The 1st US Ersatz Math Olympiad

Solutions and Results

Evan Chen, Brice Huang, and Carl Schildkraut

May 11, 2021

Contents
1 Summary 3
1.1 Thanks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Results 5
2.1 Top Scores . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Special awards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3 Honorable mentions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.4 Distinction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

3 Solutions to the Problems 8

3.1 Solution to USEMO1, by Robin Son . . . . . . . . . . . . . . . . . . . . . 8
3.2 Solution to USEMO2, by Carl Schildkraut . . . . . . . . . . . . . . . . . . 10
3.3 Solution to USEMO3, by Nikolai Beluhov . . . . . . . . . . . . . . . . . . 14
3.4 Solution to USEMO4, by Robin Son . . . . . . . . . . . . . . . . . . . . . 16
3.5 Solution to USEMO5, by Ankan Bhattacharya . . . . . . . . . . . . . . . 17
3.6 Solution to USEMO6, by Ankan Bhattacharya . . . . . . . . . . . . . . . 18

4 Marking schemes 21
4.1 Marking scheme for problem 1 . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2 Marking scheme for problem 2 . . . . . . . . . . . . . . . . . . . . . . . . 21
4.3 Marking scheme for problem 3 . . . . . . . . . . . . . . . . . . . . . . . . 22
4.4 Marking scheme for problem 4 . . . . . . . . . . . . . . . . . . . . . . . . 22
4.5 Marking scheme for problem 5 . . . . . . . . . . . . . . . . . . . . . . . . 23
4.6 Marking scheme for problem 6 . . . . . . . . . . . . . . . . . . . . . . . . 23

5 Statistics 25
5.1 Summary of scores for USEMO 2019 . . . . . . . . . . . . . . . . . . . . . 25
5.2 Problem statistics for USEMO 2019 . . . . . . . . . . . . . . . . . . . . . 25
5.3 Rankings for USEMO 2019 . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5.4 Histogram for USEMO 2019 . . . . . . . . . . . . . . . . . . . . . . . . . . 26
5.5 Full stats for USEMO 2019 . . . . . . . . . . . . . . . . . . . . . . . . . . 26

2
1 Summary
The first USEMO was held on May 23, 2020 and May 24, 2020. We had a total of
approximately 239 contestants who started the contest.
We held true to our word in producing what we think was a genuinely IMO-level
competition, and in fact we overshot the mark on the first day, with no 7’s awarded on
Problem 2 and no points awarded on Problem 3. The second day seems closer to true
IMO difficulty.
We hope no participants are discouraged by not making much progress on the problems.
Given the difficulty of the competition, solving any single problem is a fine achievement.
I made the deliberate decision to not “water down” the exam, despite the fact that it
is open to everyone, because I believe the students willing to dedicate nine hours of a
weekend to an exam like this will have the courage and determination to overcome initial
failures and great challenges.
The grading was strenuous, but the team of graders eventually completed the task
with expertise and grace. But that said, I think until we have a larger base of repeat
volunteers, we are unlikely to offer a multi-division contest next time, despite there being
many such requests. I remain hopeful in a few years that our presence may grow enough
to produce enough volunteers to do such an expansion.
Looking forward, I am hoping that we may run the next USEMO in the fall of 2020.

§1.1 Thanks
I am indebted to many individuals for the formulation of this contest.
I would like to thank the Art of Problem Solving for offering the software and platform
for us to run the competition. Special thanks to Amanda Reilly who led the development
of the software. Other names I encountered from the office include Jeremy Copeland,
Andres Lebbos, Eric Olson, Shannon Rogers, Richard Rusczyk, and Deven Ware. Surely
there are others who worked behind the scenes who I did not even get to see, and I am
thankful for their time as well.
I’d like to extend a thanks to Andrew Gu, Ankan Bhattacharya, Brice Huang, Carl
Schildkraut, David Altizio, Evan Chen, Jeffery Li, Michael Diao, Nikolai Beluhov, Robin
Son, Tristan Shin, Varun K-pati, Yannick Yao for suggesting at least one problem for the
competition, even if the problem was ultimately not selected.
The review of problems was carried out by Alex Rudenko, Anant Mudgal, Andrew Gu,
Ankan Bhattacharya, Brice Huang, Carl Schildkraut, David Altizio, Evan Chen, James
Lin, Michael Ren, Mihir Singhal, Milan Haiman, Nikolai Beluhov, Tristan Shin, Vincent
Huang, Yang Liu, and Zack Chroman. Many thanks for your time in helping select the
exam.
Last but certainly not least, I would also like to thank everyone who offered to help
grade the USEMO (even if real life got in the way — as it does in these challenging
COVID-19 times — and made it impossible for you to follow through on the offer).
These are Abrar Fiaz, Alex Rudenko, Anant Mudgal, Anders Olsen, Andrew Gu, Ankan
Bhattacharya, Arman Raayatsanati, Aron Thomas, Ashwin Sah, Bobby Shen, Brandon
Wang, Brian Chen, Brice Huang, Carl Schildkraut, Cathy Ye, Colin Tang, Danielle
Wang, Dominick Joo, Eric Zhang, Evan Chen, Farrell Wu, Foyez Alauddin, Hadyn Tang,
Henry Weng, James Lin, Jeck Lim, Jeffery Li, Jennifer Wang, Jit Wu Yap, Kevin Sun,

3
May 2020 The 1st US Ersatz Math Olympiad

Mehmet Kaysi, Michael Ren, Mihir Singhal, Milan Haiman, Nikolai Beluhov, Orlin
Kuchumbov, Rohan Goyal, Tahmid Hameem Chowdhury Zarif, Tom Luo, Tristan Shin,
Valentio Iverson, Victor Wang, Vincent Huang, Yang Liu, Yannick Yao, Yundi Duan,
Zack Chroman, and Zhou Li.

4
2 Results
If you won one of the seven awards, please reach out to usemo@evanchen.cc to claim
your prize!

§2.1 Top Scores

Congratulations to the top three scorers, who win the right to propose problems to future
contests.

1st place Jeffrey Kwan (33 points)

2nd place Jaedon Whyte (30 points)

3rd place Luke Robitaille (29 points)

§2.2 Special awards

See the Rules for a description of how these are awarded. Ties are broken by elegance of
solution (obviously subjective); when this occurs, runner-ups are noted below as well.

Top female Ali Cy (20 points)

Youth prize Ethan Liu (22 points)

Top day 1 Ankit Bisain (runner-up: Gopal Goel)

Top day 2 Noah Walsh

§2.3 Honorable mentions

This year we award Honorable Mention to anyone scoring at least 22 points. The HM’s
are listed below in alphabetical order.

Daniel Hong

Eddie Chen

Ethan Liu

Grant Yu

Noah Walsh

Sean Jinxiang Li

5
May 2020 The 1st US Ersatz Math Olympiad

§2.4 Distinction
We award Distinction to anyone scoring at least 14 points (two fully solved problems).
The Distinction awards are listed below in alphabetical order.

Akash Das

Alex Xu

Ali Cy

Andrew Gu

Andrew Wen

Andrew Yuan

Ankit Bisain

Bradley Guo

Brandon Chen

Brian Liu

Charley Cheng

Daniel Xu

Derek Liu

Easton Singer

Edward Yu

Espen Slettnes

Ethan Zhou

Gopal K. Goel

Grace Wang

Holden Mui

Isaac Zhu

Jason Cheah

Jeffrey Chen

Jeffrey Liu

Jeffrey Lu

Justin Yu

Karthik Seetharaman

Karthik Vedula

6
May 2020 The 1st US Ersatz Math Olympiad

Kevin Wu

Kristie Sue

Luke Choi

Mason Fang

Maximus Lu

Maxwell Sun

Nicholas Song

Nilay Mishra

Niyanth Rao

Paul Hamrick

Rafael

Reagan Choi

Rich Wang

Rishabh Das

Ryan Li

Samuel Wang

Sanjana Das

Serena An

Shreyas Ramamurthy

Srinath Mahankali

Sumith Nalabolu

Vittal Thirumalai

William Wang

William Yue

Yunseo Choi

7
3 Solutions to the Problems
§3.1 Solution to USEMO1, by Robin Son
Let ABCD be a cyclic quadrilateral. A circle centered at O passes through B and D and meets
lines BA and BC again at points E and F (distinct from A, B, C). Let H denote the orthocenter
of triangle DEF . Prove that if lines AC, DO, EF are concurrent, then triangles ABC and EHF
are similar.

Define G as the intersection of AC and EF .

Claim — Quadrilateral DCGF is cyclic.

First proof. Because ]DCG = ]DCA = ]DBA = ]DBE = ]DF E = ]DF G

Second proof. Follows since D is Miquel point of GABF .

O
G

D B

H
F

Claim — If G lies on line DO, we have AC ⊥ BD.

Proof. We have

]BDG = ]BDO = 90◦ − ]DEB = 90◦ − ]DF B

= 90◦ − ]DF C = 90◦ − ]DGC.

To finish,

]HEF = 90◦ − ]EF D = 90◦ − ]EBD = 90◦ − ]ABD = ]CAB.

Similarly ]HF E = ]ACB and the proof is done.

8
May 2020 The 1st US Ersatz Math Olympiad

Remark. The original version of this problem was in the converse direction: showing that
AC ⊥ BD implied the concurrence. Unfortunately, this turns out to be susceptible to
Cartesian coordinates by setting the x and y axes along these lines, as well as complex
methods.
Interestingly, it does not appear to be easy to show directly that the converse of the
problem implies the original statement (other than actually solving the problem, and
adapting the proof). Note in particular that the case where E = A and F = C is a
counterexample to the converse direction as stated.

9
May 2020 The 1st US Ersatz Math Olympiad

§3.2 Solution to USEMO2, by Carl Schildkraut

Let Z[x] denote the set of single-variable polynomials in x with integer coefficients. Find all
functions θ : Z[x] → Z[x] (i.e. functions taking polynomials to polynomials) such that
• for any polynomials p, q ∈ Z[x], θ(p + q) = θ(p) + θ(q);
• for any polynomial p ∈ Z[x], p has an integer root if and only if θ(p) does.

The answer is that

θ(x) = r(x) · p(±x + c)
for any choice of c ∈ Z, r(x) without an integer root, with the choice of sign fixed. For
the converse direction we present two approaches.

First solution It’s clear that this works, so we prove it is the only one. Let r(x) = θ(1),
which has no integer root since the constant 1 has no roots at all.

Part 1. We fix a positive integer n and start by determining θ(xn ) which is the bulk
of the problem. Let f (x) = θ(xn ). We look at

θ(axn + b) = a · f (x) + b · r(x).

Let g(x) = f (x)/r(x), a quotient of two polynomials whose denominator never vanishes.
By using the problem condition in both directions, varying x ∈ Z and −b/a ∈ Q, we find
that
f (x)
takes on exactly the values . . . , (−2)n , (−1)n , 0n , 1n , 2n , 3n , . . . for x ∈ Z
r(x)
So let g(x) = f (x)/r(x) now.

Claim (Rational functions can’t be integer-valued forever) — Since g maps integers

to integers, it must actually be a polynomial with rational coefficients.

Proof. We will only need the condition that g maps integers to integers.
If not, then by the division algorithm, we have g(x) = d(x) + ff12 (x)
(x) for some polynomials
d(x), f1 (x), f2 (x) in Q[x] with deg f2 > deg f1 ≥ 0. There exists an integer D such that
D · d(x) ∈ Z[x] (say the lcm of the denominators of the coefficients of g).
But for large enough integers x the value of ff12 (x)
(x) is a nonzero and has absolute value
1
less than D . This is a contradiction.

Remark. You can’t drop the condition that g has rational (rather than integer) coefficients
in the proof above; consider g(x) = 21 x(x + 1) for example.
A common wrong approach is to try to use the same logic on θ(xn )/θ(xn−1 ) for n ≥ 2.
This doesn’t work since θ(xn ) and θ(xn−1 ) could have a common root for n ≥ 2 and therefore
the problem condition essentially says nothing.

Let C be an integer divisible by every denominator in the coefficients of g. Then

apparently
h(x) = C n · g(x)
is a polynomial which only takes only nth powers as x ∈ Z.

10
May 2020 The 1st US Ersatz Math Olympiad

Claim (Polya and Szego) — Since h is a polynomial with integer coefficients whose
only values are nth powers, it must itself be the nth power of a polynomial.

Proof. This is a classical folklore problem, but we prove it for completeness.

Decompose h into irreducible factors as
h(x) = c · f0 (x)e0 · f1 (x)e1 · f2 (x)e2 · f3 (x)e3 · · · · · fm (x)em
where the fi are nonconstant and c is an integer, and ei > 0 for all i > 0. We also assume
m > 0.
We use the following facts:
• In general, if A(x), B(x) ∈ Z[x] are coprime, then gcd(A, B) is bounded by some
constant CA,B . This follows by Bezout lemma.
• If A(x) ∈ Z[x] is a nonconstant polynomial, then there are infinitely many primes
dividing some element in the range of A. This is called Schur’s theorem.
• Let A(x) ∈ Z[x] be an irreducible polynomial, and let A0 (x) be its derivative.
Then if p is prime and p > CA,A0 , and p has root in Fp , then there exists x with
νp (A(x)) = 1. This follows by Hensel lemma.
Now for the main proof. By the above facts and the Chinese remainder theorem
(together with Dirichlet theorem), we can select enormous primes p1 < p2 < · · · < pm < q
(exceeding c, e, max ei , max Cfi ,x , max Cfi ,fj for all i and j) and a single integer N
satisfying the following constraints:
• νpi (fi (N )) = 1 for all i = 1, . . . , m, by requiring N ≡ ti (mod p2i ) for suitable
constant ti not divisible by pi (because of Hensel lemma);
• pi - fj (N ) whenever i 6= j; this follows by the fact that pi > Cfi ,fj ;
Now look at the value of f (N ). It has
νp1 (f (N )) = e1
νp2 (f (N )) = e2
..
.
νpm (f (N )) = em .
Now f (N ) is a nth power so n divides all of e1 , . . . , em . Finally c must be an nth power
too.
So h(x) is an nth power; thus so is g(x). Let’s write g(x) = p(x)n then; so we find that
the range of p(x) contains either k or −k, for every integer k. For density reasons, this
forces p to be linear, and actually of the form p(x) = ±x + c for some constant c.

Part 2. We have now shown θ(xn ) = (±x + c)n r(x), for every n, for some sign and
choice of c depending possibly on n. It remains to show that the choices of signs and
constants are compatible across the different values of n. So let’s verify this.
By applying a suitable transformation on x let’s assume θ(x) = x for simplicity. Then
look at θ(xn + ax) = (±x + c)n + ax for choices of integers a. This is apparently supposed
6 0, this means x1 (±x + c)n can take any
to have a root for each choice of a, but if c =
integer value, which is obviously not true for density reasons. This means c = 0, so it
shows θ(xn ) = ±xn for any integer n.
Finally, by considering θ(xn + x − 2) = ±xn − x + 2, we see the sign must be + for the
RHS to have an integer root. This finishes the proof.

11
May 2020 The 1st US Ersatz Math Olympiad

Second solution, outline (by contestants) The solution is like the previous one, but
replaces the high-powered Polya and Szego with the following simpler result.

Claim (Odd-degree polynomials are determined by their range) — Let P (x) ∈ Z[x] be
an odd-degree polynomial. Let Q(x) be another polynomial with the same range as
P over Z. Then P (x) = Q(±x + c) for some ± and c.

Proof. First, Q also has odd degree since it must be unbounded in both directions. By
negating if needed, assume Q has positive leading coefficient.
Take a sufficiently large integer n0 such that P (x) and Q(x) are both strictly increasing
for x ≥ n0 , and moreover P (n0 ) > maxx<n0 P (x), Q(n0 ) > maxx<n0 P (x). Then take an
even larger integer n1 > n0 such that min(P (n1 ), Q(n1 )) > max(P (n0 ), Q(n0 )). Choose
n2 > n0 such that P (n1 ) = Q(n2 ). We find that this implies

P (n1 ) = Q(n2 )
P (n1 + 1) = Q(n2 + 1)
P (n1 + 2) = Q(n2 + 2)
P (n1 + 3) = Q(n2 + 3)

and so on. So P is a shift of Q as needed.

This is enough to force θ(xn ) = (±x + c)n r(x) when n is odd. When n is even, for
each integer k one can consider

θ(kxn+3 + xn ) = kθ(xn+3 ) + θ(xn )

and use the claim on θ(xn+3 ) and θ(kxn+3 + xn ) to pin down θ(xn ).

Third solution (from author) The answers are as before and we prove only the converse
direction.

Lemma
Given two polynomials P, Q ∈ Z[x], if P + nQ has an
integer root for all n, then
either P and Q share an integer root or P (x) = x+m
k Q(x) for some integers m, k
with k 6= 0.

Proof. Let d = gcd(P (0), Q(0)) so P (0) = dr and Q(0) = ds. Now, for an integer root
kn of P + nQ,
kn |P (0) + nQ(0) = dr + nds = d(r + ns).
Let p be a prime ≡ r mod s, of which there are infinitely many by Dirichlet’s theorem.
Now, for n = p−r
s , we have
kn |dp.
As the divisors of dp are exactly those of d times 1 or p, there exists a (not necessarily
positive) divisor j of d and a t ∈ {1, p} so that kn = dt for infinitely many n. In the
first case, we have that P (j) + nQ(j) = 0 for infinitely many n and some fixed j, which
implies that j is a root of both P and Q. In the second case, we have, noting p = r + ns,
that
P (j(r + ns)) + nQ(j(r + ns)) = 0.

12
May 2020 The 1st US Ersatz Math Olympiad

As this holds for infinitely many n, we may rewrite it as a polynomial equation

P (x) = (ax + b)Q(x)

for some rational a, b. Now, we know that (ax + b + n)Q(x) has a rational root for all
n ∈ Z. If Q has an integer root then P does as well and we are in our first case; otherwise,
n+b
a ∈ Z for all n ∈ Z. This implies that 1/a ∈ Z, let it be k. Then b/a ∈ Z; let it be m.
This finishes the proof.

Now, let Pn (x) = f (xn ). We claim that P1 (x) = (±x + t)P0 (x) for some t ∈ Z. Indeed,
P1 + nP0 has
an integer root for all n, so either P1 and P0 share an integer root or
P1 (x) = x+m
k P0 (x) for some m, k ∈ Z. They clearly cannot share a root, since P0 (x)
cannot have any integer roots. Now,

kP1 (x) + P0 (x) = (x + m + k)P0 (x)

has an integer root, so kx + 1 must as well, and thus k = ±1, as desired. Now, we see
that
θ (a(xn − cn ) + b(x − c)) = a Pn (x) − cn P0 (x) + b P1 (x) − cP0 (x)

has an integer root for any c, a, b. Let Q = Pn − cn P0 and R = P1 − cP0 . Since aQ + bR

has an integer root for all a, b ∈ Z, we can apply our lemma on both the pair (Q, R) and
(R, Q); if they do not share an integer root, then Q must be a linear polynomial times R
and R must be a linear times Q, a contradiction unless they are both 0 (in which case
they share any integer root). So, Q and R share an integer root. We have

R(x) = P1 (x) − cP0 (x) = (±x + t − c)P0 (x),

and P0 has no integer root as 1 has no integer root, so we have that ±(c − t) is the only
integer root of R and is thus also a root of Q; in particular

Pn (±(c − t)) = cn P0 (±(c − t))

for all c ∈ Z. This is a polynomial equation that holds for infinitely many c so we must
have that

Pn (±(x − t)) = xn P0 (±(x − t)) =⇒ Pn (x) = (±x + t)n P0 (x).

Thus, if Q(x) = di=0 ai xi ,

P

d d
!
X X
i
θ(Q(x)) = θ ai x = ai (±x + t)i P0 (x) = P0 (x)Q(±x + t),
i=0 i=0

finishing the proof.

13
May 2020 The 1st US Ersatz Math Olympiad

§3.3 Solution to USEMO3, by Nikolai Beluhov

Consider an infinite grid G of unit square cells. A chessboard polygon is a simple polygon (i.e. not
self-intersecting) whose sides lie along the gridlines of G.
Nikolai chooses a chessboard polygon F and challenges you to paint some cells of G green, such
that any chessboard polygon congruent to F has at least 1 green cell but at most 2020 green
cells. Can Nikolai choose F to make your job impossible?

The answer is YES, the task can be made impossible.

The solution is split into three parts. First, we describe a “polygon with holes” F . In
the second part we prove that this F works. Finally, we show how to take care of the
holes to obtain a true polygon.
Part 1. Construction. Choose large integers m ≥ 5, n ≥ 1 with m odd. We will let

s = mn

throughout the solution.

Let F0 be a square of side length s. Starting from F0 , iterate the following procedure.
Divide Fi into squares of side s/mi and poke a square hole of side 3s/mi+1 (a phase-i
hole) in the center of each such square to obtain Fi+1 . Finally, let F = Fn .
The output is shown below for m = 11 and n = 3. The claim is that for a suitable
choice of (m, n) this will serve as the desired example.

Part 2. Proof of this example. Suppose we have a green coloring as described.

For every green cell, the standard copy of F is a copy of F centered at the green cell.

14
May 2020 The 1st US Ersatz Math Olympiad

Claim — The standard copies of F completely cover the plane.

Proof. This is equivalent to every copy of F having at least one green cell, owing to
symmetry of F .

Claim — In any square C with side length 5s, there are at least n + 1 standard
copies of F .

Proof. Let C0 be a square of side 3s cocentric with C. Starting from C0 , iterate the
following procedure for 1 ≤ i ≤ n + 1:

• Let Si−1 be one of the standard copies of F that cover the center of Ci−1 ,

• If i 6= n + 1, let Ci be a phase-i hole in Si−1 that lies in the interior of Ci−1 .

Since C0 and the holes C1 , C2 , . . . , Cn are nested, the standard copies S0 , S1 , S2 , . . . ,

Sn of F are distinct. It follows that C contains at least n + 1 standard copies of F .

n
However, the area of F is s2 1 − m92 . Therefore, at least one cell c within C is
covered by at least
9 n
 
n+1
k= 1− 2
25 m
standard copies of F . The copy of F centered at c then contains at least k green cells.
When n = 25 · 2020 and m is sufficiently large, however, we get k > 2020.

Part 3. Handling the holes. We are left to show how to repair the above construc-
tion so that F becomes a true polygon.
Let D be any sufficiently large positive integer. Consider a homothetic copy FD of F
scaled by a factor of D. Cut several canals of unit width into FD so that FD continues to
be connected and every hole in FD is joined by a canal to the boundary of FD . (Canals
do not need to be straight; they may go around holes.) When FD is repaired in this way,
it becomes a true polygon FD0 .
Since the total area of all canals is proportional to D and the total area of FD is
proportional to D2 , when D becomes arbitrarily large the ratio of the area of FD0 to the
area of FD becomes arbitrarily close to one. Therefore, for all sufficiently large D our
proof that F is in fact a counterexample goes through for FD0 as well, with straightforward
adjustments.
The solution is complete.

Remark (Author comments). Some time after I came up with this problem, Ilya Bogdanov
pointed out to me that it is similar to problem 3.6 in Ilya Bogdanov and Grigory Chelnokov,
Pokritiya Kletchatimi Figurkami, Summer Conference of the Tournament of Towns, 2007,
https://www.turgor.ru/lktg/2007/3/index.php. The USEMO directors agreed that the
two problems are different enough that mine was suitable for the contest.

15
May 2020 The 1st US Ersatz Math Olympiad

§3.4 Solution to USEMO4, by Robin Son

Prove that for any prime p, there exists a positive integer n such that
1n + 2n−1 + 3n−2 + · · · + n1 ≡ 2020 (mod p).

The idea is to pick n = c · p · (p − 1) for suitable integer c. In what follows, everything

is written modulo p.

Claim — When n = c · p · (p − 1), the left-hand side is equal to

p−2 X
X p−1
ba = c · 10 + 20 + · · · + (p − 1)0

c·
a=0 b=1 + 11 + 21 + · · · + (p − 1)1
+ 12 + 22 + · · · + (p − 1)2
+ ...
+ 1p−2 + 2p−2 + · · · + (p − 1)p−2 .


Proof. In the original sum, we discard all the terms divisible by p, reduce all the bases
modulo p, and reduce all the exponents modulo p − 1 (by Fermat’s little theorem). Then
each block of p(p − 1) terms equals
10 + 2p−2 + 3p−3 + · · · + (p − 1)1
+ 1p−2 + 2p−3 + 3p−4 + · · · + (p − 1)0
+ 1p−3 + 2p−4 + 3p−5 + · · · + (p − 1)p−2
+ ...
+ 11 + 20 + 3p−2 + · · · + (p − 1)2
which rearranges to the desired sum.

Claim — We have
p−2 X
X p−1
ba ≡ −1 (mod p).
a=0 b=1

First proof. By the geometric series formula

p−2
X bp−1 − 1
ba = =0 ∀ b = 2, 3, . . . p − 1.
b−1
a=0

The terms with b = 1 contribute 10 + 11 + · · · + 1p−2 = p − 1 and done.

Second proof. In fact, it’s a classical lemma (proved in the same way, using primitive
roots) that
p−1
(
X
a −1 p − 1 | a
b ≡ (mod p)
b=1
0 p − 1 - a
so this is immediate.
Thus we simply need to select c ≡ −2020 (mod p) and win (and c > 0).

16
May 2020 The 1st US Ersatz Math Olympiad

§3.5 Solution to USEMO5, by Ankan Bhattacharya

Let P be a regular polygon, and let V be the set of its vertices. Each point in V is colored red,
white, or blue. A subset of V is patriotic if it contains an equal number of points of each color,
and a side of P is dazzling if its endpoints are of different colors.
Suppose that V is patriotic and the number of dazzling edges of P is even. Prove that there
exists a line, not passing through any point of V, dividing V into two nonempty patriotic subsets.

We prove the contrapositive: if there is no way to split V into two patriotic sets, then
the number of dazzling
√ edges is odd.
Let ζ = − 21 + 23 i be a root of unity. Read the n vertices of the polygon in order
starting from any point. In the complex plane, start from the origin and, corresponding
to red, white, or blue, move by 1, ζ, or ζ 2 , respectively, to get a path. The diagram below
shows an example (where black stands in for white, for legibility reasons).

Note that:

• The path we get is actually a closed loop, since V was assumed to be patriotic.

• Because there is no nontrivial patriotic subset, this closed loop does not intersect
itself, so it corresponds to some polygon Q.

We have to show the number m of vertices of Q (corresponding to dazzling edges) is

odd. Let x and y denote the number of 60◦ and 300◦ angles, so 60x+300y = 180(x+y −2).
This gives x − y = 3 so x + y is odd.

17
May 2020 The 1st US Ersatz Math Olympiad

§3.6 Solution to USEMO6, by Ankan Bhattacharya

Let ABC be an acute scalene triangle with circumcenter O and altitudes AD, BE, CF . Let X,
Y , Z be the midpoints of AD, BE, CF . Lines AD and Y Z intersect at P , lines BE and ZX
intersect at Q, and lines CF and XY intersect at R.
Suppose that lines Y Z and BC intersect at A0 , and lines QR and EF intersect at D0 . Prove
that the perpendiculars from A, B, C, O to the lines QR, RP , P Q, A0 D0 , respectively, are
concurrent.

We present two solutions.

Radical axis approach (author’s solution) The main idea is to show that (DEF ) and
(XY Z) has radical axis A0 D0 .
Let H be the orthocenter of 4ABC. We’ll let (AH), (BH), (CH) denote the circles
with diameters AH, BH, CH.

X
D0
P Q
O
Y
N
F
H
R
Z

B D M A0 C

Claim — Points H, D, Y , Z are cyclic.

Proof. Let M be the midpoint of BC. We claim they lie on a circle with HM .
Clearly ∠HDM = 90◦ . The segment Y M is the B-midline of 4BEC, so Y M k EC ⊥
HY : thus ∠HY M = 90◦ . Similarly ∠HZM = 90◦ .

18
May 2020 The 1st US Ersatz Math Olympiad

Claim — The point P is the radical center of (HB), (HC), (XY Z), (HY ZD).
Also, QR is the radical axis of (HA) and (XY Z).

Proof. First part since P H · P D = P Y · P Z; second part by symmetric claims.

We are now ready for the key claim.

Claim (Key claim) — The points A0 and D0 lie on the radical axis of (DEF ) and
(XY Z).

Proof. The radical center of (DEF ), (XY Z), (HY ZD) is A0 = Y Z ∩BC, and the radical
center of (DEF ), (XY Z), (HA) is D0 = EF ∩ QR, so we’re done.

Let S be the center of (XY Z) and T the reflection of H over S. Let N denote the
nine-point center.

Claim (Concurrence) — The point T is the concurrency point in the problem.

Proof. The line through the centers of (HA) and (XY Z) is perpendicular to the radical
axis QR. Now, a homothety with center H and scale 2 sends these centers to A and T ,
so AT ⊥ QR. Similarly, BT ⊥ RP and CT ⊥ P Q.
Similarly from N S ⊥ A0 D0 , a dilation at H by a factor of 2 shows OT ⊥ A0 D0 , as
desired.

Remark (Author comments on problem creation). The main goal was to create a problem
to showcase the midpoints of the altitudes: while they arise due to the midpoint of altitude
lemma (Lemma 4.14 in EGMO), I have rarely seen them studied in their own right. This
problem strives to be a synthesis of properties relating to the midpoints of altitudes.

Remark. An original, more long-winded version of the problem asks to show that if B 0 , C 0 ,
E 0 , F 0 are defined similarly, then all six points are collinear and perpendicular to OT . The
second approach below proves this.

Orthology approach (from contestants) Define B 0 , C 0 , E 0 , F 0 in an analogous fashion,

Claim — Points A0 , B 0 , C 0 , D0 , E 0 , F 0 are collinear.

Proof. Three applications of Desargue:

• ABC and XY Z are perspective at H so A0 , B 0 , C 0 are collinear.

• DEF and P QR are perspective at H so D0 , E 0 , F 0 are collinear.

• C 0 F R and B 0 EQ are perspective through A-altitude so B 0 C 0 , EF , QR are concur-

rent (at D0 ).

19
May 2020 The 1st US Ersatz Math Olympiad

Claim — The perpendiculars from A, B, C to QR, RP , P Q are concurrent.

Proof. This follows from the fact that 4ABC and 4P QR are orthologic with one
orthology center at O.

Claim — The perpendiculars from A, O, C to QR, D0 F 0 , P Q are concurrent.

Proof. This follows from the fact that 4D0 F 0 Q and 4AOC are orthologic with one
orthology center at E (note that AO ⊥ ED0 F ).

Remark. This solution does not even use the fact that X, Y , Z were the midpoints of the
altitudes!

20
4 Marking schemes
§4.1 Marking scheme for problem 1
Most solutions are worth 0 or 7. The following partial items are available but not additive:
• 5 points for proving if AC, DO, EF are concurrent implies AC ⊥ BD but not
finishing.
• 4 points for solutions for which spiral similarity justification is entirely absent, but
which would be complete if these details were supplied correctly.
• 1 point for proving that AC ⊥ BD is equivalent to the problem.
There is no deduction for configuration issues (Such as not using directed angles) or small
typos in angle chasing.
No points awarded for noting ∠ABC = ∠EHF , or proving/noting that D is the
Miquel point of AEF C but not making further progress on the problem.
Computational approaches which are not completed are judged by any geometric
content and do not earn other marks.

§4.2 Marking scheme for problem 2

For solutions which are not complete, the following items are available but not additive:
• 0 points for the correct answer.
• 1 point for proving θ(1) divides θ(P ) over Q[x]
• 1 point for showing θ(x), θ(x2 ), . . . all have a common integer root, or that every
pair does.
• 1 point for showing θ(x)/θ(1) is a linear polynomial with rational coefficients.
• 2 points are awarded for a solution that starts to make progress on θ(xn ) for n ≥ 2,
by proving some main lemma or claim. Showing that θ(x) is a linear multiple of
θ(1) does not earn this point.
θ(x ) n
(A common wrong approach is to claim that the rational function θ(x n−1 ) takes
n
on every integer value; this does not work since θ(x ) and θ(x n−1 ) could have a
common root for n ≥ 2.)
For solutions which are complete with errors, the following deductions apply, and all
deductions are additive:
• −1 point for an incorrect answer. This may include forgetting the ±1, for example.
• −1 point for a minor error. This most commonly applies to students who took
the quotient of two integer polynomials and assumed the coefficients were integers
when in fact they could be rational numbers, but the solution can be easily patched
once this is pointed out.
• −2 points for a more significant error that is easily fixable.

21
May 2020 The 1st US Ersatz Math Olympiad

§4.3 Marking scheme for problem 3

• 0 points for the correct answer

• 6 points for a solution that uses a set of grid cells which do not form a polygon,
but is otherwise correct (this includes sets which are not connected)

• 7 points for a correct solution

Any partial credit is done on case-by-case basis.

§4.4 Marking scheme for problem 4

For solutions which are not complete, the following items apply but are not additive.

• 0 points for no progress.

• 0 points for considering n = cp(p − 1) and nothing else.

• 0 points for partial steps in computing this sum, e.g. using primitive roots or the
geometric series but not tying it together.

• 1 point for showing the sum when n = cp(p−1) is c times the sum when n = p(p−1),
but not evaluating the latter sum.

• 1 point for evaluating the sum when n = p(p − 1) but not finishing the problem.

For essentially complete solutions, the following deductions could apply and are additive:

• −1 point: the student shows how to solve the problem for any negative number in
place of 2020, but doesn’t realize you can wrap around.

• −1 point: the student calculates the sum at n = p(p − 1) and from this assumes
that the sum of any n consecutive terms is −1 mod p.

• −1 point: the student states with no proof that the sum of xk as x varies across
a residue system mod p is −1 if p − 1 divides k and 0 otherwise. (Mentioning
primitive roots is OK.)

• −1 point: calculation error leading to wrong final answer

Do not deduct for the following errors:

• The student forgets things like 0p−1 = 0 instead of 1, as long as the errors do not
change the final result.

• The student uses negative exponents on multiples of p, as long as the solution

would be correct if the negative powers were replaced with “corresponding” positive
powers

22
May 2020 The 1st US Ersatz Math Olympiad

§4.5 Marking scheme for problem 5

• 0 points for an incorrect solution.

• 0 points for a 1-dimensional tracker argument, to prove e.g. the problem with two
colors instead of three.

• 1 point for making a broken-line polygon (with any set of angles) and not finishing.
This includes both:
– Making a polygon using 1, ω, ω 2 vectors and not attempting the angle-sum
argument.
– Making a polygon with a different set of vectors summing to 0, such that
finishing with the angle-sum argument is hard.

• 6 points for a correct solution with a minor error.

The most common form of this will likely be the official solution with a different
set of vectors, with a mistake in the resulting algebra. The mistake must be easily
fixable.

• 7 points for a correct solution.

A number of solutions make a polygon with 1, ω, ω 2 vectors and then try to “smooth
away” 300 degree angles. For this to be graded 7− , the smoothing argument must be
very explicit. It should be able to handle extremely concave shapes, like the one below.
Otherwise, just award the 1 point for considering the broken-line polygon.

§4.6 Marking scheme for problem 6

As this problem is difficult, there are not many correct solutions, so we may manually
flag any unusual cases to review as a group. Hence, this rubric is very sparse on details,
outlining only the common 0+ or 7− cases.
The following partial items are not additive.

• 0 points for proving just the concurrence of the first three perpendiculars: this is a
known follows from the fact that ABC and P QR are orthologic

23
May 2020 The 1st US Ersatz Math Olympiad

• 0 points for H, Y , Z, D cyclic

• 1 point for proving A0 B 0 C 0 D0 E 0 F 0 are collinear; no points for just conjecturing this.

• 1 point for realizing that A0 D0 is the radical axis of the two relevant circles, even
without proof.

• 1 point for realizing that the concurrence point is the reflection of H across the
center of (XY Z), even without proof.

• 2 points for noticing D0 QF 0 and ACO are orthologic (or analogous).

24
5 Statistics
A large number of students started the contest but submitted no files. This skews the
statistics a lot, but there isn’t a real way for me to discard them without losing some
legitimate zeros as well. Thus the difficulty of the competition is somewhat exaggerated.

§5.1 Summary of scores for USEMO 2019

N 239 1st Q 0 Max 33
µ 6.36 Median 2 Top 3 29
σ 7.49 3rd Q 14 Top 12 20

§5.2 Problem statistics for USEMO 2019

P1 P2 P3 P4 P5 P6
0 132 202 239 140 217 233
1 11 20 0 5 7 2
2 0 2 0 3 0 0
3 0 1 0 0 1 0
4 2 8 0 0 0 0
5 1 3 0 2 1 1
6 0 3 0 24 0 0
7 93 0 0 65 13 3
Avg 2.82 0.38 0.00 2.59 0.44 0.12

§5.3 Rankings for USEMO 2019

Sc Num Cu Per Sc Num Cu Per Sc Num Cu Per
42 0 0 0.00% 28 0 3 1.26% 14 26 62 25.94%
41 0 0 0.00% 27 0 3 1.26% 13 7 69 28.87%
40 0 0 0.00% 26 1 4 1.67% 12 2 71 29.71%
39 0 0 0.00% 25 0 4 1.67% 11 0 71 29.71%
38 0 0 0.00% 24 1 5 2.09% 10 1 72 30.13%
37 0 0 0.00% 23 0 5 2.09% 9 3 75 31.38%
36 0 0 0.00% 22 4 9 3.77% 8 4 79 33.05%
35 0 0 0.00% 21 2 11 4.60% 7 34 113 47.28%
34 0 0 0.00% 20 2 13 5.44% 6 5 118 49.37%
33 1 1 0.42% 19 5 18 7.53% 5 0 118 49.37%
32 0 1 0.42% 18 3 21 8.79% 4 0 118 49.37%
31 0 1 0.42% 17 3 24 10.04% 3 1 119 49.79%
30 1 2 0.84% 16 0 24 10.04% 2 1 120 50.21%
29 1 3 1.26% 15 12 36 15.06% 1 8 128 53.56%
0 111 239 100.00%

25
May 2020 The 1st US Ersatz Math Olympiad

§5.4 Histogram for USEMO 2019

111

34

26

12

8
7
5 5
4 4
3 3 3
2 2 2
1 1 1 1 1 1 1 1
1 2 3 4 5 6 8 9 0 1 2 3 5 6 7 8 9 0 2 3 4 5 6 7 9 0 1 2 3 4 6 7 8 9 0 1
0 7 14 21 28 35 42

§5.5 Full stats for USEMO 2019

Rank P1 P2 P3 P4 P5 P6 Σ
1. 7 5 0 7 7 7 33
2. 7 4 0 7 7 5 30
3. 7 1 0 7 7 7 29
4. 7 4 0 7 7 1 26

26
May 2020 The 1st US Ersatz Math Olympiad

Rank P1 P2 P3 P4 P5 P6 Σ
5. 7 4 0 6 0 7 24
6. 7 1 0 7 7 - 22
6. 7 1 - 7 7 0 22
6. 7 1 - 7 7 - 22
6. 7 1 - 7 7 - 22
10. 7 0 - 7 7 - 21
10. 7 6 0 7 1 0 21
12. 7 6 0 7 0 0 20
12. 7 6 0 7 0 0 20
14. 7 0 - 5 7 0 19
14. 7 1 - 6 5 - 19
14. 7 4 - 7 1 - 19
14. 7 5 0 7 0 - 19
14. 7 5 - 7 0 - 19
19. 7 4 0 6 1 0 18
19. 7 4 - 7 0 0 18
19. 7 4 - 7 0 - 18
22. 7 1 0 6 3 0 17
22. 7 3 0 7 0 0 17
22. 7 4 - 6 0 - 17
25. 1 - - 7 7 - 15
25. 7 0 - 7 1 0 15
25. 7 0 - 7 - 1 15
25. 7 1 0 7 0 0 15
25. 7 1 0 7 0 - 15
25. 7 1 0 7 - 0 15
25. 7 1 0 7 - - 15
25. 7 1 - 7 0 0 15
25. 7 1 - 7 0 - 15
25. 7 1 - 7 - 0 15
25. 7 1 - 7 - - 15
25. 7 2 - 6 0 - 15
37. 0 - - 7 7 0 14
37. 7 0 0 6 1 - 14
37. 7 0 0 7 0 0 14
37. 7 0 0 7 0 0 14
37. 7 0 0 7 0 0 14
37. 7 0 0 7 0 0 14
37. 7 0 0 7 0 0 14
37. 7 0 0 7 0 0 14
37. 7 0 0 7 0 0 14
37. 7 0 0 7 0 - 14
37. 7 0 0 7 - 0 14
37. 7 0 0 7 - - 14
37. 7 0 0 7 - - 14
37. 7 0 - 7 0 0 14
37. 7 0 - 7 0 0 14
37. 7 0 - 7 0 0 14
37. 7 0 - 7 0 - 14

27
May 2020 The 1st US Ersatz Math Olympiad

Rank P1 P2 P3 P4 P5 P6 Σ
37. 7 0 - 7 - - 14
37. 7 0 - 7 - - 14
37. 7 0 - 7 - - 14
37. 7 0 - 7 - - 14
37. 7 1 0 6 0 0 14
37. 7 1 0 6 - 0 14
37. 7 1 - 6 - 0 14
37. 7 - - 7 - - 14
37. 7 - - 7 - - 14
63. 0 0 - 6 7 - 13
63. 7 0 0 6 0 0 13
63. 7 0 0 6 - - 13
63. 7 0 0 6 - - 13
63. 7 0 - 6 0 0 13
63. 7 - - 6 0 0 13
63. 7 - - 6 - - 13
70. 5 0 0 7 0 0 12
70. 7 0 - 5 - - 12
72. 4 0 - 6 0 - 10
73. 7 0 - 2 - 0 9
73. 7 0 - 2 - - 9
73. 7 1 - 1 - - 9
76. 1 1 - 6 0 - 8
76. 1 - - 7 - - 8
76. 7 0 0 0 1 - 8
76. 7 0 - 1 - - 8
80. 0 0 0 7 0 0 7
80. 0 0 0 7 0 - 7
80. 0 0 0 7 0 - 7
80. 0 0 - 7 - - 7
80. 0 0 - 7 - - 7
80. 1 0 - 6 0 - 7
80. 7 0 0 0 0 0 7
80. 7 0 0 0 0 0 7
80. 7 0 0 0 - 0 7
80. 7 0 0 0 - - 7
80. 7 0 0 0 - - 7
80. 7 0 0 - - 0 7
80. 7 0 0 - - - 7
80. 7 0 0 - - - 7
80. 7 0 0 - - - 7
80. 7 0 - 0 0 0 7
80. 7 0 - 0 0 - 7
80. 7 0 - 0 - 0 7
80. 7 0 - 0 - - 7
80. 7 0 - 0 - - 7
80. 7 0 - - 0 - 7
80. 7 0 - - - - 7
80. 7 0 - - - - 7

28
May 2020 The 1st US Ersatz Math Olympiad

Rank P1 P2 P3 P4 P5 P6 Σ
80. 7 - - 0 - - 7
80. 7 - - 0 - - 7
80. 7 - - - 0 0 7
80. 7 - - - - - 7
80. - 0 - 7 - - 7
80. - - - 7 0 - 7
80. - - - 7 - - 7
80. - - - 7 - - 7
80. - - - 7 - - 7
80. - - - 7 - - 7
80. - - - 7 - - 7
114. 0 0 0 6 - - 6
114. 0 - - 6 - - 6
114. 4 2 - - - - 6
114. - 0 - 6 0 - 6
114. - - - 6 - - 6
119. 1 0 - 1 1 - 3
120. - - - 2 - - 2
121. 1 0 0 0 0 0 1
121. 1 0 0 0 0 0 1
121. 1 - 0 - - - 1
121. 1 - - - - - 1
121. 1 - - - - - 1
121. 1 - - - - - 1
121. - 0 - 1 - - 1
121. - - - 1 - - 1
129. 0 0 0 0 0 0 0
129. 0 0 0 0 0 0 0
129. 0 0 0 0 0 0 0
129. 0 0 0 0 0 0 0
129. 0 0 0 0 - 0 0
129. 0 0 0 0 - - 0
129. 0 0 0 - - - 0
129. 0 0 0 - - - 0
129. 0 0 0 - - - 0
129. 0 0 - 0 0 0 0
129. 0 0 - 0 0 0 0
129. 0 0 - 0 - - 0
129. 0 0 - - - - 0
129. 0 0 - - - - 0
129. 0 - - 0 0 0 0
129. 0 - - - - - 0
129. 0 - - - - - 0
129. 0 - - - - - 0
129. - 0 0 0 0 - 0
129. - 0 - 0 - - 0
129. - 0 - - - - 0
129. - 0 - - - - 0
129. - 0 - - - - 0

29
May 2020 The 1st US Ersatz Math Olympiad

Rank P1 P2 P3 P4 P5 P6 Σ
129. - - - 0 0 - 0
129. - - - 0 - 0 0
129. - - - 0 - 0 0
129. - - - 0 - - 0
129. - - - 0 - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0

30
May 2020 The 1st US Ersatz Math Olympiad

Rank P1 P2 P3 P4 P5 P6 Σ
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0
129. - - - - - - 0

31