HMMT November 2022

November 12, 2022

Team Round
1. [20] Two linear functions f (x) and g(x) satisfy the properties that for all x,
• f (x) + g(x) = 2
• f (f (x)) = g(g(x))
and f (0) = 2022. Compute f (1).
Proposed by: Vidur Jasuja
Answer: 1
Solution 1: Firstly, f (x) and g(x) must intersect – otherwise, f (x) = g(x) = 1, which can’t be true.
Secondly, suppose they intersect at a, so that f (a) = g(a) = c. Then f (c) = g(c) =⇒ f (c) = g(c) = 1.
But then, a = c, and c = 1. So f (1) = 1, and we’re done.

Solution 2: We will solve the problem manually, setting f (x) = ax + 2022 and g(x) = −ax − 2020.
Then f (f (x)) = a2 x + 2022a + 2022, while g(g(x)) = a2 x + 2020a − 2020.
Then, a = −2021, so then f (1) = 1.

2. [25] What is the smallest r such that three disks of radius r can completely cover up a unit disk?
Proposed by: Gabriel Wu
√
3
Answer: 2

Solution: Look at the circumference of the unit disk. Each of the disks must be capable of √ covering
up at least 31 of the circumference, which means it must be able to cover a chord of length 3. Thus,
√
3
2 is a lower bound for r. This bound is achievable: place the three centers of the disks symmetrically
at a distance of 12 from the center of the unit disk.

3. [30] Find the number of ordered pairs (A, B) such that the following conditions hold:
• A and B are disjoint subsets of {1, 2, . . . , 50}.
• |A| = |B| = 25
• The median of B is 1 more than the median of A.
Proposed by: Papon Lapate
24 2


Answer: 12

Solution: The median of both sets, which we will call a and b respectively, are more than exactly 12
of the members in their own set. Since a and b are consecutive, they must also be higher than the
lower half of the other set and lower than the higher half of the other set, meaning that they are both
higher than exactly 24 numbers in {1, 2, ..., 50} − {a, b}. Thus, a = 25 and b = 26.
The 24 lower numbers can be divided into the two groups (with 12 in each group) in 24


12 ways. Simi-
24 2
larly, the 24 higher numbers can be divided into the two groups in 24



12 ways. Thus, the answer is 12 .
4. [35] You start with a single piece of chalk of length 1. Every second, you choose a piece of chalk that
you have uniformly at random and break it in half. You continue this until you have 8 pieces of chalk.
What is the probability that they all have length 81 ?
Proposed by: Evan Erickson, Gabriel Wu
1
Answer: 63

Solution 1: There are 7! total ways to break the chalks. How many of these result in all having length
1
8 ? The first move gives you no choice. Then, among the remaining 6 moves, you must apply 3 breaks
6


on the left side and 3 breaks on the right side, so there are 3 = 20 ways to order those. On each side,
you can either break the left side or the right side first. So the final answer is

20 · 22 1
= .
7! 63

Solution 2: We know there are 7! ways to break the chalk in total.

Now, if we break up the chalk into 8 pieces, we can visualize the breaks as a binary decision tree.
Each round we select a node and break that corresponding piece of chalk, expanding it into two branch
nodes. The final tree of our desired configuration will have three layers.
We can figure out how many different ordering we can do this in with recursion. If bn is the number
of ways to expand a binary tree with n layers, we have b1 = 1. Now when we expand a node with
k + 1 layers, we will expand either the k-layered tree on the left or right, these moves can be ordered
k+1
−2
in 22k −1


ways. For each one of these trees, there are bk ways to decide these moves. So we have
2k+1 −2 2
bk+1 = 2k −1 bk . So b2 = 21 · 12 = 2, b3 = 63 · 22 = 20 · 22 . Thus, the final answer is

20 · 22 1
= .
7! 63

5. [40] A triple of positive integers (a, b, c) is tasty if lcm(a, b, c) | a + b + c − 1 and a < b < c. Find the
sum of a + b + c across all tasty triples.
Proposed by: Isaac Zhu
Answer: 44
Solution: The condition implies c | b + a − 1. WLOG assume c > b > a; since b + a − 1 < 2c we must
have b + a − 1 = c. Substituting into b | a + c − 1 and a | c + b − 1 gives

b | 2a − 2

a | 2b − 2.
Since 2a − 2 < 2b we must either have a = 1 (implying a = b, bad) or 2a − 2 = b =⇒ a | 4a − 6 =⇒
a = 2, 3, 6. If a = 2 then b = 2. Otherwise, if a = 3 we get (3, 4, 6) and if a = 6 we get (6, 10, 15), so
answer is 13 + 31 = 44.

6. [45] A triangle XY Z and a circle ω of radius 2 are given in a plane, such that ω intersects segment XY
at the points A, B, segment Y Z at the points C, D, and segment ZX at the points E, F . Suppose that
XB > XA, Y D > Y C, and ZF > ZE. In addition, XA = 1, Y C = 2, ZE = 3, and AB = CD = EF .
Compute AB.
Proposed by: Rishabh Das
 √
Answer: 10 − 1
Solution: Let d = AB and x = d/2 for ease of notation. Let the center of (ABCDEF ) be I. Because
AB = CD = EF , the distance from I to AB, CD, and EF are the same, so I is the incenter of
△XY Z. Let △XY Z have inradius r.
By symmetry, we have XF = 1, Y B = 2, and ZD = 3. Thus, △XY Z has side lengths d + 3, d + 4,
and d + 5. Heron’s Formula gives the area of △XY Z is
p p
K = (3x + 6) (x + 2) (x + 1) (x + 3) = (x + 2) 3 (x + 1) (x + 3),
while K = rs gives the area of △XY Z as
K = (3x + 6) r.
Equating our two expressions for K, we have
p p
(x + 2) 3 (x + 1) (x + 3) = (3x + 6) r =⇒ 3 (x + 1) (x + 3) = 3r =⇒ (x + 1) (x + 3) = 3r2 .
The Pythagorean Theorem gives x2 + r2 = 4, so r2 = 4 − x2 . Plugging this in and expanding gives
(x + 1) (x + 3) = 3 4 − x2 =⇒ 4x2 + 4x − 9 = 0.


√ √
This has roots x = −1±2 10 , and because x > 0, we conclude that d = 10 − 1.

7. [45] Compute the number of ordered pairs of positive integers (a, b) satisfying the equation
gcd(a, b) · a + b2 = 10000.

Proposed by: Vidur Jasuja

Answer: 99
Solution 1: Let gcd(a, b) = d, a = da′ , b = db′ . Then, d2 (a′ + b′2 ) = 1002 . Consider each divisor d of
2
100. Then, we need to find the number of solutions in coprime integers to a′ + b′2 = 100 d2 . Note that
2
every b′ < 100/d coprime to 100 100
d2 satisfies this equation, which is equivalent to being coprime to d , so
100


then there are φ d choices for
each d, except for d = 100, which would count the solution (0, 100).
Then we just need d|n φ 100
P
d − 1 = 100 − 1 = 99.

Solution 2: Note that b must be at most 99 in order for a to be positive. Now we claim that each
choice of b ∈ {1, 2, . . . , 99} corresponds to exactly one value of a that satisfies the equation.
To see why this is true, we rewrite the formula as
gcd(a, b) · a = (100 − b)(100 + b).
For any prime p, let vp (n) be the largest integer k such that pk |n. We wish to show that once we fix b,
the value vp (a) is uniquely determined for all p (which will give us a unique solution for a). Applying
vp to both sides of the equation gives us
min(vp (b), vp (a)) + vp (a) = vp (100 − b) + vp (100 + b).
We immediately see that there is at most one solution for vp (a) since the left-hand side increases with
vp (a). Further, as we increment vp (a), the left-hand side takes on all even numbers up to 2vp (b),
and then all integers larger than 2vp (b). So to show that a solution exists, we need to prove that the
right-hand side is always either even or at least 2vp (b).
If vp (b) ̸= vp (100), then vp (100 − b) = vp (100 + b) = min(vp (100), vp (b)). In this case, vp (100 −
b) + vp (100 + b) is clearly even. Otherwise, assume vp (b) = vp (100). Then vp (100 − b) ≥ vp (b) and
vp (100 + b) ≥ vp (b), so vp (100 − b) + vp (100 + b) ≥ 2vp (b) as desired. Thus, there is always a unique
solution for vp (a). Once we fix 1 ≤ b ≤ 99, the value of a is uniquely determined, so the answer is 99.
8. [50] Consider parallelogram ABCD with AB > BC. Point E on AB and point F on CD are marked
such that there exists a circle ω1 passing through A, D, E, F and a circle ω2 passing through B, C, E, F .
If ω1 , ω2 partition BD into segments BX, XY , Y D in that order, with lengths 200, 9, 80, respectively,
compute BC.
Proposed by: Albert Wang
Answer: 51
Solution: We want to find AD = BC = EF . So, let EF intersect BD at O. It is clear that
△BOE ∼ △DOF . However, we can show by angle chase that △BXE ∼ △DY F :

∠BEG = ∠ADG = ∠CBH = ∠DF H.

This means that EF partitions BD and XY into the same proportions, i.e. 200 to 80. Now, let
a = 200, b = 80, c = 9 to make computation simpler. O is on the radical axis of ω1 , ω2 and its power
respect to the two circles can be found to be
 
ac bc abc(a + b + c)
a+ =
a+b a+b (a + b)2
c(a+b+c)
However, there is now x for which OE = ax, OF = bx by similarity. This means x2 = (a+b)2 .
Notably, we want to find (a + b)x, which is just
p √
c(a + b + c) = 9 · 289 = 51.

9. [50] Call an ordered pair (a, b) of positive integers fantastic if and only if a, b ≤ 104 and

gcd(a · n! − 1, a · (n + 1)! + b) > 1

for infinitely many positive integers n. Find the sum of a + b across all fantastic pairs (a, b).
Proposed by: Pitchayut Saengrungkongka
Answer: 5183
Solution: We first prove the following lemma, which will be useful later.
Lemma: Let p be a prime and 1 ≤ n ≤ p − 1 be an integer. Then, n!(p − 1 − n)! ≡ (−1)n−1 (mod p).

Proof. Write

n!(p − n − 1)! = (1 · 2 · · · n)((p − n − 1) · · · 2 · 1)

≡ (−1)p−n−1 (1 · 2 · · · n)((n + 1) · · · (p − 2)(p − 1)) (mod p)
n
= (−1) (p − 1)!
≡ (−1)n−1 (mod p)

(where we have used Wilson’s theorem). This implies the result.

Now, we begin the solution. Suppose that a prime p divides both a · n! − 1 and a · (n + 1)! + b. Then,
since
−b ≡ a · (n + 1)! ≡ (n + 1) · (a · n!) ≡ (n + 1) (mod p),
we get that p | n + b + 1. Since we must have n < p (or else p | n!), we get that, for large enough n,
n = p − b − 1. However, by the lemma,

a(−1)b−1 ≡ a · b!(p − 1 − b)! = a · b!n! ≡ b! (mod p).

 This must hold for infinitely many p, so a = (−1)b−1 b!. This forces all fantastic pairs to be in form
((2k − 1)!, 2k − 1).
Now, we prove that these pairs all work. Take n = p − 2k for all large primes p. Then, we have
a · n! ≡ (2k − 1)!(p − 2k)!
≡ (−1)2k ≡ 1 (mod p)
a · (n + 1)! ≡ (n + 1) · (a · n!)
≡ (p − 2k + 1) · 1 ≡ −(2k − 1) (mod p),
so p divides the gcd.
The answer is (1 + 1) + (6 + 3) + (120 + 5) + (5040 + 7) = 5183.

10. [60] There is a unit circle that starts out painted white. Every second, you choose uniformly at random
an arc of arclength 1 of the circle and paint it a new color. You use a new color each time, and new
paint covers up old paint. Let cn be the expected number of colors visible after n seconds. Compute
limn→∞ cn .
Proposed by: Gabriel Wu
Answer: 4π
Solution 1: Notice that colors always appear in contiguous arcs on the circle (i.e. there’s never a
color that appears in two disconnected arcs). So the number of distinct visible colors is equal to the
number of radii that serve as boundaries between colors. Each time we place a new color, we create 2
1
more of these radii, but all of the previous radii have a p = 2π chance of being covered up.
It is well-known that, given an event that occurs with probability p, it takes an expected p1 trials for
it to happen – this means that any given radii has an expected lifespan of 2π before it gets covered
up (i.e. it remains visible for 2π turns on average). Thus, over the course of n ≫ 1 turns there are 2n
total radii created, each lasting an average of 2π turns, so there are 4πn
n = 4π radii visible on average
each turn.
A more rigorous way to see this is the two radii created on the most recent turn have a probability 1
of being exposed; the two radii created last turn each have a probability 1 − p of being exposed; the
two radii created two turns ago each have a probability (1 − p)2 of being exposed, and so on. Thus,
on turn n, the expected number of exposed radii is
2(1 + (1 − p) + (1 − p)2 + · · · + (1 − p)n−1 ).
2
This geometric series converges to p as n grows.

Solution 2: Notice that colors always appear in contiguous arcs on the circle (i.e. there’s never a
color that appears in two disconnected arcs). So the number of distinct visible colors is equal to the
number of radii that serve as boundaries between colors. Each time we place a new color, we create
2 more of these radii. However, if the expected number of colors after turn n is constant, we must
expect to remove 2 of these radii each turn. This is only possible when there are 4π total radii, since
1
we expect to remove 2π of them each time.

Solution 3: Consider the probability that the k-th last added arc is visible. Suppose there are j arcs
after the k-th last arc that partially covers this arc. Then the probability that the k-th last arc is
still visible is j+1
2j , since this is equivalent to randomly choosing j positions within the k-th last arc to
place an arc in, then randomly choosing a direction, and there are 2j ways to choose directions and
2
j + 1 of them are good. The probability that any arc partially covers the k-th last arc is 2π . Putting
everything together, the probability that the k-th last arc is visible is
k−1
X j + 1 k − 1   2 j  k−1−j
2
· · · 1 −
j=0
2j j 2π 2π
so the answer is
n k−1
X j + 1 k − 1  2 j  k−1−j
X 2
1+ · · · 1−
j=0
2j j 2π 2π
k=2

(as the last arc is definitely visible).

We can write this as
n k−1    j  k−1−j
X X k−1 1 2
1+ (j + 1) · · · 1− .
j 2π 2π
k=2 j=0

Now,
k−1    j  k−1−j  k−1
X k−1 1 2 1 2
1· · · 1− = +1−
j=0
j 2π 2π 2π 2π
k−1 k−2



by binomial theorem. We can write j · j = (k − 1) j−1 , so

k−1    j  k−1−j k−1

X k − 2   1 j  k−1−j
X k−1 1 2 2
j· · · 1− = (k − 1) · · 1−
j=0
j 2π 2π j=1
j−1 2π 2π
k−2
X   j+1  k−2−j
k−2 1 2
= (k − 1) · · 1−
j=0
j 2π 2π
 k−2
k−1 1 2
= · +1− .
2π 2π 2π

Therefore the answer is

n   k−2
X 1 k−1 1
1+ 1− + 1−
2π 2π 2π
k=2

which is an arithmetic sequence times a geometric sequence. Standard techniques simplify this to 4π.

Solution 4: We solve for the lifespan of an arc. Let f (x) represent the expected number of turns
1
an arc of length 2πx will remain visible. Our final answer will be to calculate f ( 2π ) Then we get the
recurrence Z x
2π − 1
f (x) = 1 + ( − x)f (x) + 2 f (x) dx
2π 0

The 1 term comes from counting the fact that the arc is visible during the current turn. The ( 2π−12π −
x)f (x) term comes from the fact that there is a 2π−1
2π − x chance that the next arc will not intersect the
current arc, in which case the current arc would get an extra f (x) turns to live. The integral comes
from the fact that we want to take the average of f (y) for y ∼r [0, x], which corresponds to the next
arc covering up 2π(x − y) of the current one.
If we differentiate both sides, we end up with a differential equation. Solve it via separation. We end
1
up with f (x) = 4π 2 x + 2π. Plugging in x = 2π gets us 4π.
x
In general, the answer is f (x) = k2 + k1 , where we are placing arcs of length 2πk.