposed by: Dylan Liu
√
Answer: 2−1
a+1 −a+1
Say a2 = a. Then using the recursion equation, we have a3 = −1, a4 = a−1 , a5 = a+1 , a6 = − a1 ,
a7 = − a22a
−1 , and a8 = 1.
Now we have a10 a8 a7 = a10 + a8 + a7 . No value of a10 can satisfy this equation iff a8 a7 = 1 and
a8 + a7 6= 0. Since a8 is 1, we want 1 = a7 = − a22a 2
−1 , which gives a + 2a − 1 = 0. The only positive
√
root of this equation is 2 − 1.
This problem can also be solved by a tangent substitution. Write an = tan αn . The given condition
becomes
αn+2 + αn + αn−1 = 0.
 We are given α0 = 0, α1 = π/4, and α2 ∈ (0, π/2). Using this, we can recursively compute α3 , α4 , . . .
in terms of α2 until we get to α10 = 3π 4 − 2α2 . For a10 not to exist, we need α ≡ π/2 mod π. The
√ 10
only possible value of α2 ∈ (0, π/2) is α2 = π/8, which gives a2 = tan π/8 = 2 − 1.
33. [17] A circle Γ with center O has radius 1. Consider pairs (A, B) of points so that A is inside the
circle and B is on its boundary. The circumcircle Ω of OAB intersects Γ again at C 6= B, and line AC
intersects Γ again at X 6= C. The pair (A, B) is called techy if line OX is tangent to Ω. Find the area
of the region of points A so that there exists a B for which (A, B) is techy.
Proposed by: Carl Schildkraut and Milan Haiman
3π
Answer:
4
We claim that (A, B) is techy if and only if OA = AB.
Note that OX is tangent to the circle (OBC) if and only if OX is perpendicular to the angle bisector
of ∠BOC, since OB = OC. Thus (A, B) is techy if and only if OX is parallel to BC. Now since
OC = OX,

OX k BC ⇐⇒ ∠BCA = ∠OXA ⇐⇒ ∠BCA = ∠ACO ⇐⇒ OA = AB.

From the claim, the desired region of points A is an annulus between the circles centered at O with
radii 21 and 1. So the answer is 3π
4 .

34. [20] A polynomial P with integer coefficients is called tricky if it has 4 as a root.
A polynomial is called k-tiny if it has degree at most 7 and integer coefficients between −k and k,
inclusive.
A polynomial is called nearly tricky if it is the sum of a tricky polynomial and a 1-tiny polynomial.
Let N be the number of nearly tricky 7-tiny polynomials. Estimate N .
j
k
E 4
An estimate of E will earn 20 min N ,
E N points.
Proposed by: Carl Schildkraut
Answer: 64912347
A tricky 7-tiny polynomial takes the form

(c6 x6 + . . . + c1 x + c0 )(x − 4).

For each fixed value of k, ck − 4ck+1 should lie in [−7, 7], so if we fix ck , there are around 15/4 ways
of choosing ck+1 . Therefore if we pick c0 , . . . , c6 in this order, there should be around (15/4)7 tricky
7-tiny polynomials.
A 1-tiny polynomial takes the form ε6 x7 + · · · + ε1 x + ε0 with εi ∈ {−1, 0, +1}, so there are 38 1-tiny
polynomials.
A nearly tricky 7-tiny polynomial P takes the form Q+T where Q is roughly a tricky 7-tiny polynomial,
and T is 1-tiny. Furthermore, there Pis a unique decomposition Q + T because T (4) = P (4) and each
integer n can be written in the form εk 4k in at most one way. Therefore the number of nearly tricky
7 8
7-tiny is around (15/4) · 3 ≈ 68420920, which is worth 16 points.
The exact answer can be found by setting up recurrences. Let t(d, `) be the number of polynomials of
degree at most i of the form

(`xd−1 + cd−2 xd−2 + · · · + c0 )(x − 4) + (εd−1 xd−1 + · · · + ε1 x + ε0 ).

which has integer coefficients between −7 and 7 except the leading term `xd . It follows that t(0, 0) =
1, t(0, k) = 0 for all k 6= 0, and t(d + 1, `) can be computed as follows: for each value of cd−1 , there are
 t(d, cd−1 ) ways to pick cd−2 , . . . , c0 , εd−1 , . . . , ε0 , and exactly w(cd−1 − 4`) ways of picking εd , where
w(k) = min(9 − |k|, 3) for |k| ≤ 8 and 0 otherwise. Therefore setting c = cd−1 − 4` we have
8
X
t(d + 1, `) = t(d, c + 4`)w(c).
c=−8

The number of nearly tricky 7-tiny polynomials is simply t(8, 0), which can be computed to be 64912347
using the following C code.

int w(int a){

if(a < -9 || a > 9) return 0;
else if(a == -8 || a == 8) return 1;
else if(a == -7 || a == 7) return 2;
else return 3;
}

int main()
{
int m=8,n=7,r=4,d,l,c,c4l;
int mid = 2 + n/r;
int b = 2*mid+1;
long int t[500][500];
for(l=0; l<b; l++){
t[0][l] = (l == mid) ? 1 : 0;
}
for(d=0; d<m+1; d++){
for(l=0; l<b; l++){
t[d+1][l] = 0;
for(c=-8; c<9; c++){
c4l = c + 4*(l-mid) + mid;
t[d+1][l] += (c4l >= 0 && c4l <= 2*mid) ? t[d][c4l]*w(c) : 0;
}
}
}
printf("%ld",t[8][mid]);
}

35. [20] You are trying to cross a 400 foot wide river. You can jump at most 4 feet, but you have many
stones you can throw into the river. You will stop throwing stones and cross the river once you have
placed enough stones to be able to do so. You can throw straight, but you can’t judge distance very
well, so each stone ends up being placed uniformly at random along the width of the river. Estimate
the expected number N of stones you must throw before you can get across the river.
j
k
E 3
An estimate of E will earn 20 min N ,
E N points.
Proposed by: Carl Schildkraut and Milan Haiman
Answer: 712.811
If we divide the river into 100 4-foot sections, then to be able to cross we need to get at least one stone
into each section. On average, this takes
100 100 100
+ + ··· + ≈ 100 ln 100
100 99 1
100
stone throws (it takes 100−k moves on average to get a stone into a new section if k sections already
have a stone). So the answer is at least 100 ln 100 ≈ 450.
 On the other hand, if we divide the river into 200 2-foot sections, then once we have a stone in each
section we are guaranteed to be able to cross. By a similar argument, we obtain that the answer is at
most 200 ln 200 ≈ 1050.
Estimates near these bounds earn about 5 to 7 points. An estimate in between can earn close to 20
points.
To compute the answer (almost) exactly, we use the following argument.
Scale the problem so the river is of size 1, and the jumps are of size 0.01. Suppose that after n throws,
the stones thrown are located at positions 0 < x1 < x2 < · · · < xn < 1. Let x0 = 0, xn+1 = 1, r = 0.01.
Define P (n) to be the probability that you still cannot cross
P∞ the river after n throws. In other words,
there exists i such that xi+1 − xi > r. Then our answer is n=0 P (n).
By PIE we can write
∞  
i−1 n + 1
X
P (n) = (−1) max(1 − ir, 0)n .
i=1
i
based on which intervals xi+1 − xi have length greater than r. Now we switch the order of summation:
∞ ∞ X ∞   ∞ ∞  
X X n+1 X X n+1
P (n) = (−1)i−1 max(1 − ir, 0)n = (−1)i−1 max(1 − ir, 0)n .
n=0 n=0 i=1
i i=1 n=0
i

Let x = max(1 − ir, 0). Then

∞  ∞ 
xi−1
 
X n+1 X i+j
xn = xi−1 xj = .
n=0
i j=0
i (1 − x)i+1

Thus, our answer is

b1/rc
X (1 − ir)i−1
(−1)i−1 ≈ 712.811,
i=1
(ir)i+1
where the last approximation uses the C++ code below.

#include <bits/stdc++.h>

using namespace std;

typedef long double ld;

int main() {
ld sum = 0, r = 0.01;
for (int i = 1; ; ++i) {
ld x = 1-r*i; if (x <= 0) break;
ld ex = pow(x/(1-x),i-1)/(1-x)/(1-x);
if (i&1) sum += ex;
else sum -= ex;
}
cout << fixed << setprecision(8) << sum;
}

36. [20] Let N be the number of sequences of positive integers (a1 , a2 , a3 , . . . , a15 ) for which the polynomials

x2 − ai x + ai+1

each have an integer root for every 1 ≤ i ≤ 15, setting a16 = a1 . Estimate N .
j
k
E 2
An estimate of E will earn 20 min N ,
E N points.
Proposed by: Krit Boonsiriseth
Answer: 1409
We note that ai+1 = x(ai − x) for some positive integer x, so ai+1 ≥ ai − 1. So, the only way ai can
decrease is decreasing by 1.
As it cannot decrease that quickly, we will make the assumption that if ai ≥ 10, ai+1 = ai − 1, as
otherwise it will increase at least above 16 at which point it will take many moves to go back down
below 10. Write that a → b if b is a possible value of ai+1 given a = ai . We have

5 → 6, 6 → 5, 8, 9, 7 → 6, 8 → 7, 9 → 8,

and in addition by going to 10 and above, 7 can go to 9 in 2 or 4 steps, 8 can in 4, 7, 8 steps, and
9 can in 6, 10, 12 steps. We see from this that the vast majority of sequences should pass through 8.
By looking at cycles from 8, we can determine exactly when a sequence can start at 8 and return to
8 (there is one way in 3 steps, two in 4 steps, etc.), and from there we can generate a list of types of
sequences by when 8s occur. By dividing by the number of 8s and multiplying by 15, we can get the
number of sequences that include 8, which gives us an estimate of 1235, giving us 15 points. As we
note that this is a lower estimate, we may round up slightly to get better results.
To find the exact answer, we will first show that no element larger than 32 can occur in the sequence.
Reorder the sequence to make a1 maximal; we have

ai+1 ≥ ai − 1 =⇒ a15 ≥ a1 − 14.

Also, since a1 > a15 , a1 ≥ 2a15 − 4, giving

a1 + 4
a1 − 14 ≤ =⇒ a1 ≤ 32.
2
We then construct the following Python code:

def p36(max_val,length):
L=[[i] for i in range(1,max_val+1)]
for j in range(length):
newL=[]
for k in L:
poss=[x*(k[-1]-x) for x in range(1,k[-1]//2+1)]
for t in poss:
if 1<=t<=max_val:
newL.append(k+[t])
L=newL
return len(L)

print(p36(32,15))

This gives the exact answer of 1409.