HMMT February 2024

February 17, 2024

Algebra and Number Theory Round
1. Suppose r, s, and t are nonzero reals such that the polynomial x2 + rx + s has s and t as roots, and
the polynomial x2 + tx + r has 5 as a root. Compute s.
2. Suppose a and b are positive integers. Isabella and Vidur both fill up an a × b table. Isabella fills it
up with numbers 1, 2, . . . , ab, putting the numbers 1, 2, . . . , b in the first row, b + 1, b + 2, . . . , 2b in the
second row, and so on. Vidur fills it up like a multiplication table, putting ij in the cell in row i and
column j. (Examples are shown for a 3 × 4 table below.)

1 2 3 4 1 2 3 4
5 6 7 8 2 4 6 8
9 10 11 12 3 6 9 12
Isabella’s Grid Vidur’s Grid

Isabella sums up the numbers in her grid, and Vidur sums up the numbers in his grid; the difference
between these two quantities is 1200. Compute a + b.
3. Compute the sum of all two-digit positive integers x such that for all three-digit (base 10) positive
integers a b c, if a b c is a multiple of x, then the three-digit (base 10) number b c a is also a multiple of
x.
4. Let f (x) be a quotient of two quadratic polynomials. Given that f (n) = n3 for all n ∈ {1, 2, 3, 4, 5},
compute f (0).
5. Compute the unique ordered pair (x, y) of real numbers satisfying the system of equations
x 1 y 1
√ − =7 and √ + = 4.
2
x +y 2 x 2
x +y 2 y

6. Compute the sum of all positive integers n such that 50 ≤ n ≤ 100 and 2n + 3 does not divide 2n! − 1.
7. Let P (n) = (n − 13 )(n − 23 ) . . . (n − 403 ) for positive integers n. Suppose that d is the largest positive
integer that divides P (n) for every integer n > 2023. If d is a product of m (not necessarily distinct)
prime numbers, compute m.
8. Let ζ = cos 2π 2π
13 + i sin 13 . Suppose a > b > c > d are positive integers satisfying
√
|ζ a + ζ b + ζ c + ζ d | = 3.

Compute the smallest possible value of 1000a + 100b + 10c + d.

9. Suppose a, b, and c are complex numbers satisfying

a2 = b − c,
b2 = c − a, and
c2 = a − b.

Compute all possible values of a + b + c.

10. A polynomial f ∈ Z[x] is called splitty if and only if for every prime p, there exist polynomials
gp , hp ∈ Z[x] with deg gp , deg hp < deg f and all coefficients of f − gp hp are divisible by p. Compute
the sum of all positive integers n ≤ 100 such that the polynomial x4 + 16x2 + n is splitty.
 HMMT February 2024
February 17, 2024
Algebra and Number Theory Round
1. Suppose r, s, and t are nonzero reals such that the polynomial x2 + rx + s has s and t as roots, and
the polynomial x2 + tx + r has 5 as a root. Compute s.
Proposed by: Rishabh Das
Answer: 29
Solution: The first equation implies st = s, so t = 1. Then x2 + x + r has 5 as a root, so r + 30 = 0,
implying r = −30. Finally, x2 − 30x + s has 1 as a root, so s = 29 .
Remark: We missed the case of s = t, so x2 + rx + s has s = t as one root, and 1 as the other root
(by Vieta’s). This means r = −s − 1. Then

x2 + tx + r = x2 + sx − (s + 1) = (x + (s + 1))(x − 1)

has 5 as a root, so s = −6 is another solution. During the competition, both the answers −6 and 29
(as well as “29 or −6”) were accepted.

2. Suppose a and b are positive integers. Isabella and Vidur both fill up an a × b table. Isabella fills it
up with numbers 1, 2, . . . , ab, putting the numbers 1, 2, . . . , b in the first row, b + 1, b + 2, . . . , 2b in the
second row, and so on. Vidur fills it up like a multiplication table, putting ij in the cell in row i and
column j. (Examples are shown for a 3 × 4 table below.)

1 2 3 4 1 2 3 4
5 6 7 8 2 4 6 8
9 10 11 12 3 6 9 12
Isabella’s Grid Vidur’s Grid

Isabella sums up the numbers in her grid, and Vidur sums up the numbers in his grid; the difference
between these two quantities is 1200. Compute a + b.
Proposed by: Rishabh Das
Answer: 21
n(n+1)
Solution: Using the formula 1 + 2 + · · · + n = 2 , we get

ab(ab + 1) a(a + 1) b(b + 1) ab (2(ab + 1) − (a + 1)(b + 1))

− · =
2 2 2 4
ab(ab − a − b + 1)
=
4
ab(a − 1)(b − 1)
=
4
a(a − 1) b(b − 1)
= · .
2 2

This means we can write the desired equation as

a(a − 1) · b(b − 1) = 4800.

Assume b ≤ a, so we know b(b − 1) ≤ a(a − 1), so b(b − 1) < 70. Thus, b ≤ 8.

If b = 7 or b = 8, then b(b − 1) has a factor of 7, which 4800 does not, so b ≤ 6.
 If b = 6 then b(b − 1) = 30, so a(a − 1) = 160, which can be seen to have no solutions.
If b = 5 then b(b − 1) = 20, so a(a − 1) = 240, which has the solution a = 16, giving 5 + 16 = 21 .
We need not continue since we are guaranteed only one solution, but we check the remaining cases
for completeness. If b = 4 then a(a − 1) = 4800 12 = 400, which has no solutions. If b = 3 then
a(a − 1) = 4800
6 = 800 which has no solutions. Finally, if b = 2 then a(a − 1) = 4800
2 = 2400, which has
no solutions.
The factorization of the left side may come as a surprise; here’s a way to see it should factor without
doing the algebra. If either a = 1 or b = 1, then the left side simplifies to 0. As a result, both a − 1
and b − 1 should be a factor of the left side.

3. Compute the sum of all two-digit positive integers x such that for all three-digit (base 10) positive
integers a b c, if a b c is a multiple of x, then the three-digit (base 10) number b c a is also a multiple of
x.
Proposed by: Karthik Venkata Vedula
Answer: 64
Solution: Note that abc0 − bca = a(104 − 1) must also be a multiple of x. Choosing a = 1 means that
x divides 103 − 1, and this is clearly a necessary and sufficient condition. The only two-digit factors of
103 − 1 are 27 and 37, so our answer is 27 + 37 = 64 .

4. Let f (x) be a quotient of two quadratic polynomials. Given that f (n) = n3 for all n ∈ {1, 2, 3, 4, 5},
compute f (0).
Proposed by: Pitchayut Saengrungkongka
24
Answer: 17

Solution: Let f (x) = p(x)/q(x). Then, x3 q(x) − p(x) has 1, 2, 3, 4, 5 as roots. Therefore, WLOG, let

x3 q(x) − p(x) = (x − 1)(x − 2)(x − 3)(x − 4)(x − 5) = x5 − 15x4 + 85x3 − . . .

Thus, q(x) = x2 −15x+85, so q(0) = 85. Plugging x = 0 in the above equation also gives −p(0) = −120.
120 24
Hence, the answer is 85 = 17 .
Remark. From the solution above, it is not hard to see that the unique f that satisfies the problem is

225x2 − 274x + 120

f (x) = .
x2 − 15x + 85

5. Compute the unique ordered pair (x, y) of real numbers satisfying the system of equations
x 1 y 1
p − =7 and p + = 4.
x2 + y2 x x2 + y2 y

Proposed by: Pitchayut Saengrungkongka

− 13 13


Answer: 96 , 40

Solution 1: Consider vectors  p   

x/px2 + y 2 −1/x
and .
y/ x2 + y 2 1/y
 √ √
 
7
They are orthogonal and add up to , which have length 72 + 42 = 65. The first vector has
4
√
length 1, so by Pythagorean’s theorem, the second vector has length 65 − 1 = 8, so we have
1 1 p
2
+ 2 = 64 =⇒ x2 + y 2 = ±8xy.
x y
However, the first
p equation indicates that x < 0, while the second equation indicates that y > 0, so
xy < 0. Thus, x2 + y 2 = −8xy. Plugging this into both of the starting equations give
1 1 1 1
− − = 7 and − + = 4.
8y x 8x y

− 13 13


Solving this gives (x, y) = 96 , 40 , which works.

Solution 2: Let x = r cos θ and y = r sin θ. Then our equations read

1
cos θ − =7
r cos θ
1
sin θ + = 4.
r sin θ
Multiplying the first equation by cos θ and the second by sin θ, and then adding the two gives 7 cos θ +
4 sin θ = 1. This means

4 sin θ = 1 − 7 cos θ =⇒ 16 sin2 θ = 1 − 14 cos θ + 49 cos2 θ =⇒ 65 cos2 θ − 14 cos θ − 15 = 0.

3 5 3
This factors as (13 cos θ + 5)(5 cos θ − 3) = 0, so cos θ is either 5 or − 13 . This means either cos θ = 5
and sin θ = − 45 , or cos θ = − 13
5
and sin θ = 12
13 .
The first case, plugging back in, makes r a negative number, a contradiction, so we take the second
case. Then x = cos 1θ−7 = − 13 1 13
− 13 13


96 and y = 4−sin θ = 40 . The answer is (x, y) = 96 , 40 .

6. Compute the sum of all positive integers n such that 50 ≤ n ≤ 100 and 2n + 3 does not divide 2n! − 1.
Proposed by: Pitchayut Saengrungkongka
Answer: 222
Solution: We claim that if n ≥ 10, then 2n + 3 - 2n! − 1 if and only if both n + 1 and 2n + 3 are prime.
If both n + 1 and 2n + 3 are prime, then assume 2n + 3 | 2n! − 1. By Fermat Little Theorem,
2n + 3 | 22n+2 + 1. However, since n + 1 is prime, gcd(2n + 2, n!) = 2, so 2n + 3 | 22 − 1 = 3, a
contradiction.
If 2n + 3 is composite, then ϕ(2n + 3) is even and is at most 2n, so ϕ(2n + 3) | n!, done.
If n + 1 is composite but 2n + 3 is prime, then 2n + 2 | n!, so 2n + 3 | 2n! − 1.
The prime numbers between 50 and 100 are 53, 59, 61, 67, 71, 73, 79, 83, 89, 97. If one of these is n + 1,
then the only numbers that make 2n + 3 prime are 53, 83, and 89, making n one of 52, 82, and 88.
These sum to 222 .

7. Let P (n) = (n − 13 )(n − 23 ) . . . (n − 403 ) for positive integers n. Suppose that d is the largest positive
integer that divides P (n) for every integer n > 2023. If d is a product of m (not necessarily distinct)
prime numbers, compute m.
Proposed by: Nithid Anchaleenukoon
Answer: 48
 Solution: We first investigate what primes divide d. Notice that a prime p divides P (n) for all
n ≥ 2024 if and only if {13 , 23 , . . . , 403 } contains all residues in modulo p. Hence, p ≤ 40. Moreover,
x3 ≡ 1 must not have other solution in modulo p than 1, so p 6≡ 1 (mod 3). Thus, the set of prime
divisors of d is S = {2, 3, 5, 11, 17, 23, 29}.
Next,
j kthe main claim is that for all prime p ∈ S, the minimum value of νp (P (n)) across all n ≥ 2024
is 40
p . To see why, note the following:
j k
40
• Lower Bound. Note that for all n ∈ Z, one can group n − 13 , n − 23 , . . . , n − 403 into p
contiguous blocks of size p. Since p 6≡ 1 (mod 3), x3 span through all residues modulo p, so each
block
j k will have one number divisible by p. Hence,j among k n − 13 , n − 23 , . . . , n − 403 , at least
40
p are divisible by p, implying that νp (P (n)) > 40
p .

• Upper Bound. We pick any n such that νp (n) = 1 so that only terms in form n − p3 , n − (2p)3 ,
2
. . . are divisible by
j p.k Note that these terms are not divisible by p either, so in this case, we
40
have νp (P (n)) = p .
j k
40
Hence, νp (d) = p for all prime p ∈ S. Thus, the answer is

X  40  
40
            
40 40 40 40 40 40
= + + + + + + = 48 .
p 2 3 5 11 17 23 29
p∈S

8. Let ζ = cos 2π 2π
13 + i sin 13 . Suppose a > b > c > d are positive integers satisfying
√
|ζ a + ζ b + ζ c + ζ d | = 3.

Compute the smallest possible value of 1000a + 100b + 10c + d.

Proposed by: Rishabh Das
Answer: 7521
0 0 0
Solution: We may as well take d = 1 and shift the other variables down by d to get |ζ a +ζ b +ζ c +1| =
√
3. Multiplying by its conjugate gives
0 0 0 0 0 0
(ζ a + ζ b + ζ c + 1)(ζ −a + ζ −b + ζ −c + 1) = 3.

Expanding, we get X
1+ ζ x−y = 0,
x,y∈S, x6=y
0 0 0
where S = {a , b , c , 0}.
This is the sum of 13 terms, which hints that S − S should form a complete residue class mod 13. We
can prove this with the fact that the minimal polynomial of ζ is 1 + x + x2 + · · · + x12 .
The minimum possible value of a0 is 6, as otherwise every difference would be between −5 and 5 mod
13. Take a0 = 6. If b0 ≤ 2 then we couldn’t form a difference of 3 in S, so b0 ≥ 3. Moreover, 6−3 = 3−0,
so 3 6∈ S, so b0 = 4 is the best possible. Then c0 = 1 works.
If a0 = 6, b0 = 4, and c0 = 1, then a = 7, b = 5, c = 2, and d = 1, so the answer is 7521 .
9. Suppose a, b, and c are complex numbers satisfying
a2 = b − c,
b2 = c − a, and
c2 = a − b.
Compute all possible values of a + b + c.
Proposed by: Rishabh Das
√
Answer: 0, ±i 6
Solution: Summing the equations gives a2 + b2 + c2 = 0 and summing a times the first equation and
etc. gives a3 + b3 + c3 = 0. Let a + b + c = k. Then a2 + b2 + c2 = 0 means ab + bc + ca = k 2 /2, and
a3 + b3 + c3 = 0 =⇒ −3abc = a3 + b3 + c3 − 3abc = (a + b + c)(a2 + b2 + c2 − ab − bc − ca) = −k 3 /2,
so abc = k 3 /6.
This means a, b, and c are roots of the cubic
x3 − kx2 + (k 2 /2)x − (k 3 /6) = 0
for some k.
Next, note that
X
a4 + b4 + c4 = a(ka2 − (k 2 /2)a + (k 3 /6))
cyc
X
= k(ka2 − (k 2 /2)a + (k 3 /6)) − (k 2 /2)a2 + (k 3 /6)a
cyc
X
= (k 2 /2)a2 − (k 3 /3)a + (k 4 /6)
cyc

= −k 4 /3 + k 4 /2
= k 4 /6.
After this, there are two ways to extract the values of k.
• Summing squares of each equation gives
X
a4 + b4 + c4 = (a − b)2 = 2(a2 + b2 + c2 ) − 2(ab + bc + ca) = −k 2 ,
cyc

so
k4 √
= −k 2 =⇒ k = 0, ±i 6 .
6
• Summing a2 times the first equation, etc. gives
X k6
a4 + b4 + c4 = a2 (b − c) = −(a − b)(b − c)(c − a) = −a2 b2 c2 = − ,
cyc
36
so
k4 k6 √
=− =⇒ k = 0, ±i 6 .
6 36
√ √
We can achieve k = 0 with a = b = c = 0. Letting a, b, and c be the roots of x3 − (i 6)x2 − 3x + (i 6)
will force one of a2 = b − c and all other equalities or a2 − c − b and all other equalities
√ to hold, if the
latter happens, swap b and c. Finally, for these (a, b, c), take (−a, −c, −b) to get −i 6. Thus, all of
these are achievable.
10. A polynomial f ∈ Z[x] is called splitty if and only if for every prime p, there exist polynomials
gp , hp ∈ Z[x] with deg gp , deg hp < deg f and all coefficients of f − gp hp are divisible by p. Compute
the sum of all positive integers n ≤ 100 such that the polynomial x4 + 16x2 + n is splitty.
Proposed by: Pitchayut Saengrungkongka
Answer: 693
Solution: We claim that x4 + ax2 + b is splitty if and only if either b or a2 − 4b is a perfect square.
(The latter means that the polynomial splits into (x2 − r)(x2 − s)).
Assuming the characterization, one can easily extract the answer. For a = 16 and b = n, one of
n and 64 − n has to be a perfect square. The solutions to this that are at most 64 form 8 pairs
that sum to 64 (if we include 0), and then we additionally have 81 and 100. This means the sum is
64 · 8 + 81 + 100 = 693 .
Now, we move on to prove the characterization.
Necessity.
Take a prime p such that neither a2 − 4b nor b is a quadratic residue modulo p (exists by Dirichlet +
CRT + QR). Work in Fp . Now, suppose that

x4 + ax2 + b = (x2 + mx + n)(x2 + sx + t).

Then, looking at the x3 -coefficient gives m + s = 0 or s = −m. Looking at the x-coefficient gives
m(n − t) = 0.
• If m = 0, then s = 0, so x4 +ax2 +b = (x2 +n)(x2 +t), which means a2 −4b = (n+t)2 −4nt = (n−t)2 ,
a quadratic residue modulo p, contradiction.
• If n = t, then b = nt is a square modulo p, a contradiction. (The major surprise of this problem
is that this suffices, which will be shown below.)
Sufficiency.
Clearly, the polynomial splits in p = 2 because in F2 [x], we have x4 + ax2 + b = (x2 + ax + b)2 . Now,
assume p is odd.
If a2 − 4b is a perfect square, then x4 + ax2 + b splits into (x2 − r)(x2 − s) even in Z[x].
If b is a perfect square, then let b = k 2 . We then note that
 2 2

• x4 + ax2 + b splits in form (x2 − r)(x2 − s) if a −4k p = 1.
 
2k−a
• x4 + ax2 + b splits in form (x2 + rx + k)(x2 − rx + k) if a = 2k − r2 , or p = 1.
 
−2k−a
• x4 + ax2 + b splits in form (x2 + rx − k)(x2 − rx − k) if a = −2k − r2 , or p = 1.

Since (2k − a)(−2k − a) = a2 − 4k 2 , it follows that at least one of these must happen.
 HMMT February 2024
February 17, 2024
Combinatorics Round
1. Compute the number of ways to divide a 20 × 24 rectangle into 4 × 5 rectangles. (Rotations and
reflections are considered distinct.)

2. A lame king is a chess piece that can move from a cell to any cell that shares at least one vertex with
it, except for the cells in the same column as the current cell.
A lame king is placed in the top-left cell of a 7 × 7 grid. Compute the maximum number of cells it can
visit without visiting the same cell twice (including its starting cell).
3. Compute the number of ways there are to assemble 2 red unit cubes and 25 white unit cubes into a
3 × 3 × 3 cube such that red is visible on exactly 4 faces of the larger cube. (Rotations and reflections
are considered distinct.)
4. Sally the snail sits on the 3 × 24 lattice of points (i, j) for all 1 ≤ i ≤ 3 and 1 ≤ j ≤ 24. She wants to
visit every point in the lattice exactly once. In a move, Sally can move to a point in the lattice exactly
one unit away. Given that Sally starts at (2, 1), compute the number of possible paths Sally can take.

5. The country of HMMTLand has 8 cities. Its government decides to construct several two-way roads
between pairs of distinct cities. After they finish construction, it turns out that each city can reach
exactly 3 other cities via a single road, and from any pair of distinct cities, either exactly 0 or 2 other
cities can be reached from both cities by a single road. Compute the number of ways HMMTLand
could have constructed the roads.

6. In each cell of a 4 × 4 grid, one of the two diagonals is drawn uniformly at random. Compute the
probability that the resulting 32 triangular regions can be colored red and blue so that any two regions
sharing an edge have different colors.
7. There is a grid of height 2 stretching infinitely in one direction. Between any two edge-adjacent cells
of the grid, there is a door that is locked with probability 12 independent of all other doors. Philip
starts in a corner of the grid (in the starred cell). Compute the expected number of cells that Philip
can reach, assuming he can only travel between cells if the door between them is unlocked.

···
···
···
8. Rishabh has 2024 pairs of socks in a drawer. He draws socks from the drawer uniformly at random,
without replacement, until he has drawn a pair of identical socks. Compute the expected number of
unpaired socks he has drawn when he stops.
9. Compute the number of triples (f, g, h) of permutations on {1, 2, 3, 4, 5} such that

f (g(h(x))) = h(g(f (x))) = g(x),

g(h(f (x))) = f (h(g(x))) = h(x), and
h(f (g(x))) = g(f (h(x))) = f (x)

for all x ∈ {1, 2, 3, 4, 5}.

10. A peacock is a ten-digit positive integer that uses each digit exactly once. Compute the number of
peacocks that are exactly twice another peacock.
 HMMT February 2024
February 17, 2024
Combinatorics Round
1. Compute the number of ways to divide a 20 × 24 rectangle into 4 × 5 rectangles. (Rotations and
reflections are considered distinct.)
Proposed by: Derek Liu
Answer: 6
Solution: For convenience, say the edge of length 20 is vertical.
Consider some vertical line inside the rectangle. It must pass through rectangles of some heights adding
to 20. In particular, these heights correspond to ways to add up to 20 with fours and fives, which is
either 4 + 4 + 4 + 4 + 4 or 5 + 5 + 5 + 5. These options correspond to columns of rectangles with width
5 or 4, respectively. In particular, we need to span the width of the original 20 × 24 rectangle using
these columns, meaning that we can just count the number of ways to add to 24 with fours and fives.

4 4 4 4 4 4 4 5 5 5 5

There are two ways to do this: either 4 + 4 + 4 + 4 + 4 + 4 or 4 + 5 + 5 + 5 + 5. Considering the orders

in which the second sum can be written, we get an answer of 1 + 5 = 6 .

2. A lame king is a chess piece that can move from a cell to any cell that shares at least one vertex with
it, except for the cells in the same column as the current cell.
A lame king is placed in the top-left cell of a 7 × 7 grid. Compute the maximum number of cells it can
visit without visiting the same cell twice (including its starting cell).
Proposed by: Arul Kolla
Answer: 43
Solution: Color the columns all-black and all-white, alternating by column. Each move the lame king
takes will switch the color it’s on. Assuming the king starts on a black cell, there are 28 black and 21
white cells, so it can visit at most 22 + 21 = 43 cells in total, which is easily achievable:
3. Compute the number of ways there are to assemble 2 red unit cubes and 25 white unit cubes into a
3 × 3 × 3 cube such that red is visible on exactly 4 faces of the larger cube. (Rotations and reflections
are considered distinct.)
Proposed by: Albert Wang
Answer: 114
Solution:
We do casework on the two red unit cubes; they can either be in a corner, an edge, or the center of
the face.
• If they are both in a corner, they must be adjacent – for each configuration, this corresponds to
an edge, of which there are 12.

• If one is in the corner and the other is at an edge, we have 8 choices to place the corner. For the
edge, the red edge square has to go on the boundary of the faces touching the red corner square,
and there are six places here. Thus, we get 8 · 6 = 48 configurations.

• If one is a corner and the other is in the center of a face, we again have 8 choices for the corner and
3 choices for the center face (the faces not touching the red corner). This gives 8·3 = 8+8+8 = 24
options.

• We have now completed the cases with a red corner square! Now suppose we have two edges: If
we chose in order, we have 12 choices for the first cube. For the second cube, we must place the
edge so it covers two new faces, and thus we have five choices. Since we could have picked these
edges in either order, we divide by two to avoid overcounting, and we have 12 · 5/2 = 30 in this
case.
 Now, since edges and faces only cover at most 2 and 1 face respectively, no other configuration works.
Thus we have all the cases, and we add: 12 + 48 + 24 + 30 = 114 .

4. Sally the snail sits on the 3 × 24 lattice of points (i, j) for all 1 ≤ i ≤ 3 and 1 ≤ j ≤ 24. She wants to
visit every point in the lattice exactly once. In a move, Sally can move to a point in the lattice exactly
one unit away. Given that Sally starts at (2, 1), compute the number of possible paths Sally can take.
Proposed by: Isabella Zhu
Answer: 4096

Solution 1:
On her first turn, Sally cannot continue moving down the middle row. She must turn either to the
bottom row or the top row. WLOG, she turns to the top row, and enters the cell (3, 1) and we will
multiply by 2 later. Then, we can see that the path must finish in (1, 1). So, we will follow these two
branches of the path, one for the start and one for the end. These branches must both move one unit
up, and then one of the paths must move into the center row. Both branches move up one unit, and
then the path in the middle row must go back to fill the corner. After this, we have exactly the same
scenario as before, albeit with two fewer rows. So, for each additional two rows, we have a factor of
two and thus there are 212 = 4096 paths.
Solution 2:
We solve this problem for a general 3 by 2n grid.
On her first turn, Sally cannot continue moving down the middle row. She must turn either to the
topmost row or the bottommost row. WLOG, she turns to the top row.
Suppose Sally returns to the middle row k times. There are k ”blocks”. However, 2k of the squares are
already occupied by Sally’s row shifts. Thus, we are solving

2(x1 + x2 + . . . xk ) = 2(n − k).

There are
n (
∑ )
n−k+k−1
= 2n−1
k−1
k=1

solutions. We multiply by 2 to get 2n .

For n = 12, this evaluates to 212 = 4096 .
5. The country of HMMTLand has 8 cities. Its government decides to construct several two-way roads
between pairs of distinct cities. After they finish construction, it turns out that each city can reach
exactly 3 other cities via a single road, and from any pair of distinct cities, either exactly 0 or 2 other
cities can be reached from both cities by a single road. Compute the number of ways HMMTLand
could have constructed the roads.
Proposed by: Derek Liu
Answer: 875
Solution: Let the cities be numbered 1, 2, 3, 4, 5, 6, 7, 8. WLOG, 1 is connected to 2, 3, and 4. First
suppose 2 and 3 are connected; then 3 and 1 share a second common neighbor, which must be 4 (as
1 is not connected to anything else). Likewise 2 and 4 are connected, and so 5, 6, 7, 8 are pairwise
connected as well, so the graph consists of two disjoint copies of K4 :

1 2 5 6

4 3 8 7

()
1 8
There are 2 4 = 35 ways to partition the 8 vertices into two groups of 4, so there are 35 such graphs.
Otherwise, none of 2, 3, 4 are connected to each other. Then 2 and 3 must share a common neighbor, as
must 3 and 4, and 2 and 4. If these are the same neighbor, this vertex would share all three neighbors
with 1, so they must be pairwise distinct. The last vertex must then be connected to these three,
creating a cube graph.

4 5

1 3

8 6

2 7

8!
A cube has 48 symmetries, so the number of such graphs is 48 = 840.
The total is 35 + 840 = 875 .

6. In each cell of a 4 × 4 grid, one of the two diagonals is drawn uniformly at random. Compute the
probability that the resulting 32 triangular regions can be colored red and blue so that any two regions
sharing an edge have different colors.
Proposed by: Derek Liu
1
Answer: 512

Solution: Give each cell coordinates from (1, 1) to (4, 4).

Claim. The grid has a desired coloring if and only if every vertex not on the boundary meets an even
number of edges and diagonals.
Proof. If this were not the case, the odd number of regions around the vertex would have to alternate
between the two colors, which is clearly impossible. In the event that every vertex has an even number
of incident edges, it is not hard to show that the grid is always colorable.
 We claim the diagonals drawn in the cells of form (1, a) and (a, 1) for 1 ≤ a ≤ 4 uniquely determine
the rest (for a valid coloring to exist). Indeed, given the diagonals for any three cells around a vertex,
we can uniquely determine the fourth one using the parity in the claim above. If (1, 1), (1, 2), (2, 1)
are fixed, so is (2, 2); likewise so are (2, 3) and (2, 4), etc. until the whole grid is fixed.

The solid lines force the dotted lines as described above.

Thus, once the seven cells along the top row and leftmost column are determined, the remaining nine
1
have a 219 = chance of being selected in a way that admits a coloring.
512

7. There is a grid of height 2 stretching infinitely in one direction. Between any two edge-adjacent cells
of the grid, there is a door that is locked with probability 12 independent of all other doors. Philip
starts in a corner of the grid (in the starred cell). Compute the expected number of cells that Philip
can reach, assuming he can only travel between cells if the door between them is unlocked.

···
···
···
Proposed by: Jacob Paltrowitz
32
Answer: 7

Solution: For clarity, we will number our grid, with (0,0) being the corner that Philip starts in, and
the grid stretching in the positive x direction, i.e. all elements of the grid are of the form (x, y), with
y ∈ {0, 1} and x ∈ N.
We will use recursion and casework. Let A be the expected number of reachable cells given that the
door between (0, 0) and (0, 1) is unlocked, and B be the expected number of cells given that door is
closed. Since that door is locked 12 of the time, our answer is A+B
2 .
We can write recurrence relations by considering the different configurations of the doors in the first 4
cells.
For the sake of writing, let W be the (0, 0) − (0, 1) door, X be the (0, 0) − (1, 0) door, Y be the
(0, 1) − (1, 1) door, and Z be the (1, 0) − (1, 1) door.
Let’s start with the case where W is unlocked and compute A:

Y Y Y

W ? W ? W ?

X X X
 Case 1: W is unlocked. Shaded cells represent inaccessible cells, and the arrows show Philip’s
movements between cells.

• If X, Y are both locked, then Philip can reach exactly 2 rooms. This occurs with probability 14 .
• If both of X, Y are unlocked, then we have exactly the same case of A, except with the ability to
reach two extra cells. This occurs with probability 14 .
• If exactly one of X, Y are unlocked, we have back to the original case, except with the ability to
access two more cells, which occurs with probability 12 .

So, we get the equation:

( )
1 1 1 A+B
A = (2) + (A + 2) + +2
4 4 2 2

Now, let’s consider when W is locked.

Y Y Y

W ? W ? W ?

X X X

Case 2: W is locked.

• If X is locked, Philip can only reach one cell. This occurs with probability 12 .
• If X is unlocked and Y is locked, we have exactly the original problem, except with the ability to
reach one more cell. This occurs with probability 14 .
• In the case where X is unlocked and Y is unlocked, this is the same as the original configuration,
except with the ability to reach one extra cell (the start) and possibly the cell at (0, 1).
Now, let’s compute the probability that Philip can reach (0, 1) in this case. This is the probability
that Philip can reach (1, 1) since Y is unlocked. We can compute that the probability that Philip
can reach (1, 1) from (1, 0) is equal to
∑∞
1
23n+1
n=0

by looking at the minimum distance Philip has to go to the right before getting back to (1, 1).
This is a geometric series with sum 74 . So, in this case, on average Philip can reach 1 + 47 more
cells than the original case. This case occurs with probability 14 .

So, we can write the equation:

( ) ( )
1 1 A+B 1 A+B 4
B = (1) + +1 + +1+
2 4 2 4 2 7

40 24 A+B 32
Solving the system of these two linear equations, we get A = 7 ,B = 7 and 2 = .
7
8. Rishabh has 2024 pairs of socks in a drawer. He draws socks from the drawer uniformly at random,
without replacement, until he has drawn a pair of identical socks. Compute the expected number of
unpaired socks he has drawn when he stops.
Proposed by: Rishabh Das
42024
Answer: (4048) − 2
2024

Solution 1: We solve for the expected number of total socks drawn and subtract two at the end.
Let En be the expected number of socks drawn for n pairs of socks, so that E1 = 2. Suppose there
are n pairs of socks, Rishabh continued to draw socks until the drawer was empty, and without loss of
generality let the last sock drawn be red. If we ignore the two red socks, the process is equivalent to
drawing from a drawer with n − 1 pairs of socks. Let k be the number of socks drawn until a pair of
k
identical socks is found, after ignoring the two red socks. Then the first red sock has probability 2n−1
of being before this stopping point, so the expected value is k + 2n−1 = k · 2n−1 . Since the expected
k 2n

value of k is En−1 , we have

2n
En = · En−1 .
2n − 1
Applying this recurrence, we get
(2n)!! 2n · n! 4n · (n!)2 4n
En = = = = (2n) .
(2n − 1)!! (2n − 1)!! (2n)! n

42024
Subtracting two and plugging in n = 2024 gives a final answer of (4048) − 2 .
2024

Solution 2: Let P (k) denote the probability that Rishabh draws more than k socks. We compute
P (k) for all 0 ≤ k ≤ 2024 (and note P (k) = 0 for larger k).
The number of ways to draw k socks, none identical to each other, is
2024!
4048 · 4046 · · · · (4050 − 2k) = 2k · ,
(2024 − k)!
while the total number of ways to draw k socks is
4048!
4048 · 4047 · · · (4049 − k) = .
(4048 − k)!
Thus,
( )
2k · 2024!
(2024−k)! 2024! 2k (4048 − k)! 1 4048 − k
P (k) = = · = (4048) · 2k .
4048!
(4048−k)!
4048! (2024 − k)! 2024
2024
The expected number of socks drawn is
∑
2024 ( )
1 4048 − k
P (0) + P (1) + · · · + P (2024) = (4048) 2 k
.
2024
2024
k=0

This sum is equivalent to Putnam 2020 A2. We claim that it is equal to 42024 . We do this via a
counting argument: we count how many ways there are to choose at least half of the elements from
4049
the set {1, 2, . . . , 4049}. On the one hand that is 2 2 = 42024 . On the other hand, letting k + 1 be
(4048−k)
the 2025th largest element chosen, there are 2024 ways to choose the elements larger than it, and
2k ways to choose the elements smaller than it. Varying k, we get
∑
2024 ( )
k 4048 − k
2 = 42024 .
2024
k=0
 This means the expected number of socks is
42024
(4048) ,
2024

42024
and subtracting two for the matching pair gives (4048) − 2 .
2024

9. Compute the number of triples (f, g, h) of permutations on {1, 2, 3, 4, 5} such that

f (g(h(x))) = h(g(f (x))) = g(x),

g(h(f (x))) = f (h(g(x))) = h(x), and
h(f (g(x))) = g(f (h(x))) = f (x)

for all x ∈ {1, 2, 3, 4, 5}.

Proposed by: Rishabh Das
Answer: 146
Solution: Let f g represent the composition of permutations f and g, where (f g)(x) = f (g(x)) for all
x ∈ {1, 2, 3, 4, 5}.
Evaluating f ghf h in two ways, we get

f = gf h = (f gh)f h = f ghf h = f (ghf )h = f hh,

so hh = 1. Similarly, we get f , g, and h are all involutions. Then

f gh = g =⇒ f g = gh,

so f g = gh = hf . Let x := f g = gh = hf . Then

x3 = (f g)(gh)(hf ) = 1.

We can also show that f g = gh = hf along with f, g, h being involutions is enough to recover the
initial conditions, so we focus on satisfying these new conditions.
() ( 5 )
If x = 1, then f = g = h is an involution. There are 1 + 52 + 12 2,2,1 = 26 involutions, so this case
gives 26 solutions.
Suppose x ̸= 1. Then since x3 = 1, x is composed of a 3-cycle and two fixed points, of which there are
20 choices. WLOG x = (1 2 3). It can be checked that {1, 2, 3} must map to itself for all of f, g, h and
also {4, 5}. We can either have all of f, g, h map 4 and 5 to themselves or each other. Restricted to
{1, 2, 3}, they are some rotation of (1 2), (2 3), (1 3). Each of the 20 cases thus gives 2 · 3 = 6 triples, so
overall we get 20 · 6 = 120.
The final answer is 26 + 120 = 146 .

10. A peacock is a ten-digit positive integer that uses each digit exactly once. Compute the number of
peacocks that are exactly twice another peacock.
Proposed by: Albert Wang
Answer: 184320
Solution:
We begin with the following observation:

Claim 1. Let x be a peacock. Then, 2x is a peacock if and only if:

 • the multiplication x · 2 uses five carries,
• each of the pairs of digits (0, 5), (1, 6), (2, 7), (3, 8), (4, 9) receives exactly one carry.
• The leading digit is not 5, 6, 7, 8, 9.

Proof. After the multiplication of x · 2, we will have a ten digit number. Let’s first consider the output
without carrying. It consists of the digits 0, 2, 4, 6, 8 twice each, occupying positions where pairs of
digits (0, 5), (1, 6), (2, 7), (3, 8), (4, 9) were in x. However, we guaranteed that one digit from each pair
received a carry, meaning all ten digits are present after adding in the carries.

We will now biject all peacocks to the following combination of objects:

• a queue of low digits 0, 1, 2, 3, 4, in any order with the constraint that 0 is not first,
• a queue of high digits 5, 6, 7, 8, 9, in any order, and
• of each of the pairs of digits (0, 5), (1, 6), (2, 7), (3, 8), (4, 9) mark one of them to receive a carry,
except we are not allowed to mark the final digit in the high queue.
We construct a correspondence from these objects to peacocks by accumulating digits to an initially
empty string. We’ll say that we poll a queue by popping its front entry and appending it to the end of
this string. First, poll the low queue. Then, if we have just polled a marked digit, poll the high queue;
otherwise, poll the low queue. We repeat this until all queues are emptied.
As an example of this process, let our low queue be 1, 4, 0, 2, 3, our high queue be 8, 5, 9, 6, 7, and mark
the digits 0, 1, 2, 3, 9 marked to receive a carry. Our steps are as follows:
• Poll the low queue, so that our string is now 1.
• Since 1 was marked to receive a carry, we poll the high queue, making our string 18.
• Since 8 was not marked, we poll the low queue to reach 184.
• Since 4 was not marked, we poll the low queue to reach 1840.
• Since 0 was marked, we poll the high queue to reach 18405.
• etc.
In the end, we will construct the peacock 1840529637, which is the one shown earlier to work.

Claim 2. Any string of digits x constructed through this process will be a peacock that satisfies
the constraints outlined in Claim 1.

Low queue 1 4 0 2 3

High queue 8 5 9 6 7

The order in which digits get polled to construct 1840529637; note the 4 connected components in the
high queue. The circled digits are those that have been marked for carrying.

Proof. We first argue that all digits end up being polled. In particular, if a high digit is marked, let’s
connect it by an edge to the digit on its right (using the requirement that the last digit is not marked).
If h of the high digits are marked, then we will have 5 − h connected components among these high
digits. However, we then have 5 − h marked digits in the low queue, and every time we poll a marked
low digit we will end up polling all digits from the next connected component in the high queue.
So, all digits end up being polled. Notice that our marked digits will always be followed immediately
by a high digit, satisfying the first and second conditions of the claim. As we do not start with a high
digit, the third constraint is satisfied. Therefore any peacock x output by this process will also have
2x a peacock.

Since we always use all the digits, this process is evidently injective. To map from peacocks back
to these sequences of digits, we can just let the queues be the order of appearances of the low and high
digits in the peacock, and mark the carried digits accordingly. Indeed, we notice that this mapping is
also injective.
Using this bijection, we just need to find the number of initial settings of the queues and marked digits.
There are 4 · 4! ways to order the low number queue. There are then 5! ways to order the high number
queue. Finally, of each of the four pairs of digits not inluding the final high digit, there are 24 ways to
mark them. This gives an answer of

4 · 4! · 5! · 24 = 184320 .
 HMMT February 2024
February 17, 2024
Geometry Round
1. Inside an equilateral triangle of side length 6, three congruent equilateral triangles of side length x
with sides parallel to the original equilateral triangle are arranged so that each has a vertex on a side
of the larger triangle, and a vertex on another one of the three equilateral triangles, as shown below.

6
A smaller equilateral triangle formed between the three congruent equilateral triangles has side length
1. Compute x.
2. Let ABC be a triangle with ∠BAC = 90◦ . Let D, E, and F be the feet of altitude, angle bisector,
and median from A to BC, respectively. If DE = 3 and EF = 5, compute the length of BC.
3. Let Ω and ω be circles with radii 123 and 61, respectively, such that the center of Ω lies on ω. A chord
of Ω is cut by ω into three segments, whose lengths are in the ratio 1 : 2 : 3 in that order. Given that
this chord is not a diameter of Ω, compute the length of this chord.
4. Let ABCD be a square, and let ℓ be a line passing through the midpoint of segment AB that intersects
segment BC. Given that the distances from A and C to ℓ are 4 and 7, respectively, compute the area
of ABCD.
5. Let ABCD be a convex trapezoid such that ∠DAB = ∠ABC = 90◦ , DA = 2, AB = 3, and BC = 8.
Let ω be a circle passing through A and tangent to segment CD at point T . Suppose that the center
of ω lies on line BC. Compute CT .
6. In triangle ABC, a circle ω with center O passes through B and C and intersects segments AB and AC
again at B ′ and C ′ , respectively. Suppose that the circles with diameters BB ′ and CC ′ are externally
tangent to each other at T . If AB = 18, AC = 36, and AT = 12, compute AO.
7. Let ABC be an acute triangle. Let D, E, and F be the feet of altitudes from A, B, and C to sides BC,
CA, and AB, respectively, and let Q be the foot of altitude from A to line EF . Given that AQ = 20,
BC = 15, and AD = 24, compute the perimeter of triangle DEF .
8. Let ABT CD be a convex pentagon with area 22 such that AB = CD and the circumcircles of triangles
T AB and T CD are internally tangent. Given that ∠AT D = 90◦ , ∠BT C = 120◦ , BT = 4, and CT = 5,
compute the area of triangle T AD.
9. Let ABC be a triangle. Let X be the point on side AB such that ∠BXC = 60◦ . Let P be the point
on segment CX such that BP ⊥ AC. Given that AB = 6, AC = 7, and BP = 4, compute CP .
10. Suppose point P is inside quadrilateral ABCD such that
∠P AB = ∠P DA,
∠P AD = ∠P DC,
∠P BA = ∠P CB, and
∠P BC = ∠P CD.
If P A = 4, P B = 5, and P C = 10, compute the perimeter of ABCD.
 HMMT February 2024
February 17, 2024
Geometry Round
1. Inside an equilateral triangle of side length 6, three congruent equilateral triangles of side length x
with sides parallel to the original equilateral triangle are arranged so that each has a vertex on a side
of the larger triangle, and a vertex on another one of the three equilateral triangles, as shown below.

A smaller equilateral triangle formed between the three congruent equilateral triangles has side length
1. Compute x.
Proposed by: Rishabh Das
5
Answer: 3

Solution:

O√
3
6

√ √
3 3
2 x

Let x be the side length of the shaded triangles. Note that the centers of the triangles with side lengths
1 and 6 coincide; call this common center O.
√
The distance from O to a side of the equilateral triangle with side length √ 1 is 3/6. Similarly the
distance from O to a side of the equilateral triangle with side length 6 is 3. Notice the difference of
these two distances is exactly the length of the altitude of one of shaded triangles. So
√ √
√ 3 3 5
3− = x =⇒ x = .
6 2 3
2. Let ABC be a triangle with ∠BAC = 90◦ . Let D, E, and F be the feet of altitude, angle bisector,
and median from A to BC, respectively. If DE = 3 and EF = 5, compute the length of BC.
Proposed by: Jerry Liang
Answer: 20
Solution 1:
A

3 5
B C
D E F

Since F is the circumcenter of △ABC, we have that AE bisects ∠DAF . So by the angle bisector
theorem, we can set AD = 3x and AF = 5x. Applying Pythagorean theorem to △ADE then gives
(3x)2 + (5 + 3)2 = (5x)2 =⇒ x = 2.

So AF = 5x = 10 and BC = 2AF = 20 .

Solution 2: Let BF = F C = x. We know that △BAD ∼ △ACD so BA BD DA

AC = DA = DC and thus
√ √
BA BD x−8 AB BE x−5
AC = DC = x+8 . By Angle Bisector Theorem, we also have AC = EC = x+5 , which means that
√
x−8 x−5
= =⇒ (x − 8)(x + 5)2 = (x + 8)(x − 5)2
x+8 x+5
which expands to
x3 + 2x2 − 55x − 200 = x3 − 2x2 − 55x + 200 =⇒ 4x2 = 400.
This solves to x = 10, and so BC = 2x = 20 .

3. Let Ω and ω be circles with radii 123 and 61, respectively, such that the center of Ω lies on ω. A chord
of Ω is cut by ω into three segments, whose lengths are in the ratio 1 : 2 : 3 in that order. Given that
this chord is not a diameter of Ω, compute the length of this chord.
Proposed by: Benjamin Kang, Holden Mui, Pitchayut Saengrungkongka
Answer: 42
Solution: Denote the center of Ω as O. Let the chord intersect the circles at W, X, Y, Z so that
W X = t, XY = 2t, and Y Z = 3t. Notice that Y is the midpoint of W Z; hence OY ⊥ W XY Z.
The fact that ∠OY X = 90◦ means X is the antipode of O on ω, so OX = 122. Now applying power
of point to X with respect to Ω gives
245 = 1232 − OX 2 = W X · XZ = 5t2 =⇒ t = 7.
Hence the answer is 6t = 42 .
 Z

O
X
W

4. Let ABCD be a square, and let ℓ be a line passing through the midpoint of segment AB that intersects
segment BC. Given that the distances from A and C to ℓ are 4 and 7, respectively, compute the area
of ABCD.
Proposed by: Ethan Liu
Answer: 185
Solution:

ℓ′
A B

D C

Consider the line ℓ′ through B parallel to ℓ, and drop perpendiculars from A to ℓ′ and C to ℓ′ . Note
that because ℓ passes through the midpoint of segment AB, the distance from B to ℓ is 4. Thus, the
distances from A to ℓ′ and from C to ℓ′ are 4 + 4 = 8 and 4 + 7 = 11, respectively. Let P be the foot
from A to ℓ′ . Rotating the square 90◦ from B to A sends the altitude from C to ℓ′ to the segment
along
√ ℓ′ between √B and the foot from A to ℓ′ ; hence BP = 11. So the side length of the square is
AP + BP 2 = 82 + 112 , which means the area of the square is 82 + 112 = 185 .
2
5. Let ABCD be a convex trapezoid such that ∠DAB = ∠ABC = 90◦ , DA = 2, AB = 3, and BC = 8.
Let ω be a circle passing through A and tangent to segment CD at point T . Suppose that the center
of ω lies on line BC. Compute CT .
Proposed by: Pitchayut Saengrungkongka
√ √
Answer: 4 5− 7
Solution:

T
D

P A B A′

Let A′ be the reflection of A across BC, and let P = AB ∩ CD. Then since the center of ω lies on BC,
we have that ω passes through A′ . Thus, by power of a point, P T 2 = P A · P A′ . By similar triangles,
we have
PA PB PA PA + 3
= =⇒ = =⇒ P A = 1,
AD BC 2 8
√ √ √
and A′ P = 1 + 2 · 3 = 7, so P T = 7. But by the Pythagorean Theorem, P C = P B 2 + BC 2 = 4 5,
√ √
and since T lies on segment CD, it lies between C and P , so CT = 4 5 − 7 .

6. In triangle ABC, a circle ω with center O passes through B and C and intersects segments AB and AC
again at B ′ and C ′ , respectively. Suppose that the circles with diameters BB ′ and CC ′ are externally
tangent to each other at T . If AB = 18, AC = 36, and AT = 12, compute AO.
Proposed by: Ethan Liu
65
Answer: 3

Solution 1:
 A

C′

B′

MB MC
T

O
B

By Radical Axis Theorem, we know that AT is tangent to both circles. Moreove, consider power
of a point A with respect to these three circles, we have AB · AB ′ = AT 2 = AC · AC ′ . Thus
2
122
AB ′ = 12 ′
18 = 8, and AC = 36 = 4. Consider the midpoints MB , MC of segments BB , CC ,
′ ′
◦
respectively. We have ∠OMB A = ∠OMC A = 90 , so O is the antipode of A in (AMB MC ). Notice
that △AMB T ∼ △AOMC , so AM AO
C
= AM
AT . Now, we can do the computations as follow:
B

AMB · AMC
AO =
( AT )( )
AB + AB ′ AC + AC ′ 1
=
2 2 AT
( )( )
8 + 18 36 + 4 1 65
= = .
2 2 12 3

7. Let ABC be an acute triangle. Let D, E, and F be the feet of altitudes from A, B, and C to sides BC,
CA, and AB, respectively, and let Q be the foot of altitude from A to line EF . Given that AQ = 20,
BC = 15, and AD = 24, compute the perimeter of triangle DEF .
Proposed by: Isabella Zhu
√
Answer: 8 11
Solution:
 A

T
E
Q

F
H

B D C

Note that A is the excenter of △DEF and AQ is the length of the exradius. Let T be the tangency
point of the A-excircle to line DF . We have AD = AT = 20. It is well known that the length of DT
is the semiperimeter of DEF . Note that △ADT is a right triangle, so
AT 2 + DT 2 = AD2
which implies √ √
DT = 242 − 202 = 4 11.
√ √
Thus, the perimeter of △DEF is 2 · 4 11 = 8 11 .

8. Let ABT CD be a convex pentagon with area 22 such that AB = CD and the circumcircles of triangles
T AB and T CD are internally tangent. Given that ∠AT D = 90◦ , ∠BT C = 120◦ , BT = 4, and CT = 5,
compute the area of triangle T AD.
Proposed by: Pitchayut Saengrungkongka
√
Answer: 64(2 − 3)
Solution: Paste △T CD outside the pentagon to get △ABX ∼
= △DCT . From the tangent circles
condition, we get
∠XBT = 360◦ − ∠XBA − ∠ABT
= 360◦ − ∠DCT − ∠ABT
= 360◦ − 270◦ = 90◦

∠XAT = 90◦ − ∠BXA − ∠AT B

= 90◦ − ∠CT D − ∠AT B
= 90◦ − (120◦ − 90◦ ) = 60◦ .
Moreover, if x = AT and y = T D, then notice that
[ABT CD] = [ABT ] + [CDT ] + [AT D]
= [XAT ] − [XBT ] + [AT D]
1 1 1
= xy sin 60◦ − · 4 · 5 + xy
2 √ 2 2
2+ 3
= xy − 10,
4
 so we have
4 √ 1 √
xy = 32 · √ = 128(2 − 3) =⇒ [AT D] = xy = 64(2 − 3) .
2+ 3 2

B D

X
T
5
5
4 C
y
y
x
B D

60◦

9. Let ABC be a triangle. Let X be the point on side AB such that ∠BXC = 60◦ . Let P be the point
on segment CX such that BP ⊥ AC. Given that AB = 6, AC = 7, and BP = 4, compute CP .
Proposed by: Pitchayut Saengrungkongka
√
Answer: 38 − 3
Solution: Construct
√ parallelogram
√ BP CQ. We have CQ = 4, ∠ACQ = 90◦ , and ∠ABQ = 120◦ .
Thus, AQ = AC + CQ = 65, so if x = CP = BQ, then by Law of Cosine, x2 + 6x + 62 = 65.
2 2
√
Solving this gives the answer x = 38 − 3 .
 A

X
120◦
60◦
6 7

4 x

B 120◦ C

x 4

10. Suppose point P is inside quadrilateral ABCD such that

∠P AB = ∠P DA,
∠P AD = ∠P DC,
∠P BA = ∠P CB, and
∠P BC = ∠P CD.

If P A = 4, P B = 5, and P C = 10, compute the perimeter of ABCD.

Proposed by: Rishabh Das
√
9 410
Answer: 5

Solution:
 A D

4 5
8

N P′
P M

4
5 10

B C

First of all, note that the angle conditions imply that ∠BAD + ∠ABC = 180◦ , so the quadrilateral
is a trapezoid with AD ∥ BC. Moreover, they imply AB and CD are both tangent to (P AD) and
(P BC); in particular AB = CD or ABCD is isosceles trapezoid. Since the midpoints of AD and BC
clearly lie on the radical axis of the two circles, P is on the midline of the trapezoid.
Reflect △P AB over the midline and translate it so that D = B ′ and C = A′ . Note that P ′ is still
on the midline. The angle conditions now imply P DP ′ C is cyclic, and P P ′ bisects CD. This means
10 · 4 = P C · CP ′ = P D · DP ′ = 5 · P D, so P D = 8.
Now P DP ′ C is a cyclic quadrilateral with side lengths 10, 8, 5, 4 in that order. Using standard cyclic
quadrilateral facts (either law of cosines or three applications
√
on Ptolemy √on the three possible quadri-
laterals formed with these side lengths) we get CD = 5 and P P ′ = 410
2 410
2 . Finally, note that P P
′

is equal to the midline of the trapezoid, so the final answer is

√
′9 410
2 · CD + 2 · P P = .
5
 HMMT February 2024
February 17, 2024
Team Round
1. [20] Let a1 , a2 , a3 , . . . , a100 be integers such that

a21 + a22 + a23 + · · · + a2100

= 100.
a1 + a2 + a3 + · · · + a100
Determine, with proof, the maximum possible value of a1 .
2. [25] Nine distinct positive integers summing to 74 are put into a 3×3 grid. Simultaneously, the number
in each cell is replaced with the sum of the numbers in its adjacent cells. (Two cells are adjacent if
they share an edge.) After this, exactly four of the numbers in the grid are 23. Determine, with proof,
all possible numbers that could have been originally in the center of the grid.
3. [25] Let ABC be a scalene triangle and M be the midpoint of BC. Let X be the point such that
CX ∥ AB and ∠AM X = 90◦ . Prove that AM bisects ∠BAX.
4. [30] Each lattice point with nonnegative coordinates is labeled with a nonnegative integer in such a
way that the point (0, 0) is labeled by 0, and for every x, y ≥ 0, the set of numbers labeled on the
points (x, y), (x, y + 1), and (x + 1, y) is {n, n + 1, n + 2} for some nonnegative integer n. Determine,
with proof, all possible labels for the point (2000, 2024).
5. [40] Determine, with proof, whether there exist positive integers x and y such that x + y, x2 + y 2 , and
x3 + y 3 are all perfect squares.

6. [45] Let Q be the set of rational numbers. Given a rational number a ̸= 0, find, with proof, all functions
f : Q → Q satisfying the equation

f (f (x) + ay) = af (y) + x

for all x, y ∈ Q.

7. [50] Let ABCDEF be a regular hexagon with P as a point in its interior. Prove that of the three
values tan ∠AP D, tan ∠BP E, and tan ∠CP F , two of them sum to the third one.
8. [50] Let P be a point in the interior of quadrilateral ABCD such that the circumcircles of triangles
P DA, P AB, and P BC are pairwise distinct but congruent. Let the lines AD and BC meet at X. If
O is the circumcenter of triangle XCD, prove that OP ⊥ AB.

9. [55] On each cell of a 200 × 200 grid, we place a car, which faces in one of the four cardinal directions.
In a move, one chooses a car that does not have a car immediately in front of it, and slides it one cell
forward. If a move would cause a car to exit the grid, the car is removed instead. The cars are placed
so that there exists a sequence of moves that eventually removes all the cars from the grid. Across all
such starting configurations, determine the maximum possible number of moves to do so.

10. [60] Across all polynomials P such that P (n) is an integer for all integers n, determine, with proof, all
possible values of P (i), where i2 = −1.
 HMMT February 2024
February 17, 2024
Team Round
1. [20] Let a1 , a2 , a3 , . . . , a100 be integers such that

a21 + a22 + a23 + · · · + a2100

= 100.
a1 + a2 + a3 + · · · + a100
Determine, with proof, the maximum possible value of a1 .
Proposed by: Nilay Mishra
Answer: 550
Solution 1: We can rearrange the equation as follows:

a21 + a22 + · · · + a299 + a2100 = 100(a1 + a2 + · · · + a99 + a100 )

(a21 − 100a1 ) + (a22 − 100a2 ) + (a23 − 100a3 ) + · · · + (a2100 − 100a100 ) = 0

(a1 − 50)2 + (a2 − 50)2 + (a3 − 50)2 + · · · + (a100 − 50)2 = 100 · 502 = 5002 .
Thus (a1 − 50)2 ≤ 5002 and so a1 ≤ 550. Equality holds when a1 = 550 and ai = 50 for all i > 1.
Therefore, the maximum possible value of a1 is 550 .

Solution 2: Let k = 1
99 (a2 + a3 + · · · + a100 ). Note, that by the Cauchy-Schwarz inequality, we have:

99 · (a22 + a23 + · · · + a2100 ) ≥ (a2 + a3 + · · · + a100 )2 = (99k)2

and so:
a21 + a22 + a23 + · · · + a2100 a2 + 99k 2
100 = ≥ 1
a1 + a2 + a3 + · · · + a100 a1 + 99k
Now consider the function:
99x2 + t2
ft (x) =
99x + t
for all reals t. Note that for t = a1 , we have that ft (k) ≤ 100 for some real k. Furthermore, because
a21 + a22 + · · · + a2100 > 0, we have a1 + a2 + · · · + a100 > 0, and so ft (x) > 0 for all valid x.

We assume a1 > 0, since we are trying to maximize this element and some positive solution exists
(i.e. (100, 100, . . . , 100)). Hence, the minimum point of ft (x) over x > 0 needs to be less than 100. To
compute this, set the derivative to be zero as follows:

100t2
ft′ (x) = 1 − =0
(t + 99x)2
t
which has solutions at x = 11 and x = − 9t . Taking a second derivative shows that the point ( 11
t 2t
, 11 )
is the only valid minimum point in the first quadrant. Hence, we must have 11 ≤ 100 and so
2t

a1 = t ≤ 550 , as desired. Equality holds at (550, 50, . . . , 50).

2. [25] Nine distinct positive integers summing to 74 are put into a 3×3 grid. Simultaneously, the number
in each cell is replaced with the sum of the numbers in its adjacent cells. (Two cells are adjacent if
they share an edge.) After this, exactly four of the numbers in the grid are 23. Determine, with proof,
all possible numbers that could have been originally in the center of the grid.
Proposed by: Rishabh Das
Answer: 18
 Solution: Suppose the initial grid is of the format shown below:
 
a b c
d e f 
g h i

After the transformation, we end with

   
a n bn cn b+d a+c+e b+f
 dn e n fn  = a + e + g b + d + f + h c + e + i
g n hn in d+h g+e+i f +h

Since d ̸= f , an = b + d ̸= b + f = cn . By symmetry, no two corners on the same side of the grid may
both be 23 after the transformation.
Since c ̸= g, bn = a + c + e ̸= a + e + g = dn . By symmetry, no two central-edge squares sharing a
corner may both be 23 after the transformation.
Assume for the sake of contradiction that en = 23. Because an , cn , gn , in < en , none of an , cn , gn , in
can be equal to 23. Thus, 3 of bn , dn , fn , hn must be 23. WLOG assume bn = dn = fn = 23. Thus is
a contradiction however, as bn ̸= dn . Thus, en ̸= 23.
This leaves the case with two corners diametrically opposite and two central edge squares diametrically
opposite being 23. WLOG assume an = bn = hn = in = 23.
Thus, 92 = 4 · 23 = an + bn + hn + in = (b + d) + (a + c + e) + (e + g + i) + (f + h) = (a + b + c + d +
e + f + g + h + i) + e. Since a + b + c + d + e + f + g + h + i = 74, this means that e = 92 − 74 = 18.
 
4 16 2
One possible example of 18 working is 6 18 7. Thus the only possible value for the center is 18.
1 17 3

3. [25] Let ABC be a scalene triangle and M be the midpoint of BC. Let X be the point such that
CX ∥ AB and ∠AM X = 90◦ . Prove that AM bisects ∠BAX.
Proposed by: Pitchayut Saengrungkongka
Solution:

X
B
M C
Y
 Let Y be the intersection of lines AB and XM . Since BY ∥ CX, we have ∠Y BM = ∠XCM .
Furthermore, we have BM = CM , since M is the midpoint of BC. Thus,

△BM Y ∼
= △CM X.
Thus, M Y = M X. Combined with the condition AM ⊥ XY , we get that AY X is an isosceles triangle
with median AM . Therefore, AM bisects ∠Y AX which is the same as ∠BAX and we are done.

4. [30] Each lattice point with nonnegative coordinates is labeled with a nonnegative integer in such a
way that the point (0, 0) is labeled by 0, and for every x, y ≥ 0, the set of numbers labeled on the
points (x, y), (x, y + 1), and (x + 1, y) is {n, n + 1, n + 2} for some nonnegative integer n. Determine,
with proof, all possible labels for the point (2000, 2024).
Proposed by: Nithid Anchaleenukoon
Answer: 0, 3, 6, 9, . . . , 6048
Solution: We claim the answer is all multiples of 3 from 0 to 2000 + 2 · 2024 = 6048. First, we prove
no other values are possible. Let ℓ(x, y) denote the label of cell (x, y).

The label is divisible by 3.

Observe that for any x and y, ℓ(x, y), ℓ(x, y + 1), and ℓ(x + 1, y) are all distinct mod 3. Thus, for any
a and b, ℓ(a + 1, b + 1) cannot match ℓ(a + 1, b) or ℓ(a, b + 1) mod 3, so it must be equivalent to ℓ(a, b)
modulo 3.

Since ℓ(a, b + 1), ℓ(a, b + 2), ℓ(a + 1, b + 1) are all distinct mod 3, and ℓ(a + 1, b + 1) and ℓ(a, b) are
equivalent mod 3, then ℓ(a, b), ℓ(a, b + 1), ℓ(a, b + 2) are all distinct mod 3, and thus similarly ℓ(a, b +
1), ℓ(a, b + 2), ℓ(a, b + 3) are all distinct mod 3, which means that ℓ(a, b + 3) must be neither ℓ(a, b + 1)
or ℓ(a, b + 2) mod 3, and thus must be equal to ℓ(a, b) mod 3.

.. .. .. ... .. .. .. .. .. .. .. ..
. . . . . . . . . . .
··· ··· ··· ···
→ → →
··· 2 ··· 2 0 ··· 2 0 ···
0 ··· 0 1 ··· 0 1 ··· 0 1 2 ···

These together imply that

ℓ(w, x) ≡ ℓ(y, z) mod 3 ⇐⇒ w − x ≡ y − z mod 3.

It follows that ℓ(2000, 2024) must be equivalent to ℓ(0, 0) mod 3, which is a multiple of 3.

The label is at most 6048.

Note that since ℓ(x + 1, y), ℓ(x, y + 1), and ℓ(x, y) are 3 consecutive numbers, ℓ(x + 1, y) − ℓ(x, y)
and ℓ(x, y + 1) − ℓ(x, y) are both ≤ 2. Moreover, since ℓ(x + 1, y + 1) ≤ ℓ(x, y) + 4, since it is
also the same mod 3, it must be at most ℓ(x, y) + 3. Thus, ℓ(2000, 2000) ≤ ℓ(0, 0) + 3 · 2000, and
ℓ(2000, 2024) ≤ ℓ(2000, 2000) + 2 · 24, so ℓ(2000, 2024) ≤ 6048.

Construction.
Consider lines ℓn of the form x + 2y = n (so (2000, 2024) lies on ℓ6048 ). Then any three points of the
form (x, y), (x, y + 1), and (x + 1, y) lie on three consecutive lines ℓn , ℓn+1 , ℓn+2 in some order. Thus,
for any k which is a multiple of 3, if we label every point on line ℓi with max(i mod 3, i − k), any three
consecutive lines ℓn , ℓn+1 , ℓn+2 will either be labelled 0, 1, and 2 in some order, or n − k, n − k + 1,
n − k + 2, both of which consist of three consecutive numbers. Below is an example with k = 6.
 8
7 8 9 10 11 12 13 14 15
6
5 6 7 8 9 10 11 12 13
4
3 4 5 6 7 8 9 10 11
2
1 2 3 4 5 6 7 8 9
0 →
2 0 1 2 3 4 5 6 7
1
0 1 2 0 1 2 3 4 5
2
1 2 0 1 2 0 1 2 3
0
0 1 2 0 1 2 0 1

Any such labelling is valid, and letting k range from 0 to 6048, we see (2000, 2024) can take any label
of the form 6048 − k, which spans all such multiples of 3.

Hence the possible labels are precisely the multiples of 3 from 0 to 6048.

5. [40] Determine, with proof, whether there exist positive integers x and y such that x + y, x2 + y 2 , and
x3 + y 3 are all perfect squares.
Proposed by: Rishabh Das
Answer: Yes
Solution: Take (x, y) = (184, 345). Then x + y = 232 , x2 + y 2 = 3912 , and x3 + y 3 = 68772 .
Remark. We need x + y, x2 + y 2 , x2 − xy + y 2 to be perfect squares. We will find a, b such that
a2 + b2 , a2 − ab + b2 are perfect squares, and then let x = a(a + b) and y = b(a + b). Experimenting
with small Pythagorean triples gives a = 8, b = 15 as a solution.
Remark. The smallest solution we know of not of the form (184k 2 , 345k 2 ) is

(147 916 017 521 041, 184 783 370 001 360).

6. [45] Let Q be the set of rational numbers. Given a rational number a ̸= 0, find, with proof, all functions
f : Q → Q satisfying the equation

f (f (x) + ay) = af (y) + x

for all x, y ∈ Q.
Proposed by: Yichen Xiao


f (x) = x
Answer: f (x) = −x


f (x) = x + c for all rational numbers c iff a = 2

Solution: Let P (x, y) denote the functional equation. From P (x, 0), we have f (f (x)) = x + af (0).
Thus, the tripling trick gives f (x + af (0)) = f (f (f (x))) = f (x) + af (0).
 Now, here is the main idea: P (f (x), y) gives


f f (f (x)) + ay = af (y) + f (x)
f (x + af (0) + ay) = f (x) + af (y)
f (x + ay) = f (x) + af (y) − af (0).

In particular, plugging in x = 0 into this equation gives f (ay) = af (y) + (1 − a)f (0), so inserting it
back to the same equation gives

f (x + ay) = f (x) + f (ay) − f (0),

for all rational numbers x, y. In particular, the function g(x) = f (x) − f (0) is additive, so f is linear.
Let f (x) = bx + c. By substituting it in, we have P (x, y) iff

f (ay + bx + c) = a(by + c) + x
aby + b2 x + bc + c = aby + ac + x
(b2 − 1)x + (b + 1 − a)c = 0.

Since x is arbitrary, we can state that b2 − 1 = 0 and (b + 1 − a)c = 0, thus b = ±1. As a ̸= 0, we

know b + 1 − a = 0 only if b = 1 and a = 2. When a ̸= 2 or b ̸= 1, we know the only solutions are
b = ±1, c = 0, while for a = 2, b = 1, the equation is automatically satisfied, so the final answer is


f (x) = x
f (x) = −x


f (x) = x + c for all rational number c iff a = 2

Solution 2: We will only prove that f is linear. Then, proceed as in the end of Solution 1.
We know f (f (x)) = af (0)+x, so as af (0)+x can take any rational number when x takes every rational
number, the range of f is Q, and so f is surjective. If f (x1 ) = f (x2 ), we have x1 = f (f (x1 )) − af (0) =
f (f (x2 )) − af (0) = x2 , so x1 = x2 , implying f being injective. Thus, f is bijective.
From P (x, 0), we still get f (f (x)) = x + af (0).
Thus, from P (f −1 (0)), y/a), we can get f (y) = af (y/a) + f −1 (0). Plugging again P (f (x), y/a), we
have f (f (f (x)) + y) = f (x + y + af (0)) = af (y/a) + f (x) = f (x) + f (y) − f −1 (x).
Thus, we know f (x + y) = f (x − af (0)) + f (y) − f −1 (0) = f (x) + f (y) − f (0). Hence, the function
g(x) = f (x) − f (0) is additive, so g(x) = kx for some rational number k. Thus, f is a linear function,
and we can proceed as in above solution.

7. [50] Let ABCDEF be a regular hexagon with P as a point in its interior. Prove that of the three
values tan ∠AP D, tan ∠BP E, and tan ∠CP F , two of them sum to the third one.
Proposed by: Albert Wang
Solution 1:
 E D

X
P

O
F C
Y

Z
Q A B

WLOG let the side length of the hexagon be 1. Let O be the center of the hexagon. Consider drawing
in the circles (AP D), (BP E), and (CP F ). Note that O lies on the radical axis of all three circles,
since AO · OD = BO · OE = CO · OF . Since P also lies on the radical axis, all three circles are coaxial.
Let X, Y , and Z be the centers of (AP D), (BP E), and (CP F ), respectively. Since the circles are
coaxial, X, Y , and Z are collinear. WLOG Y lies on segment XZ. Note that XO ⊥ AD, Y O ⊥
BE =⇒ ∠XOY = 60◦ . Similarly, we have ∠Y OZ = 60◦ . Now, inverting at O and using van
Schooten’s Theorem gives that 1/OY = 1/OX + 1/OZ.
Furthermore, we have
1 AO 1
∠AP D = 180◦ − ∠AXD = 180◦ − ∠AXO =⇒ tan ∠AP D = − tan ∠AXO = − =− .
2 OX OX
Similarly, we have tan ∠BP E = − OY
1
and tan ∠CP F = − OZ
1
. Therefore, two of these tangent values
sum to the third, as desired.

Solution 2:
Firstly, note that AD, BE, and CF are diameters of the circle (ABCDEF ), so the angles ∠AP D,
∠BP E, ∠CP F are all obtuse. Therefore, the desired tangents are well-defined.
WLOG let the side length of the hexagon be 1. Let O be the center of the hexagon, and let OP = x.
Finally, let ∠AOP = θ.
√ √
Now, by Law of Cosines, we have AP = x2 + 1 − 2x cos θ and DP = x2 + 1 + 2x cos θ. Now, by
Law of Cosines again we have
2(x2 + 1) − 4 x2 − 1
cos ∠AP D = p =q
2 (x2 + 1)2 − 4x2 cos2 θ (1 − x2 )2 + 4x2 sin2 θ

2x| sin θ|
=⇒ tan ∠AP D = .
x2 − 1
 Similarly, ∠(BE, OP ) and ∠(CF, OP ) are θ +60◦ and θ +120◦ , respectively (here we are using directed
angles). Therefore, the desired three values are
 
2x| sin θ| 2x| sin(θ + 60◦ )| 2x| sin(θ + 120◦ )|
{tan ∠AP D, tan ∠BP E, tan ∠CP F } = , , .
x2 − 1 x2 − 1 x2 − 1
◦ ◦
We can scale down the tangent values by x22x −1 to get {| sin θ|, | sin(θ + 60 )|, | sin(θ + 120 )|}. Now,
consider an equilateral triangle with vertices at the third roots of unity rotated by θ degrees counter-
clockwise. The three values represent the distances from the three vertices to the real axis. Since the
centroid of this triangle is the origin (lying on the real axis), two of these quantities must sum to the
third, as desired.

Solution 3: We will show either the three sum to 0 or two of them sum to the third one; since they’re
all negative, the former case is actually impossible.
Let (ABCDEF ) be the unit circle, with a = 1, b = ω, and so on, where ω = eπi/3 . Then ∠AP D is
the argument of
1−p (1 − p)(ω 3 − p) ω 3 − p − ω 3 p + |p|2
= = .
ω3 − p (ω 3 − p)(ω 3 − p) 1 + |p|2 − ω 3 p − ω 3 p
Then tan ∠AP D is the imaginary part divided by the real part of this, which is
1 p−p
− · 2 = c · dist(P, AD)
i |p| − 1

for some constant c. (Note that tan AP D might actually be the negative of this, depending on direction;
this is why we added the remark at the beginning about them possibly summing to 0.)
Similarly, tan ∠BP E = c · dist(P, BE) and tan ∠CP F = c · dist(P, CF ). It suffices to show that two
dist(P, AD), dist(P, BE), and dist(P, CF ) sum to the third. However, this is easy; without loss of
generality let P be inside OAB, and let the hexagon have side length 1. Then
√
3
dist(P, AD) + dist(P, BE) = − dist(P, AB) = dist(P, CF ),
2
as desired.

8. [50] Let P be a point in the interior of quadrilateral ABCD such that the circumcircles of triangles
P DA, P AB, and P BC are pairwise distinct but congruent. Let the lines AD and BC meet at X. If
O is the circumcenter of triangle XCD, prove that OP ⊥ AB.
Proposed by: Pitchayut Saengrungkongka
Solution 1:
 D

C′ C ′′ C

B′ A B

Because the circles have equal radii, ∠P DA = ∠ABP , so if (P DA) intersects line AB again at a point
B ′ , then we have ∠P B ′ B = ∠P BB ′ , which means P B = P B ′ , similarly for the second intersection of
(P CB) with AB, A′ ; thus, (P DA) and (P CB) are congruent mirror images across the P -altitude, as
they are (P AB ′ ) and (P BA′ ), respectively.
Consider C ′ , the reflection of C across the P -altitude. We want to prove that C ′ lies on (XCD), as
then the circumcenter of (XCD) will lie on the perpendicular bisector of CC ′ . Because of our earlier
observation, C ′ , D, P, A are concyclic.
We present two approaches to finishing the angle chase from here:
• Add point C ′′ , the intersection of CC ′ with (XDA). Because AB is parallel to the line between
the centers, and so is C ′′ C, then ABCC ′′ is a parallelogram; thus,

∠C ′ DA = ∠C ′ C ′′ A = ∠C ′ CB = ∠C ′ CX.

• Add point B ′ , the reflection of B over the P -altitude. Note that B ′ lies on (XDA); in particular,
C ′ CBB ′ is an isosceles trapezoid, as C ′ B ′ is the reflection of CB over the P -altitude of △P AB.
Thus,
∠C ′ DA = 180 − ∠C ′ B ′ A = 180 − ∠C ′ B ′ B = ∠C ′ CB = ∠C ′ CX.
Remark. It is possible to do the last angle chasing without adding any additional points (beyond C ′ ).
However, the details are much messier.

Solution 2:
 H′

D′′
′
D
C′

C ′′

X′

A′ B′

Invert about P . Because the circles (P DA), (P AB), (P BC) all have equal radii and pass through P ,
the resulting lines D′ A′ , A′ B ′ , B ′ C ′ are equal distances away from P ; letting H ′ be the intersection of
lines D′ A′ and B ′ C ′ , it follows that P is an incenter or excenter of △A′ B ′ H ′ . Also, in the original
diagram, the P -altitude of △P AB includes the second intersection of the circles (P DA) and (P CB)
(as in the first solution); thus this P -altitude inverts to line P H ′ . Finally, X ′ is the intersection of
(P D′ A′ ) and (P B ′ C ′ ).
Note that the center of (XCD) lies on the P -altitude of P AB iff the inverse of the center of (X ′ C ′ D′ )
does. Thus we want to show that the center of (X ′ C ′ D′ ) lies on H ′ P . Let D′′ be the second intersection
of X ′ C ′ D′ with B ′ H ′ . Then

∡(D′ D′′ , H ′ P ) = ∡D′ D′′ H ′ + ∡D′′ H ′ P

= ∡D′ X ′ C ′ + ∡B ′ H ′ P
= ∡D′ X ′ P + ∡P X ′ C ′ + ∡P H ′ A′
= ∡D′ A′ P + ∡P B ′ C ′ + ∡P H ′ A′
= ∡P A′ B ′ + ∡P B ′ H ′ + ∡P H ′ A′
= 90◦

(where the last step follows from the fact that P is an incenter or excenter of △A′ B ′ H ′ ), and

∡H ′ D′ D′′ = 180◦ − ∡D′′ HD′ − ∡D′ D′′ H

= 180◦ − ∡D′′ H ′ P − (∡D′′ H ′ P + ∡D′ D′′ H ′ )
= 90◦ − ∡D′′ H ′ P
= ∡D′ D′′ H ′ ,

so △H ′ D′ D′′ is isosceles, and thus H ′ P is the perpendicular bisector of D′ D′′ . Thus the center of
(X ′ C ′ D′ D′′ ) lies on H ′ P , which means we’re done.

Solution 3:
 D
P
C

A B

B′

A′

Let A′ be the other intersection of line P A with (P DX) and B ′ be the other intersection of P B with
(P CX). Consider circles (P A′ B ′ ) and (XCD). Note that A and B have equal power with respect
to both circles, because of (P DA′ X) and (P CB ′ X). Thus, AB is the radical axis of the two cricles.
However,
∡P A′ X = ∡P DX = ∡P DA = ∡ABP
and
∡XB ′ P = ∡XCP = ∡BCP = ∡P AB,
where the last step follows from the fact that the circles have equal radii. Because ∡B ′ P A′ = ∡BP A,
it follows that A′ , X, B ′ are collinear, and in fact △P AB ∼ △P B ′ A′ . In particular, this means
that the P -altitude of △P AB passes through the circumcenter of △P B ′ A′ , as the circumcenter and
orthocenter are isogonal conjugates. Thus, as the circumcenter of P A′ B ′ lies on the P -altitude, and
the line between the centers of (P A′ B ′ ) and (XCD) must be perpendicular to their radical axis AB,
then the circumcenter of (XCD) must lie on the P -altitude as well, completing the proof.

9. [55] On each cell of a 200 × 200 grid, we place a car, which faces in one of the four cardinal directions.
In a move, one chooses a car that does not have a car immediately in front of it, and slides it one cell
forward. If a move would cause a car to exit the grid, the car is removed instead. The cars are placed
so that there exists a sequence of moves that eventually removes all the cars from the grid. Across all
such starting configurations, determine the maximum possible number of moves to do so.
Proposed by: Ethan Zhou
Answer: 6014950
Solution:
Let n = 100. The answer is 12 n(12n2 + 3n − 1) = 6014950.
A construction for an 8 × 8 grid instead (so n = 4):
 ▼ ▼ ▼ ▼ ◀ ◀ ◀ ◀
▼ ▼ ▼ ▼ ◀ ◀ ◀ ◀
▼ ▼ ▼ ▼ ◀ ◀ ◀ ◀
▼ ▼ ▼ ▼ ◀ ◀ ◀ ◀
▶ ▶ ▶ ▶ ▶ ▶ ▶ ▶
▶ ▶ ▶ ▶ ▲ ▲ ▲ ▲
▶ ▶ ▶ ▶ ▲ ▲ ▲ ▲
▶ ▶ ▶ ▶ ▲ ▲ ▲ ▲

Label the rows and columns from 1 to 2n, and let (r, c) denote the cell at row r, column c. The cars
can be cleared in the following order:

• Remove all cars in row n.

• For each row k = n − 1, . . . , 1, move the n upward-facing cars in row k once, then remove all
remaining cars in row k.
• Now all cars in the upper-left quarter of the grid can be removed, then those in the upper-right,
then those in the lower-right.

Moreover, this starting configuration indeed requires

n2 (3n + 1) n(n + 1) 1
4· − = n(12n2 + 3n − 1)
2 2 2
moves to clear.
Now we show this is the best possible. Take some starting configuration for which it is possible for all
cars to leave. For each car c, let d(c) denote the number of moves c makes before it exits. Partition
the grid into concentric square “rings” S1 , . . . , Sn , such that S1 consists of all cells on the border of
the grid, . . . , Sn consists of the four central cells:

▼ ▼ ▼ ▼ ◀ ◀ ◀ ◀
▼ ▼ ▼ ▼ ◀ ◀ ◀ ◀
▼ ▼ ▼ ▼ ◀ ◀ ◀ ◀
▼ ▼ ▼ ▼ ◀ ◀ ◀ ◀
▶ ▶ ▶ ▶ ▶ ▶ ▶ ▶
▶ ▶ ▶ ▶ ▲ ▲ ▲ ▲
▶ ▶ ▶ ▶ ▲ ▲ ▲ ▲
▶ ▶ ▶ ▶ ▲ ▲ ▲ ▲

Since all cars can be removed, each Sk contains some car c which points away from the ring, so that
d(c) = k. Now fix some ring Sk . Then:

• If car c is at a corner of Sk , we have d(c) ≤ 2n + 1 − k.

 • Each car c on the bottom edge of Sk , say at (x, k) for k < x < 2n + 1 − k, can be paired with the
opposing car c′ at (x, 2n + 1 − k). As c, c′ cannot point toward each other, we have

d(c) + d(c′ ) ≤ (2n + 1 − k) + max{x, 2n + 1 − x}.

Likewise, we can pair each car c at (k, x) with the opposing car c′ at (2n + 1 − k, x), getting the
same bound.
• If d(c) = k, then pairing it with the opposing car c′ gives d(c) + d(c′ ) ≤ 2n + 1. Note that this is
less than the previous bound, by at least

max{x, 2n + 1 − x} − k ≥ n + 1 − k > 0.

Summing the contributions d(c) from the four corners, each pair among the non-corner cars, and a
pair involving an outward-facing car gives
!
X X
n
d(c) ≤ 4(2n + 1 − k) + 4 [(2n + 1 − k) + (2n + 1 − x)] − (n + 1 − k).
c∈Sk x=k+1

One can verify that this evaluates to 12 n(12n2 + 3n − 1); alternatively, note that equality holds in our
construction, so summing over all 1 ≤ k ≤ n must yield the desired tight upper bound.

10. [60] Across all polynomials P such that P (n) is an integer for all integers n, determine, with proof, all
possible values of P (i), where i2 = −1.
Proposed by: Pitchayut Saengrungkongka
Answer: a + bi works if and only if a, b ∈ Q and νp (a), νp (b) ≥ 0 for all p ≡ 1 (mod 4).
Solution: We claim the answer is every complex number a + bi where a and b are rationals whose
simplified denominators are not multiples of any prime congruent to 1 modulo 4. The proof consists
of two main steps: proving that powers of p ≡ 1 mod 4 can’t appear in the denominator, and showing
all possible values are attainable. We show three different methods of the former part.
Impossibility via elementary number theory
We first show that no other values are possible. P Indeed, it
is well known that any polynomial that
m
maps Z into itself must be of the form P (n) = k=0 ak nk for integers ak and where we treat the
binomials as formal polynomials. This may be proved via finite differences.


It is therefore sufficient to show that for any k, ki can be simplified to a fraction of the form a+bi c ,
where a, b, c are integers and c is not divisible by any prime that is 1 modulo 4. We have that
 
i i · (i − 1) · . . . · (i − (k − 1))
= .
k k · (k − 1) · . . . · 1

Pick any prime p that is 1 modulo 4. Since p is 1 modulo 4, there exist distinct residue classes x, y
modulo p so that x2 ≡ y 2 ≡ −1 mod p. We will show that for every integer, r ≤ k divisible by p in the
denominator, we can pair it with a disjoint set, {rx , ry }, of two positive integers less than k in these
two residue classes so that (i − rx )(i − ry ) has real and complex parts divisible by the highest power
of p dividing r. Thus any factor of p that is 1 modulo 4 in the denominator, exists in the numerator
as well, which suffices. Start with u = 1 and repeat the following process for increasing u until there
is nothing left to do.
For every positive integer j ≤ k such that pu |j, there exist unique jx′ , jy′ ∈ [j − pu , j) satisfying
jx′ ≡ x mod p, jy′ ≡ y mod p and pu |jx′ + 1, jy′ + 1. So pair j with the set {jx′ , jy′ }, and pair its old
2 2

partners if any to the old partners of jx′ and jy′ .

The important feature of this assignment process is that we always have at step u, pmin(u,vp (r)) |rx +
ry , rx2 + 1, ry2 + 1. Thus at the end of the process, (i − rx )(i − ry ) = 12 ((rx + ry )2 − (rx2 + 1) − (ry2 + 1)) −
(rx + ry )i has real and complex parts divisible by pvp (r) as claimed.
Impossibility via Gaussian Integers
We work in the ring of Gaussian Integers Z[i] = {a + bi : a, b ∈ Z}, which is sitting inside number field
Q[i] = {a + bi : a, b ∈ Q}. It’s well-known that Z[i] is a unique factorization domain. For any Gaussian
prime π, let νpi (z) denote the exponent of π in the factorization of z.
Let p ≡ 1 (mod 4) be a prime. It’s well known that p splits into two Gaussian primes, p = ππ. Note
that it suffices to show
   
k k
νπ (i(i − 1)(i − 2) . . . (i − k + 1)) ≥ νπ (k!) = + 2 + . . . (∗)
p p

because the similar statement for π will follow. The key claim is the following:

Claim. For any integer t ≥ 1 and n, at least one of numbers i − n − 1, i − n − 2, . . . , i − n − pt is

divisible by π t .

Proof. First, we show that there exists integer r such that π t | i − r. To that end, by Hensel’s lemma,
there exists an integer s for which pt | s2 + 1. Thus,

π t π t | (s − i)(s + i).

However, gcd(s − i, s + i) | 2, so π t must divide either s − i or s + i. In particular, r = s or r = −s

work.
To complete the problem, pick the unique x ∈ {1, 2, . . . , pt } such that n+x ≡ r (mod p)t , so i−(n+x) ≡
i − r (mod pt ) and hence divisible by π t .

Using the claim repeatedly, we find that among numbers i, i − 1, i − 2, . . . , i − k + 1,

j k j k
• at least kp are divisible by π (this is by selecting kp disjoint contiguous block of size p),
j k
• at least pk2 are divisible by π 2 ,
j k
• at least pk3 are divisible by π 3 ,
..
• .
Using these altogether suffices to prove (∗).
Impossibility via p-adics
Let P (i) = a + bi, and note that P (−i) = a − bi. Fix some p ≡ 1 mod 4, and consider the p-adic
integers Zp lying in Qp . Note that x2 + 1 = 0 has a solution in Zp , and hence ±i ∈ Zp . Now take
a sequence of integers mk converging to i in Zp , so −mk converges to −i. Then since polynomials
are continuous, P (mk ) converges to P (i) and P (−mk ) converges to P (−i), so P (mk ) + P (−mk )
and P (mk ) − P (−mk ) converge to 2x and 2yi respectively. Finally, since P (mk ) and P (−mk ) are
integers, they have nonnegative p-adic valuation, and so by continuity, 2x and 2y have nonegative
p-adic valuation. Thus, when written as simplified fractions, a and b cannot have any powers of p in
their denominator, as desired.
Construction
We now show that all of the claimed values are possible. The set of polynomials, P , taking Z to itself is
closed under addition and multiplication, and therefore so is the set of possible values of P (i). It clearly
contains Z[i] by taking linear polynomials. Thus it suffices to show that p−1 is attainable for every
prime p that is not 1 modulo 4. 2−1 is achieved by taking P (x) = 41 x(x−1)(x−2)(x−3)+3, so we may
focus our attention only on the case where p is 3 modulo 4. It is then further sufficient to show that some
(a+bi)p−1 is attainable for a, b ∈ Z not both divisible by p because then (a−bi)·(a+bi)p−1 = (a2 +b2 )p−1
is also obtainable and cannot be an integer since −1 is not a quadratic residue modulo p, so Bézout’s
Theorem shows that p−1 is attainable. Now with this goal in mind, observe that
 
i 12 · (12 + 12 ) · . . . · (12 + (p − 1)2 )
= .
p p · (2p − 1) · . . . · 1

has denominator divisible by p, but numerator not divisible by p
since again, −1 is not a quadratic
residue modulo p. Hence we can find some integer m so that m pi = (a + bi)p−1 where a and b that
aren’t both divisible by p as desired.
..........................................................................................................
HMMT February 2024, February 17, 2024 — GUTS ROUND

Organization Team Team ID#

1. [5] Compute the sum of all integers n such that n2 − 3000 is a perfect square.
2. [5] Jerry and Neil have a 3-sided die that rolls the numbers 1, 2, and 3, each with probability 31 . Jerry
rolls first, then Neil rolls the die repeatedly until his number is at least as large as Jerry’s. Compute the
probability that Neil’s final number is 3.
3. [5] Compute the number of even positive integers n ≤ 2024 such that 1, 2, . . . , n can be split into n
2 pairs,
and the sum of the numbers in each pair is a multiple of 3.
4. [5] Equilateral triangles ABF and BCG are constructed outside regular pentagon ABCDE. Compute
∠F EG.

..........................................................................................................
HMMT February 2024, February 17, 2024 — GUTS ROUND

Organization Team Team ID#

5. [6] Let a, b, and c be real numbers such that

a + b + c = 100,
ab + bc + ca = 20, and
(a + b)(a + c) = 24.

Compute all possible values of bc.

6. [6] In triangle ABC, points M and N are the midpoints of AB and AC, respectively, and points P and
Q trisect BC. Given that A, M , N , P , and Q lie on a circle and BC = 1, compute the area of triangle
ABC.

7. [6] Positive integers a, b, and c have the property that ab , bc , and ca end in 4, 2, and 9, respectively.
Compute the minimum possible value of a + b + c.
8. [6] Three points, A, B, and C, are selected independently and uniformly at random from the interior of
a unit square. Compute the expected value of ∠ABC.
 ..........................................................................................................
HMMT February 2024, February 17, 2024 — GUTS ROUND

Organization Team Team ID#

9. [7] Compute the sum of all positive integers n such that n2 − 3000 is a perfect square.
10. [7] Alice, Bob, and Charlie are playing a game with 6 cards numbered 1 through 6. Each player is dealt
2 cards uniformly at random. On each player’s turn, they play one of their cards, and the winner is
the person who plays the median of the three cards played. Charlie goes last, so Alice and Bob decide
to tell their cards to each other, trying to prevent him from winning whenever possible. Compute the
probability that Charlie wins regardless.
11. [7] Let ABCD be a rectangle such that AB = 20 and AD = 24. Point P lies inside ABCD such that
triangles P AC and P BD have areas 20 and 24, respectively. Compute all possible areas of triangle P AB.

12. [7] Compute the number of quadruples (a, b, c, d) of positive integers satisfying

12a + 21b + 28c + 84d = 2024.

..........................................................................................................
HMMT February 2024, February 17, 2024 — GUTS ROUND

Organization Team Team ID#

13. [9] Mark has a cursed six-sided die that never rolls the same number twice in a row, and all other outcomes
are equally likely. Compute the expected number of rolls it takes for Mark to roll every number at least
once.
14. [9] Compute the smallest positive integer such that, no matter how you rearrange its digits (in base ten),
the resulting number is a multiple of 63.
15. [9] Let a ⋆ b = ab − 2. Compute the remainder when (((579 ⋆ 569) ⋆ 559) ⋆ · · · ⋆ 19) ⋆ 9 is divided by 100.

16. [9] Let ABC be an acute isosceles triangle with orthocenter H. Let M and N be the midpoints of sides
AB and AC, respectively. The circumcircle of triangle M HN intersects line BC at two points X and Y .
Given XY = AB = AC = 2, compute BC 2 .
 ..........................................................................................................
HMMT February 2024, February 17, 2024 — GUTS ROUND

Organization Team Team ID#

17. [11] The numbers 1, 2, . . . , 20 are put into a hat. Claire draws two numbers from the hat uniformly at
random, a < b, and then puts them back into the hat. Then, William draws two numbers from the hat
uniformly at random, c < d.
Let N denote the number of integers n that satisfy exactly one of a ≤ n ≤ b and c ≤ n ≤ d. Compute
the probability N is even.
18. [11] An ordered pair (a, b) of positive integers is called spicy if gcd(a + b, ab + 1) = 1. Compute the
probability that both (99, n) and (101, n) are spicy when n is chosen from {1, 2, . . . , 2024!} uniformly at
random.

19. [11] Let A1 A2 . . . A19 be a regular nonadecagon. Lines A1 A5 and A3 A4 meet at X. Compute ∠A7 XA5 .
√
20. [11] Compute 4 55083 + 56253 + 57423 , given that it is an integer.

..........................................................................................................
HMMT February 2024, February 17, 2024 — GUTS ROUND

Organization Team Team ID#

21. [12] Kelvin the frog currently sits at (0, 0) in the coordinate plane. If Kelvin is at (x, y), either he can
walk to any of (x, y + 1), (x + 1, y), or (x + 1, y + 1), or he can jump to any of (x, y + 2), (x + 2, y) or
(x + 1, y + 1). Walking and jumping from (x, y) to (x + 1, y + 1) are considered distinct actions. Compute
the number of ways Kelvin can reach (6, 8).
22. [12] Let x < y be positive real numbers such that
√ √ √ √
x+ y=4 and x+2+ y + 2 = 5.

Compute x.
23. [12] Let ℓ and m be two non-coplanar lines in space, and let P1 be a point on ℓ. Let P2 be the point on
m closest to P1 , P3 be the point on ℓ closest to P2 , P4 be the point on m closest to P3 , and P5 be the
point on ℓ closest to P4 . Given that P1 P2 = 5, P2 P3 = 3, and P3 P4 = 2, compute P4 P5 .

24. [12] A circle is tangent to both branches of the hyperbola x2 − 20y 2 = 24 as well as the x-axis. Compute
the area of this circle.
 ..........................................................................................................
HMMT February 2024, February 17, 2024 — GUTS ROUND

Organization Team Team ID#

25. [14] Point P is inside a square ABCD such that ∠AP B = 135◦ , P C = 12, and P D = 15. Compute the
area of this square.

26. [14] It can be shown that there exists a unique polynomial P in two variables such that for all positive
integers m and n,
∑
m ∑ n
P (m, n) = (i + j)7 .
i=1 j=1

Compute P (3, −3).

27. [14] A deck of 100 cards is labeled 1, 2, . . . , 100 from top to bottom. The top two cards are drawn; one
of them is discarded at random, and the other is inserted back at the bottom of the deck. This process
is repeated until only one card remains in the deck. Compute the expected value of the label of the
remaining card.
28. [14] Given that the 32-digit integer

64 312 311 692 944 269 609 355 712 372 657

is the product of 6 consecutive primes, compute the sum of these 6 primes.

..........................................................................................................
HMMT February 2024, February 17, 2024 — GUTS ROUND

Organization Team Team ID#

29. [16] For each prime p, a polynomial P (x) with rational coefficients is called p-good if and only if there
exist three integers a, b, and c such that 0 ≤ a < b < c < p3 and p divides all the numerators of P (a),
P (b), and P (c), when written in simplest form. Compute the number of ordered pairs (r, s) of rational
numbers such that the polynomial x3 + 10x2 + rx + s is p-good for infinitely many primes p.
30. [16] Let ABC be an equilateral triangle with side length 1. Points D, E, F lie inside triangle ABC such
that A, E, F are collinear, B, F , D are collinear, C, D, E are collinear, and triangle DEF is equilateral.
Suppose that there exists a unique equilateral triangle XY Z with X on side BC, Y on side AB, and Z
on side AC such that D lies on side XZ, E lies on side Y Z, and F lies on side XY . Compute AZ.

31. [16] Ash and Gary independently come up with their own lineups of 15 fire, grass, and water monsters.
Then, the first monster of both lineups will fight, with fire beating grass, grass beating water, and water
beating fire. The defeated monster is then substituted with the next one from their team’s lineup; if there
is a draw, both monsters get defeated.
Gary completes his lineup randomly, with each monster being equally likely to be any of the three types.
Without seeing Gary’s lineup, Ash chooses a lineup that maximizes the probability p that his monsters
are the last ones standing. Compute p.
32. [16] Over all pairs of complex numbers (x, y) satisfying the equations

x + 2y 2 = x4 and y + 2x2 = y 4 ,

compute the minimum possible real part of x.

 ..........................................................................................................
HMMT February 2024, February 17, 2024 — GUTS ROUND

Organization Team Team ID#

33. [20] Let p denote the proportion of teams, out of all participating teams, who submitted a negative
response to problem 5 of the Team round (e.g. “there are no such integers”). Estimate P = ⌊10000p⌋.
An estimate of E earns max(0, ⌊20 − |P − E|/20⌋) points.
If you have forgotten, problem 5 of the Team round was the following: “Determine, with proof, whether
there exist positive integers x and y such that x + y, x2 + y 2 , and x3 + y 3 are all perfect squares.”

34. [20] Estimate the number of positive integers n ≤ 106 such that n2 + 1 has a prime factor greater than n.
Submit
( ⌊ a positive integer E.⌋) If the correct answer is A, you will receive
( )5
E 106 −E
max 0, 20 · min A , 106 −A + 0.5 points.

35. [20] Barry picks infinitely many points inside a unit circle, each independently and uniformly at random,
P1 , P2 , . . . . Compute the expected value of N , where N is the smallest integer such that PN +1 is inside
the convex hull formed by the points P1 , P2 , . . . , PN .
Submit a positive real number E. If the correct answer is A, you will receive
⌊100 · max(0.2099 − |E − A|, 0)⌋ points.
36. [20] Let ABC be a triangle. The following diagram contains points P1 , P2 , . . . , P7 , which are the following
triangle centers of triangle ABC in some order:
• the incenter I;
• the circumcenter O;
• the orthocenter H;
• the symmedian point L, which is the intersections of the reflections of B-median and C-median
across angle bisectors of ∠ABC and ∠ACB, respectively;
• the Gergonne point G, which is the intersection of lines from B and C to the tangency points of the
incircle with AC and AB, respectively;
• the Nagel point N , which is the intersection of line from B to the tangency point between B-excircle
and AC, and line from C to the tangency point between C-excircle and AB; and
• the Kosnita point K, which is the intersection of lines from B and C to the circumcenters of triangles
AOC and AOB, respectively.

P2
P1
P3
P4 P5
P6

P7

Note that the triangle ABC is not shown. Compute which triangle centers {I, O, H, L, G, N, K} corre-
sponds to Pk for k ∈ {1, 2, 3, 4, 5, 6, 7}.
Your answer should be a seven-character string containing I, O, H, L, G, N , K, or X for blank. For
instance, if you think P2 = H and P6 = L, you would answer XHXXXLX.
⌈ ⌉ If you attempt to identify
n > 0 points and get them all correct, then you will receive (n − 1)5/3 points. Otherwise, you will
receive 0 points.
 HMMT February 2024
February 17, 2024
Guts Round
1. [5] Compute the sum of all integers n such that n2 − 3000 is a perfect square.
Proposed by: Holden Mui
Answer: 0
Solution: If n2 − 3000 is a square, then (−n)2 − 3000 is also a square, so the sum is 0 .

2. [5] Jerry and Neil have a 3-sided die that rolls the numbers 1, 2, and 3, each with probability 13 . Jerry
rolls first, then Neil rolls the die repeatedly until his number is at least as large as Jerry’s. Compute
the probability that Neil’s final number is 3.
Proposed by: Rishabh Das
11
Answer: 18
1
Solution: If Jerry rolls k, then there is a 4−k probability that Neil’s number is 3, since Neil has an
equal chance of rolling any of the 4 − k integers not less than k. Thus, the answer is
 
1 1 1 11
1+ + = .
3 2 3 18

3. [5] Compute the number of even positive integers n ≤ 2024 such that 1, 2, . . . , n can be split into n
2
pairs, and the sum of the numbers in each pair is a multiple of 3.
Proposed by: Rishabh Das
Answer: 675
Solution: There have to be an even number of multiples of 3 at most n, so this means that n ≡ 0, 2
(mod 6). (We can also say that there should be an equal number of 1 (mod 3) and 2 (mod 3) numbers,
which gives the same restriction.)
We claim that all these work. We know there are an even number of multiples of 3, so we can pair
them; then we can pair 3k + 1 and 3k + 2 for all k.
2022
This means the answer is 3 + 1 = 675 .

4. [5] Equilateral triangles ABF and BCG are constructed outside regular pentagon ABCDE. Compute
∠F EG.
Proposed by: Karthik Venkata Vedula
Answer: 48◦ = 4π
15

Solution: We have ∠F EG = ∠AEG − ∠AEF . Since EG bisects ∠AED, we get ∠AEG = 54◦ .
Now, ∠EAF = 108◦ + 60◦ = 168◦ . Since triangle EAF is isosceles, this means ∠AEF = 6◦ , so the
answer is 54◦ − 6◦ = 48◦ .
 G F

C A

D E

5. [6] Let a, b, and c be real numbers such that

a + b + c = 100,
ab + bc + ca = 20, and
(a + b)(a + c) = 24.

Compute all possible values of bc.

Proposed by: Pitchayut Saengrungkongka
Answer: 224, −176
Solution: We first expand the left-hand-side of the third equation to get (a + b)(a + c) = a2 + ac +
ab + bc = 24. From this, we subtract the second equation to obtain a2 = 4, so a = ±2.
If a = 2, plugging into the first equation gives us b + c = 98 and plugging into the second equation
gives us 2(b + c) + bc = 20 ⇒ 2(98) + bc = 20 ⇒ bc = −176.
Then, if a = −2, plugging into the first equation gives us b + c = 102, and plugging into the second
equation gives us −2(b + c) + bc = 20 ⇒ −2(102) + bc = 20 ⇒ bc = 224.
Therefore, the possible values of bc are 224, −176 .

6. [6] In triangle ABC, points M and N are the midpoints of AB and AC, respectively, and points P
and Q trisect BC. Given that A, M , N , P , and Q lie on a circle and BC = 1, compute the area of
triangle ABC.
Proposed by: Rishabh Das
√
7
Answer: 12

Solution: Note that M P ∥ AQ, so AM P Q is an isosceles trapezoid. In particular, we have AM =

M B = BP = P Q = 13 , so AB = 23 . Thus ABC is isoceles with base 1 and legs 23 , and the height from
√ √
7 7
A to BC is 6 , so the area is 12 .
 A

M N

B C
Q P

7. [6] Positive integers a, b, and c have the property that ab , bc , and ca end in 4, 2, and 9, respectively.
Compute the minimum possible value of a + b + c.
Proposed by: Derek Liu
Answer: 17
Solution: This minimum is attained when (a, b, c) = (2, 2, 13). To show that we cannot do better,
observe that a must be even, so c ends in 3 or 7. If c ≥ 13, since a and b are even, it’s clear (2, 2, 13) is
optimal. Otherwise, c = 3 or c = 7, in which case bc can end in 2 only when b ends in 8. However, no
eighth power ends in 4, so we would need b ≥ 18 (and a ≥ 2), which makes the sum 2 + 18 + 3 = 23
larger than 17 .

8. [6] Three points, A, B, and C, are selected independently and uniformly at random from the interior
of a unit square. Compute the expected value of ∠ABC.
Proposed by: Akash Das
Answer: 60◦ = π
3

Solution: Since ∠ABC + ∠BCA + ∠CAB = 180◦ for all choices of A, B, and C, the expected value
is 60◦ .

9. [7] Compute the sum of all positive integers n such that n2 − 3000 is a perfect square.
Proposed by: Holden Mui, Pitchayut Saengrungkongka, Rishabh Das
Answer: 1872
Solution: Suppose n2 − 3000 = x2 , so n2 − x2 = 3000. This factors as (n − x)(n + x) = 3000. Thus,
we have n − x = 2a and n + x = 2b for some positive integers a, b such that ab = 750 and a < b.
Therefore, we have n = a + b, so the sum will be just sum of divisors of 750 = 2 · 3 · 53 , which is

(1 + 2)(1 + 3)(1 + 5 + 25 + 125) = 1872 .

Remark. Problem 1 and 9 have slightly different statements.

1. Compute the sum of all integers n such that n2 − 3000 is a perfect square.
9. Compute the sum of all positive integers n such that n2 − 3000 is a perfect square.
There are 86 teams participating in the Guts rounds. Of these, the distribution of answers to Problem
1 is as follows:
 • 49 teams submitted 0, the correct answer.
• 17 teams submitted 1872, the correct answer to Problem 9.
• 3 teams submitted 3744, twice the correct answer to Problem 9.
• 2 teams submitted each of 55 and 744.
• 1 team submitted each of the following answers: 20, 548, 1404, 1586, 1772, 1807, 1817, 1882, 2184,
2746, 4680, 7488, and 9360.
The distribution of answers to Problem 9 is as follows:
• 69 teams submitted 1872, the correct answer.
• 5 teams submitted 3744, twice the correct answer.
• 2 teams submitted 2184.
• 1 team submitted each of the following answers: 55, 205, 548, 1560, 1764, 1772, 1832, 1867, 1893,
and 3634.

10. [7] Alice, Bob, and Charlie are playing a game with 6 cards numbered 1 through 6. Each player is dealt
2 cards uniformly at random. On each player’s turn, they play one of their cards, and the winner is
the person who plays the median of the three cards played. Charlie goes last, so Alice and Bob decide
to tell their cards to each other, trying to prevent him from winning whenever possible. Compute the
probability that Charlie wins regardless.
Proposed by: Ethan Liu
2
Answer: 15

Solution: If Alice has a card that is adjacent to one of Bob’s, then Alice and Bob will play those cards
as one of them is guaranteed to win. If Alice and Bob do not have any adjacent cards, since Charlie
goes last, Charlie can always choose a card that will win.
Let A denote a card that is held by Alice and B denote a card that is held by Bob. We will consider
the ascneding order of which Alice and Bob’s cards are held.
If the ascending order in which Alice and Bob’s cards are held are ABAB or BABA, then Charlie
cannot win. In these 2 cases, there will always be 2 consecutive cards where one is held by Alice and the
other is held by Bob. Therefore, the only cases we need to consider are the ascending orders AABB,
ABBA, and their symmetric cases.
In the case AABB, we must make sure that the larger card Alice holds and the smaller card Bob holds
are not consecutive. Alice can thus have {1, 2}, {2, 3}, or {1, 3}. Casework on what Bob can have yields
5 different combinations of pairs of cards Alice and Bob can hold. Since this applies to the symmetric
case BBAA as well, we get 10 different combinations.
In the case ABBA, we see that Alice’s cards must be {1, 6} and Bob’s cards must be {3, 4}. Considering
the symmetric case BAAB as well, this gives us 2 more combinations.
Thus, there are 12 total possible combinations of Alice’s and Bob’s cards such that
Charlie

will win
regardless. The total number of ways to choose Alice’s and Bob’s cards is given by 62 42 = 90, so the
12 2
probability that Charlie is guaranteed to win is 90 = 15 .

11. [7] Let ABCD be a rectangle such that AB = 20 and AD = 24. Point P lies inside ABCD such that
triangles P AC and P BD have areas 20 and 24, respectively. Compute all possible areas of triangle
P AB.
 Proposed by: Pitchayut Saengrungkongka
Answer: 98, 118, 122, 142
Solution:

A D

P
O P

B C

There are four possible locations of P as shown in the diagram. Let O be the center. Then, [P AO] = 10
and [P BO] = 12. Thus, [P AB] = [AOB] ± [P AO] ± [P BO] = 120 ± 10 ± 12, giving the four values
98, 118, 122, and 142 .

12. [7] Compute the number of quadruples (a, b, c, d) of positive integers satisfying

12a + 21b + 28c + 84d = 2024.

Proposed by: Rishabh Das

Answer: 2024
Solution: Looking at the equation mod 7 gives a ≡ 3 (mod 7), so let a = 7a′ + 3. Then mod 4 gives
b ≡ 0 (mod 4), so let b = 4b′ . Finally, mod 3 gives c ≡ 2 (mod 3), so let c = 3c′ + 2.
Now our equation yields

84a′ + 84b′ + 84c′ + 84d = 2024 − 3 · 12 − 2 · 28 = 1932 =⇒ a′ + b′ + c′ + d = 23.

Since a, b, c, d are positive integers, we have a′ and c′ are nonnegative and b′ and d are positive. Thus,
let b′′ = b′ + 1 and
d′ = d + 1, so a′ , b′′ , c′ , d′ are nonnegative integers summing to 21. By stars and
bars, there are 24 3 = 2024 such solutions.

13. [9] Mark has a cursed six-sided die that never rolls the same number twice in a row, and all other
outcomes are equally likely. Compute the expected number of rolls it takes for Mark to roll every
number at least once.
Proposed by: Albert Wang
149
Answer: 12

Solution: Suppose Mark has already rolled n unique numbers, where 1 ≤ n ≤ 5. On the next roll,
there are 5 possible numbers he could get, with 6 − n of them being new. Therefore, the probability of
getting another unique number is 6−n 5 , so the expected number of rolls before getting another unique
5
number is 6−n . Since it always takes 1 roll to get the first number, the expected total number of rolls
 5 5 5 5 5 149
is 1 + 5 + 4 + 3 + 2 + 1 = 12 .

14. [9] Compute the smallest positive integer such that, no matter how you rearrange its digits (in base
ten), the resulting number is a multiple of 63.
Proposed by: Arul Kolla
Answer: 111 888
Solution: First, the number must be a multiple of 9 and 7. The first is easy to check and holds for
all permutations. Note that when two adjacent digits a and b are swapped, the number changes by
9(a − b) · 10k (we disregard sign), so 9(a − b) must also be a multiple of 63 for all digits a and b. In
particular, this is sufficient, since a permutation can be represented as a series of transpositions.
This means that a − b must be a multiple of 7 for all digits a and b, so either all digits are equal or
they are in {0, 7}, {1, 8}, or {2, 9}. We find the minimum for each case separately.
We first provide the following useful fact: the first repunit (numbers 1, 11, 111, …) that is a multiple
of 7 is 111111. This is because 10 mod 7 = 3, and 3 is a generator modulo 7 (of course, you can just
compute the powers of 3 by hand, and it will not take much longer).
If a number k · 1 . . . 1 is a multiple of 63, then either k or 1 . . . 1 is a multiple of 7; if it is k, then it’s
clear that we need 777 777 777 to make the sum a multiple of 9. If 1 . . . 1 is a multiple of 7, then it is
at least 111 111, then to make a multiple of 9, we need 333 333.
If the only digits are 7 and 0, then we need at least nine sevens to make the digit sum a multiple of
nine, which has more digits than 333 333.
If the only digits are 8 and 1, then we can note that since 8 and 1 are both 1 (mod 7), these numbers
are equivalent to the repunits modulo 7, so such numbers have at least six digits. The best such
six-digit number with digits summing to a multiple of 9 is 111 888, which is our new candidate.
If the only digits are 9 and 2, then by analogous logic such numbers have at least six digits. But the
smallest such number is 999 999, which is not better.
So our best answer is 111 888 . It works.

15. [9] Let a ⋆ b = ab − 2. Compute the remainder when (((579 ⋆ 569) ⋆ 559) ⋆ · · · ⋆ 19) ⋆ 9 is divided by 100.
Proposed by: Rishabh Das
Answer: 29
Solution: Note that
(10a + 9) ⋆ (10b + 9) = (100ab + 90a + 90b + 81) − 2 ≡ 90(a + b) + 79 (mod 100),
so throughout our process all numbers will end in 9, so we will just track the tens digit. Then the ”new
operation” is
a † b ≡ −(a + b) + 7 mod 10,
where a and b track the tens digits. Now
(a † b) † c ≡ (−(a + b) + 7) † c ≡ a + b − c mod 10.
Thus, our expression has tens digit congruent to
−0 + 1 − 2 + 3 − · · · − 54 + 55 − 56 − 57 + 7 ≡ −28 − 57 + 7 ≡ 2 mod 10,
making the answer 29 .
16. [9] Let ABC be an acute isosceles triangle with orthocenter H. Let M and N be the midpoints of
sides AB and AC, respectively. The circumcircle of triangle M HN intersects line BC at two points
X and Y . Given XY = AB = AC = 2, compute BC 2 .
Proposed by: Andrew Wen
√
Answer: 2( 17 − 1)
Solution:

H
M N

B Y D X C

Let D be the foot from A to BC, also the midpoint of BC. Note that DX = DY = M A = M B =
M D = N A = N C = N D = 1. Thus, M N XY is cyclic with circumcenter D and circumradius 1. H
lies on this circle too, hence DH = 1.
If we let DB = DC = x, then since △HBD ∼ △BDA,
√
p 17 − 1
BD2 = HD · AD =⇒ x2 = 4 − x2 =⇒ x4 = 4 − x2 =⇒ x2 = .
2
√
Our answer is BC 2 = (2x)2 = 4x2 = 2( 17 − 1)

17. [11] The numbers 1, 2, . . . , 20 are put into a hat. Claire draws two numbers from the hat uniformly
at random, a < b, and then puts them back into the hat. Then, William draws two numbers from the
hat uniformly at random, c < d.
Let N denote the number of integers n that satisfy exactly one of a ≤ n ≤ b and c ≤ n ≤ d. Compute
the probability N is even.
Proposed by: Rishabh Das
181
Answer: 361

Solution: The number of integers that satisfy exactly one of the two inequalities is equal to the
number of integers that satisfy the first one, plus the number of integers that satisfy the second one,
minus twice the number of integers that satisfy both. Parity-wise, this is just the number of integers
that satisfy the first one, plus the number of integers that satisfy the second one.
 The number of integers that satisfy the first one is b − a + 1. The probability this is even is 10
19 , and
9
odd is 19 . This means the answer is
102 + 92 181
= .
192 361

18. [11] An ordered pair (a, b) of positive integers is called spicy if gcd(a + b, ab + 1) = 1. Compute the
probability that both (99, n) and (101, n) are spicy when n is chosen from {1, 2, . . . , 2024!} uniformly
at random.
Proposed by: Pitchayut Saengrungkongka
96
Answer: 595

Solution: We claim that (a, b) is spicy if and only if both gcd(a+1, b−1) = 1 and gcd(a−1, b+1) = 1.
To prove the claim, we note that

gcd(a + b, ab + 1) = gcd(a + b, b(−b) + 1) = gcd(a + b, b2 − 1).

Hence, we have

gcd(a + b, ab + 1) = 1 ⇐⇒ gcd(a + b, b2 − 1) = 1
⇐⇒ gcd(a + b, b − 1) = 1 and gcd(a + b, b + 1) = 1
⇐⇒ gcd(a + 1, b − 1) = 1 and gcd(a − 1, b + 1) = 1,

proving the claim.

Thus, n works if and only if all following four conditions hold:
• gcd(n + 1, 98) = 1, or equivalently, n is neither −1 (mod 2) nor −1 (mod 7);
• gcd(n − 1, 100) = 1, or equivalently, n is neither 1 (mod 2) nor 1 (mod 5);
• gcd(n + 1, 100) = 1, or equivalently, n is neither −1 (mod 2) nor −1 (mod 5); and
• gcd(n − 1, 102) = 1, or equivalently, n is neither 1 (mod 2), 1 (mod 3), nor 1 (mod 17).
Thus, there are 1, 2, 3, 6, 17 possible residues modulo 2, 3, 5, 7, and 17, respectively. The residues are
uniformly distributed within {1, 2, . . . , 2024!}. Hence, the answer is 21 · 32 · 35 · 67 · 17
16 96
= 595 .

19. [11] Let A1 A2 . . . A19 be a regular nonadecagon. Lines A1 A5 and A3 A4 meet at X. Compute ∠A7 XA5 .
Proposed by: Nithid Anchaleenukoon
1170 ◦ 13π
Answer: 19 = 38

Solution:
 X

A5
A6 A4

A7 A3

A8 A2

A9 A1

Inscribing the nondecagon in a circle, note that

1 \ \ 1
∠A3 XA5 = (A1 A3 − A4 A5 ) = A\
5 A3 A4 = ∠A5 A3 X.
2 2
Thus A5 X = A5 A3 = A5 A7 , so
1 1
∠A7 XA5 = 90◦ − ∠XA5 A7 = ∠A1 A5 A7
2 2
1 1 13 1170 ◦
= A\ 1 A8 A7 = · · 360◦ = .
4 4 19 19

√
4
20. [11] Compute 55083 + 56253 + 57423 , given that it is an integer.
Proposed by: Rishabh Das
Answer: 855
Solution: Let a = 5625 = 752 and b = 117. Then we have

55083 + 52653 + 57423 = (a − b)3 + a3 + (a + b)3 = 3a3 + 6ab2 = 3a(a2 + 2b2 ).

We have 3a = 33 · 54 , so a2 + 2b2 = 34 · (6252 + 2 · 192 ) should be 3 times a fourth power. This means

6252 + 2 · 192 = 3x4

√ √
for some integer x. By parity, x must be odd, and also x2 3 ≈ 625. Approximating 3 even as 2, we
get x should be around 19. Then x = 17 is clearly too small, and x = 21 is too big. (You can also
check mod 7 for this latter one.) Thus, x = 19. The final answer is then

32 · 5 · 19 = 855 .

21. [12] Kelvin the frog currently sits at (0, 0) in the coordinate plane. If Kelvin is at (x, y), either he can
walk to any of (x, y + 1), (x + 1, y), or (x + 1, y + 1), or he can jump to any of (x, y + 2), (x + 2, y)
or (x + 1, y + 1). Walking and jumping from (x, y) to (x + 1, y + 1) are considered distinct actions.
Compute the number of ways Kelvin can reach (6, 8).
Proposed by: Derek Liu


Answer: 1831830 = 610 · 14
6


Solution: Observe there are 14 6 = 3003 up-right paths from (0, 0) to (6, 8), each of which are 14
steps long. Any two of these steps can be combined into one: U U , RR, and RU as jumps, and U R as
walking from (x, y) to (x + 1, y + 1). The number of ways to combine steps is the number of ways to
group 14 actions into singles and consecutive pairs, which is F15 = 610. Every path Kelvin can take
can be represented this way, so the answer is 610 · 3003 = 1831830 .

22. [12] Let x < y be positive real numbers such that

√ √ √ p
x+ y=4 and x+2+ y + 2 = 5.

Compute x.
Proposed by: Ethan Liu
49
Answer: 36

Solution: Adding and subtracting both equations gives

√ √ p √
x+2+ x+ y+2+ y =9
√ √ p √
x+2− x+ y+2− y =1
√ √ √ √ √ √ √ √
Substitute a = x + x + 2 and b = y + y + 2. Then since ( x + 2 + x)( x + 2 − x) = 2, we
have
a+b=9
2 2
+ =1
a b
Dividing the first equation by the second one gives

ab = 18, a = 3, b = 6
√ √ √ √ √
x+2+ x−( x+2− x) 3− 23
Lastly, x= 2 = 2 = 76 , so x = 49
36 .

23. [12] Let ℓ and m be two non-coplanar lines in space, and let P1 be a point on ℓ. Let P2 be the point
on m closest to P1 , P3 be the point on ℓ closest to P2 , P4 be the point on m closest to P3 , and P5 be
the point on ℓ closest to P4 . Given that P1 P2 = 5, P2 P3 = 3, and P3 P4 = 2, compute P4 P5 .
Proposed by: Luke Robitaille
√
39
Answer: 4

Solution: The figure below shows the situation of the problem when projected appropriately, which
will be explained later.
 ℓ

P1

√ 5 − h2
2
P3
√
9−

√ − h2
h2

4
P5
√ m
a2
− P2
h2
P4

Let a be the answer. By taking the z-axis to be the cross product of these two lines, we can let the lines
be on the planes z = 0 and z = h, respectively. Then, by projecting√ onto the√xy-plane, we √ get the above
diagram.
√ The projected lengths of the first four segments are 25 − h 2, 9 − h 2 , and 4 − h2 , and
a − h . By similar triangles, these lengths must form a geometric progression. Therefore, 25 − h2 ,
2 2

9 − h2 , 4 − h2 , a2 − h2 is a geometric progression. By taking consecutive differences, 16, 5, 4 − a2 is a

√
geometric progression. Hence, 4 − a2 = 25
16 =⇒ a = 39
4 .

24. [12] A circle is tangent to both branches of the hyperbola x2 −20y 2 = 24 as well as the x-axis. Compute
the area of this circle.
Proposed by: Karthik Venkata Vedula
Answer: 504π
Solution 1:
 y

Invert about the unit circle centered at the origin. ω turns into a horizontal line, and the hyperbola
turns into the following:

x2 20y 2
− 2 = 24 =⇒ x2 − 20y 2 = 24(x2 + y 2 )2 .
(x2 2
+y ) 2 (x + y 2 )2
=⇒ 24x4 + (48y 2 − 1)x2 + 24y 4 + 20y 2 = 0
=⇒ (48y 2 − 1)2 ≥ 4(24)(24y 4 + 20y 2 )
=⇒ 1 − 96y 2 ≥ 1920y 2
p
=⇒ y ≤ 1/2016.
p
This means that the horizontal line in question is y = 1/2016. This means that
√ the diameter of the
circle
√ is the reciprocal of the distance between the point and line, which is 2016, so the radius is
504, and the answer is 504π .

Solution 2: Let a be the y-coordinate of both tangency points to the hyperbola. Then, the equation
of the circle must be in the form
x2 − 20y 2 + c(y − a)2 = 24.
Comparing the y 2 -coefficient, we see that c = 21. Moreover, we need it to pass through (0, 0), so
21a2 = 24. Thus, the equation of the circle is

x2 + y 2 − 42ay + 21a2 = 24 =⇒ x2 + (y − 21a)2 = (21a)2 ,

so the radius is 21a, and the area is (441a2 )π = 504π .

25. [14] Point P is inside a square ABCD such that ∠AP B = 135◦ , P C = 12, and P D = 15. Compute
the area of this square.
Proposed by: Pitchayut Saengrungkongka
√
Answer: 123 + 6 119
Solution:
 D C

x
12

√ Q
2y 45◦
P
y
x 135◦ y

A B

Let x = AP and y = BP . Rotate △BAP by 90◦ around B to get △BCQ. Then, △BP Q is right-
isosceles, and from ∠BQC = 135◦ , we get ∠P QC = 90◦ . Therefore, by Pythagorean’s theorem,
P C 2 = x2 + 2y 2 . Similarly, P D2 = y 2 + 2x2 .
√
Thus, y 2 = 2P C 3−P D = 21, and similarly x2 = 102 =⇒ xy = 3 238.
2 2

Thus, by the Law of Cosines, the area of the square is

AB 2 = AP 2 + BP 2 − 2 cos(135◦ )(AP )(BP )
√
= x2 + y 2 + 2xy
√
= 123 + 6 119 .

26. [14] It can be shown that there exists a unique polynomial P in two variables such that for all positive
integers m and n,
X
m X n
P (m, n) = (i + j)7 .
i=1 j=1

Compute P (3, −3).

Proposed by: Pitchayut Saengrungkongka
Answer: −2445
Solution: Note that for integers m > 0, n > 1,
X
m
P (m, n) − P (m, n − 1) = (i + n)7 .
i=1

For any given positive integer m, both sides are a polynomial in n, so they must be equal as polynomials.
In particular,
X
3
P (3, x) − P (3, x − 1) = (i + x)7 = (x + 1)7 + (x + 2)7 + (x + 3)7
i=1

for all real x. Moreover, P (3, 1) − P (3, 0) = P (3, 1) =⇒ P (3, 0) = 0. Then

P (3, −3) = P (3, 0) − (17 + 27 + 37 ) − (07 + 17 + 27 ) − ((−1)7 + 07 + 17 )
= −37 − 2 · 27 − 2 = −2445 .
27. [14] A deck of 100 cards is labeled 1, 2, . . . , 100 from top to bottom. The top two cards are drawn;
one of them is discarded at random, and the other is inserted back at the bottom of the deck. This
process is repeated until only one card remains in the deck. Compute the expected value of the label
of the remaining card.
Proposed by: Albert Wang
467
Answer: 8

Solution 1: Note that we can just take averages: every time you draw one of two cards, the EV of
the resulting card is the average of the EVs of the two cards. This average must be of the form

2• · 1 + 2• · 2 + 2• · 3 + · · · + 2• · 100

where the 2• s add up to 1. Clearly, the cards further down in the deck get involved in one less layer of
averaging, and therefore 1 through 72 are weighted 2−7 while the rest are weighted 2−6 . To compute
the average now, we just add it up to get 4678 .

Solution 2: We see that in a deck with 2n cards, that after repeating the process 2n−1 times, that
each card has a chance of 12 of remaining in the deck. This means that the average of the cards in the
deck doesn’t change between 2n by 2n−1 cards. Thus, by repeating this process, we determine that
the expected value of the last card is the average of all cards whenever we start with 2n cards.
Suppose we instead start with 27 = 128 cards in the following order:

73, 73, 74, 74, . . . , 100, 100, 1, 2, 3, . . . , 72.

Thus, after 28 steps, we will be left with the original configuration. Since a power of 2 cards are in the
deck, we expect that the final card will be the average of these numbers. This is 467 8 .

28. [14] Given that the 32-digit integer

64 312 311 692 944 269 609 355 712 372 657

is the product of 6 consecutive primes, compute the sum of these 6 primes.

Proposed by: Derek Liu
Answer: 1200974
Solution: Because the product is approximately 64 · 1030 , we know the primes are all around 200000.
Say they are 200000 + xi for i = 1, . . . , 6.
Q6
By expanding i=1 (200000 + xi ) as a polynomial in 200000, we see that

31231 · 1025 = 2000005 (x1 + · · · + x6 )

plus the carry from the other terms. Note that 31231 = 975 · 32 + 31, so x1 + · · · + x6 ≤ 975.
Thus,
5 20
16(x1 x2 + x1 x3 + · · · + x5 x6 ) ≤ 16 · (x1 + · · · + x6 )2 < · 10002 < 67 · 105 ,
12 3
so the carry term from 2000004 (x1 x2 + · · · + x5 x6 ) is at most 67 · 1025 . The other terms have negligible
carry, so it is pretty clear x1 + · · · + x6 > 972, otherwise the carry term would have to be at least

31231 · 1025 − 2000005 (972) = 127 · 1025 .

 It follows that x1 + · · · + x6 lies in [973, 975], so the sum of the primes, 6 · 200000 + (x1 + · · · + x6 ), lies
in [1200973, 1200975].
As these primes are all greater than 2, they are all odd, so their sum is even. Thus it must be 1200974 .

29. [16] For each prime p, a polynomial P (x) with rational coefficients is called p-good if and only if there
exist three integers a, b, and c such that 0 ≤ a < b < c < p3 and p divides all the numerators of P (a),
P (b), and P (c), when written in simplest form. Compute the number of ordered pairs (r, s) of rational
numbers such that the polynomial x3 + 10x2 + rx + s is p-good for infinitely many primes p.
Proposed by: Pitchayut Saengrungkongka
Answer: 12
Solution: By Vieta, the sum of the roots is −10 (mod p). However, since the three roots are less than
p/3, it follows that the roots are (p − a′ )/3, (p − b′ )/3, (p − c′ )/3, where there are finitely many choices
a′ < b′ < c′ . By pigeonhole, one choice, say (u, v, w) must occur for infinitely many p. We then get
that the roots of P are −u/3, −v/3, and −w/3. Moreover, we must have that u, v, w are all 1 (mod 3)
or all 2 (mod 3), and by Vieta, we have u + v + w = 30.
The polynomial is then uniquely determined by u, v, w. Thus, it suffices to count triples u < v < w of
positive integers such that u, v, w are all 1 (mod 3) or all 2 (mod 3) and that u + v + w = 30. It’s not
very hard to list them all now.
When u, v, w ≡ 1 (mod 3), there are 7 triples: (1, 4, 25), (1, 7, 22), (1, 10, 19), (1, 13, 16), (4, 7, 19),
(4, 10, 16), and (7, 10, 13).
When u, v, w ≡ 2 (mod 3), there are 5 triples: (2, 5, 23), (2, 8, 20), (2, 11, 17), (5, 8, 17), and (5, 11, 14).
Hence, the answer is 7 + 5 = 12 .

30. [16] Let ABC be an equilateral triangle with side length 1. Points D, E, F lie inside triangle ABC
such that A, E, F are collinear, B, F , D are collinear, C, D, E are collinear, and triangle DEF is
equilateral. Suppose that there exists a unique equilateral triangle XY Z with X on side BC, Y on
side AB, and Z on side AC such that D lies on side XZ, E lies on side Y Z, and F lies on side XY .
Compute AZ.
Proposed by: Jaedon Whyte, Maxim Li
1√
Answer: 1+ 3 2

Solution:
 A

Z
E
Y
O
D
F

B X C

First, note that point X can be constructed from intersection of ⊙(DOF ) and side BC. Thus, if there
is a unique equilateral triangle, then we must have that ⊙(DOF ) is tangent to BC. Furthermore,
⊙(DOF ) is tangent to DE, so by equal tangents, we have CD = CX.
We now compute the answer. Let x = AZ = CX = CD = BF . Then, by power of point,
(1 − x)2
BF · BD = BX 2 =⇒ BD = .
x
Thus, by law of cosine on △BDC, we have that
 2
(1 − x)2 (1 − x)2
x2 + +x· =1
x x
(1 − x) 4
x2 + + (1 − x)2 =1
x2
(1 − x)4
= 2x(1 − x)
x2
1−x √
3
= 2
x
1
x= √ .
1+ 32

31. [16] Ash and Gary independently come up with their own lineups of 15 fire, grass, and water monsters.
Then, the first monster of both lineups will fight, with fire beating grass, grass beating water, and water
beating fire. The defeated monster is then substituted with the next one from their team’s lineup; if
there is a draw, both monsters get defeated.
Gary completes his lineup randomly, with each monster being equally likely to be any of the three
types. Without seeing Gary’s lineup, Ash chooses a lineup that maximizes the probability p that his
monsters are the last ones standing. Compute p.
Proposed by: Albert Wang
215
Answer: 1− 315
15
Solution: First, we show Ash cannot do better. Notice there is a 2315 chance that Gary’s i-th monster
ties or defeats Ash’s i-th monster for each i. If this is the case, Ash cannot win, as Ash’s i-th monster
 will always be defeated by Gary’s i-th monster, if not sooner. Thus, Ash wins with probability at most
15
1 − 2315 . It remains to show this is achievable.
Ash uses the lineup fire-grass-water repeated 5 times. Then, none of Gary’s monsters can defeat more
than one monster in Ash’s lineup, so Ash will win unless Gary manages to take down exactly one
monster with each of his. In particular, this means the i-th monster Gary has must tie or defeat Ash’s
i-th monster, which occurs with 23 chance with each i. Thus this construction achieves the answer of
215
1− 315 .

32. [16] Over all pairs of complex numbers (x, y) satisfying the equations

x + 2y 2 = x4 and y + 2x2 = y 4 ,

compute the minimum possible real part of x.

Proposed by: Jaedon Whyte
q √
3 1− 33
Answer: 2

Solution 1: Note the following observations:

(a) if (x, y) is a solution then (ωx, ω 2 y) is also a solution if ω 3 = 1 and ω ̸= 1.
(b) we have some solutions (x, x) where x is a solution of x4 − 2x2 − x = 0.
These are really the only necessary observations and the first does not need to be noticed immediately.
Indeed, we can try to solve this directly as follows: first, from the first equation, we get y 2 = 21 (x4 − x),
so inserting this into the second equation gives
1 4
(x − x)2 − 2x2 = y
4

2 2
x4 − x − 8x2 − 8x4 + 8x = 0

2
x8 − 2x5 − 7x2 − 8x4 + 8x = 0
x16 + · · · + 41x4 + 8x = 0
| {z }
P (x)

By the second observation, we have that x(x3 − 2x − 1) should be a factor of P . The first observation
gives that (x3 − 2ωx − 1)(x3 − 2ω 2 x − 1) should therefore also be factor. Now (x3 − 2ωx − 1)(x3 −
2ω 2 x − 1) = x6 + 2x4 − 2x3 + 4x2 − 2x + 1 since ω and ω 2 are roots of x2 + x + 1. So now we see
that the last two terms of the product of all of these is −5x4 − x. Hence the last two terms of the
polynomial we get after dividing out should be −x3 − 8, and given what we know about the degree
q be exactly x − x − 8 which has roots being
6 3
and the fact that everything is monic, the quotient must
√
the cube roots of the roots to x2 − x − 8, which are 3 1± 33
2 . Now x3 − 2x − 1 is further factorable as
√
(x − 1)(x2 − x − 1) with roots 1, 1±2 5 so it is not difficult to compare the real parts of all roots of P ,
especially since 5 are real and non-zero, and we have that Re(ωx) = − 21 x if x ∈ R. We conclude that
q √
3 1− 33
the smallest is 2 .

Solution 2: Subtracting the second equation from the first, we get:

(y + 2x2 ) − (x + 2y 2 ) = y 4 − x4 =⇒
(x − y) + 2(x2 − y 2 ) = (x2 − y 2 )(x2 + y 2 ) =⇒
 (x − y)(1 − (x + y)(x2 + y 2 + 2)) = 0
Subtracting y times the first equation from x times the second, we get:
(xy + 2y 3 ) − (xy + 2x3 ) = x4 y − xy 4 =⇒
2(y 3 − x3 ) = xy(x3 − y 3 ) =⇒
(x3 − y 3 )(2 + xy) = 0
Subtracting y 2 times the second equation from x2 times the first, we get:
(x3 + 2x2 y 2 ) − (y 3 + 2x2 y 2 ) = x6 − y 6 =⇒
x3 − y 3 = (x3 + y 3 )(x3 − y 3 ) =⇒
(x3 − y 3 )(1 − x3 − y 3 ) = 0
We have three cases.
Case 0. x = 0 Thus, (x, y) = (0, 0) is the only valid solution.
Case 1. x = ωy for some third root of unity ω. Thus, y 2 = ω 4 x2 = ωx2

x + 2y 2 = x4 =⇒
x + 2ωx2 = x4 =⇒
x(1 + ω)(2 − ωx2 ) = 1
Note that x = −ω is always a solution to the above, and so we can factor as:
x3 + 2(1 + ω)x − 1 = 0
(x + ω)(x2 − ωx − ω 2 ) = 0
and so the other solutions are of the form:
√
1± 5
x= ·ω
2
√
for the third root of unity. The minimum real part in this case is − 1+2 5
when ω = 1.
Case 2. Since x − y ̸= 0, we have xy = −2 and x + y = 1.
3 3 3 3
p √  √ 
Thus, x3 − y 3 = (x3 + y 3 )2 − 4(xy)2 = ± 33 =⇒ x3 = 1±2 33
 √ 1/3
1− 33
This yields the minimum solution of x = 2 as desired. This is satisfied by letting
 √ 1/3
y = 1+2 33 .

33. [20] Let p denote the proportion of teams, out of all participating teams, who submitted a negative
response to problem 5 of the Team round (e.g. “there are no such integers”). Estimate P = ⌊10000p⌋.
An estimate of E earns max(0, ⌊20 − |P − E|/20⌋) points.
If you have forgotten, problem 5 of the Team round was the following: “Determine, with proof, whether
there exist positive integers x and y such that x + y, x2 + y 2 , and x3 + y 3 are all perfect squares.”
Proposed by: Arul Kolla
Answer: 5568
Solution: Of the 88 teams competing in this year’s Team round, 49 of them answered negatively,
9 (correctly) provided a construction, 16 answered ambiguously or did not provide a construction, and
the remaining 14 teams did not submit to problem 5. Thus p = 49 88 ≈ 0.5568.
34. [20] Estimate the number of positive integers n ≤ 106 such that n2 + 1 has a prime factor greater than
n.
   5 
E 106 −E
Submit a positive integer E. If the correct answer is A, you will receive max 0, 20 · min A , 106 −A + 0.5
points.
Proposed by: Pitchayut Saengrungkongka
Answer: 757575
Solution: Let N denote 106 . We count by summing over potential prime factors p.
For any prime p > 2, we have that p | n2 + 1 for two values of n if p ≡ 1 (mod 4), and zero values
otherwise. Pretending these values are equally 
likely to
 be any of 1, . . . , p, we expect the number of n
2N
corresponding to a 1 (mod 4) prime to be min 2, p .
The number of primes up to x is, by the Prime Number Theorem logx x . Assuming around half of the
1
prime numbers are 1 (mod 4), we on average expect some x to be a 1 (mod 4) prime 2 log x of the time.
2
Approximating by an integral over potential primes x from 1 to N , using our approximations, gives
Z N2  
2N dx
min 2, · .
1 x 2 log x

We now approximately calculate this integral as follows:

Z N2   Z N Z N2
2N dx dx N
min 2, · = + dx
1 x 2 log x 1 log x N x log x
N
≈ + N (log log(N 2 ) − log log N )
log N
N
= + N log 2.
log N

Here, for the first integral, we estimate log x on [1, N ] by log N , and for the second integral, we use
1
that the antiderivative of x log x is log log x.
Using log 2 ≈ 0.7, one can estimate

log N = 2 log 1000 ≈ 20 log 2 ≈ 14

giving a final estimate of

106 /14 + 106 · 0.7 = 771428.
This estimate yields a score of 15. If one uses the closer estimate log 2 ≈ 0.69, one gets the final
estimate of 761428, yielding a score of 18.
Here is a code using sympy to calculate the final answer:
from sympy.ntheory import factorint
cnt = 0
for n in range(1,10 ** 6+1):
if max(factorint(n ** 2+1, multiple=True)) > n:
cnt += 1
print(cnt)

35. [20] Barry picks infinitely many points inside a unit circle, each independently and uniformly at
random, P1 , P2 , . . . . Compute the expected value of N , where N is the smallest integer such that
PN +1 is inside the convex hull formed by the points P1 , P2 , . . . , PN .
Submit a positive real number E. If the correct answer is A, you will receive ⌊100 · max(0.2099 − |E − A|, 0)⌋
points.
Proposed by: Albert Wang, Rishabh Das
Answer: 6.54
Solution: Clearly, N ≥ 3, and let’s scale the circle to have area 1. We can see that the probability
to not reach N = 4 is equal to the probability that the fourth point is inside the convex hull of the
past three points. That is, the probability is just one minus the expected area of those N points. The
area of this turns out to be really small, and is around 0.074, and so (1 − 0.074) of all sequences of
points make it to N = 4. The probability to reach to the fifth point from there should be around
(1 − 0.074)(1 − 0.074 · 2), as any four points in convex configuration can be covered with 2 triangles.
Similarly, the chance of reaching N = 6 should be around (1 − 0.074)(1 − 0.074 · 2)(1 − 0.074 · 3), and
so on. Noting that our terms eventually decay to zero around term 1/0.074 = 13, our answer should
be an underestimate. In particular, we get

3 + (1 − 0.074)(1 + (1 − 0.074 · 2)(1 + (1 − 0.074 · 3)(1 + · · · ))) ≈ 6.3.

Guessing anything slightly above this lower bound should give a positive score.
Here is a Python code that simulates the result.
from random import randrange ,getrandbits
import itertools , math
from tqdm import tqdm
import numpy as np

DEBUG = False

def unit_circle_pt ():

while True:
x = randrange(-(2 ** 32),2 ** 32+1)
y = randrange(-(2 ** 32),2 ** 32+1)
if x*x + y*y < 2 ** 64:
return (x,y)

def area_of_triangle(p1 , p2 , p3):

return abs((p2[0] - p1[0])*(p3[1] - p2[1]) - (p2[1] - p1[1])*(p3[0] - p2[0]))

def pt_inside_polygon (point , polygon):

# point is a pair
# polygon is an angle - sorted list of points that are the vertices of a convex polygon in
some order

# area of polygon

# plot the polygon and the point

if DEBUG:
import matplotlib.pyplot as plt
# plot the big circle
circle = plt.Circle ((0,0), 2 ** 32 , color='b', fill=False)
# fix view to circle
plt.xlim(-2 ** 32 ,2 ** 32)
plt.ylim(-2 ** 32 ,2 ** 32)
plt.gca().add_artist(circle)
# make the window to scale
plt.gca().set_aspect('equal ', adjustable='box')
plt.plot([x for x,_ in polygon], [y for _,y in polygon])
plt.scatter([point[0]], [point[1]])
plt.show ()

area = 0
for i in range(1, len(polygon)-1):
# add on area between points 0, i, i+1
area += area_of_triangle(polygon[0], polygon[i], polygon[i+1])
 # point is inside polygon if the area of the triangles formed by the point and each edge
of the polygon sum to the area of the
polygon
area_sum = 0
for i in range(len(polygon)):
# add on area between points point , polygon [i], polygon [(i+1)%len( polygon )]
area_sum += area_of_triangle(point , polygon[i], polygon[(i+1)%len(polygon)])

return area_sum == area

def convex_hull(points):
# sort by x, then y
points = sorted(points , key=lambda x: (x[0], x[1]))
# graham scan

# find the lowest point

lowest = points[0]
for p in points:
if p[1] < lowest[1]:
lowest = p
# sort by angle
points = sorted(points , key=lambda x: (math.atan2(x[1]-lowest[1], x[0]-lowest[0]), -x[1]
, x[0]))
# remove duplicates
points = list(k for k,_ in itertools.groupby(points))
# stack to hold the points
stack = []
for p in points:
while len(stack) > 1 and (stack[-1][0]-stack[-2][0])*(p[1]-stack[-2][1]) - (stack[-1
][1]-stack[-2][1])*(p[0]-stack[-2][
0]) <= 0:
stack.pop()
stack.append(p)
return stack

def pulse(horizon=1000):
cur = [unit_circle_pt () for _ in range(3)]

for N in range(3, horizon):

pt = unit_circle_pt ()
if pt_inside_polygon (pt , cur):
return N
cur = convex_hull(cur + [pt])

trials = 1000
blocks = 100000
cur_trials = 0
cur_sum = 0
results = []

for block in range(trials):

for _ in tqdm(range(blocks)):
results.append(pulse ())
cur_trials += blocks
mean = np.mean(results)
stddev = np.std(results)
z = 5.0
ci = (mean - z*stddev/np.sqrt(cur_trials), mean + z*stddev/np.sqrt(cur_trials))
print(block+1, mean , stddev , ci)

36. [20] Let ABC be a triangle. The following diagram contains points P1 , P2 , . . . , P7 , which are the
following triangle centers of triangle ABC in some order:
 • the incenter I;
• the circumcenter O;
• the orthocenter H;
• the symmedian point L, which is the intersections of the reflections of B-median and C-median
across angle bisectors of ∠ABC and ∠ACB, respectively;
• the Gergonne point G, which is the intersection of lines from B and C to the tangency points of
the incircle with AC and AB, respectively;
• the Nagel point N , which is the intersection of line from B to the tangency point between B-
excircle and AC, and line from C to the tangency point between C-excircle and AB; and
• the Kosnita point K, which is the intersection of lines from B and C to the circumcenters of
triangles AOC and AOB, respectively.

P2
P1
P3
P4 P5
P6

P7

Note that the triangle ABC is not shown. Compute which triangle centers {I, O, H, L, G, N, K}
corresponds to Pk for k ∈ {1, 2, 3, 4, 5, 6, 7}.
Your answer should be a seven-character string containing I, O, H, L, G, N , K, or X for blank. For
instance, if you think P2 = H and P6 = L, you would answer XHXXXLX.
  If you attempt to identify
n > 0 points and get them all correct, then you will receive (n − 1)5/3 points. Otherwise, you will
receive 0 points.
Proposed by: Kevin Zhao, Pitchayut Saengrungkongka
Answer: KOLIN GH
Solution: Let G′ be the centroid of triangle ABC. Recall the following.
• Points O, G′ , H lie on Euler’s line of △ABC with OG′ : G′ H = 1 : 2.
• Points I, G′ , N lie on Nagel’s line of △ABC with IG′ : G′ N = 1 : 2.
Thus, OI ∥ HN with OI : HN = 1 : 2. Therefore, we can detect parallel lines with ratio 2 : 1 in
the figure. The only possible pairs are P2 P4 ∥ P7 P5 . Therefore, there are two possibilities: (P2 , P7 )
and (P4 , P5 ) must be (O, H) and (I, N ) in some order. Intuitively, H should be further out, so it’s
not unreasonable to guess that P2 = O, P7 = H, P4 = I, and P5 = N . Alternatively, perform the
algorithm below with the other case to see if it fails.
To identify the remaining points, we recall that the isogonal conjugate of G and N both lie on OI
(they are insimilicenter and exsimilicenter of incircle and circumcircle, respectively). Thus, H, G, N, I
lie on isogonal conjugate of OI, known as the Feuerbach’s Hyperbola. It’s also known that OI is
tangent to this line, and this hyperbola have perpendicular asymptotes.
Using all information in the above paragraph, we can eyeball a rectangular hyperbola passing through
H, G, N, I and is tangent to OI. It’s then not hard to see that P6 = G.
Finally, we need to distinguish between symmedian and Kosnita points. To do that, recall that Kosnita
point is isogonal conjugate of the nine-point center (not hard to show). Thus, H, L, K, O lies on isogonal
conjugate of OH, which is the Jerabek’s Hyperbola. One can see that H, L, K, O lies on the same
branch. Moreover, they lie on this hyperbola in this order because the isogonal conjugates (in order)
are O, centroid, nine-point center, and H, which lies on OH in this order. Using this fact, we can
identity P5 = L and P1 = K, completing the identification.
The following is the diagram with the triangle ABC.

P2
P1
P3
P4 P5
P6
P7

B C

Here is the Asymptote code that generates the diagram in the problem.
import olympiad;
import geometry;
size(7.5cm);
pair A = (0.5,3.2);
pair B = (0,0);
pair C = (4,0);
pair O = circumcenter(triangle(A,B,C));
pair H = orthocentercenter (triangle(A,B,C));
pair L = symmedian(triangle(A,B,C));
pair Ge = gergonne(triangle(A,B,C));
pair I = incenter(triangle(A,B,C));
pair Na = A+B+C - 2I;
pair K = extension(A, circumcenter(B,O,C), B, circumcenter(A,O,C));

dot("$P_6$",Ge ,dir(-90));
dot("$P_2$",O,dir(90));
dot("$P_7$",H,dir(90));
dot("$P_3$",L,dir(135));
dot("$P_4$",I,dir(-45));
dot("$P_5$",Na ,dir(-90));
dot("$P_1$",K,dir(90));