Remarks on English

a.k.a. Advice for writing proofs

Evan Chen《陳誼廷》
13 December 2023

Exposition, criticism, appreciation, is work for second-rate

minds.

G. H. Hardy

Contents
1 Examples of good and bad writing that we’ll refer back to 1
1.1 Introductory example: USAMO 2014 Problem 1 . . . . . . . . . . . . . . 2
1.2 Advanced example: USAMO 2014 Problem 4 . . . . . . . . . . . . . . . . 3

2 Concrete advice 5
2.1 Never write wrong math . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Emphasize the point where you cross the ocean . . . . . . . . . . . . . . . 5
2.3 “Find all. . .” and “find the minimum/maximum. . .” . . . . . . . . . . . . . 6
2.4 Leave space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.5 For the love of god please include a diagram if you needed one . . . . . . 7
2.6 Pay special attention to definitions (extra important) . . . . . . . . . . . . 7
2.7 Other things . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

3 Funny quote from Steven Witten 8

A Notes specific to mathematical competitions 9

A.1 Grading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
A.2 More examples of decent write-ups . . . . . . . . . . . . . . . . . . . . . . 9
A.3 How much detail to include . . . . . . . . . . . . . . . . . . . . . . . . . . 9
A.4 Citing lemmas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
A.5 Fake-solving problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

B Checking your work 10

B.1 How to check your solutions during the year . . . . . . . . . . . . . . . . . 10
B.2 How to check your solutions during a contest . . . . . . . . . . . . . . . . 11

§1 Examples of good and bad writing that we’ll refer back to

Following the examples-first philosophy, here are two problems with commentary that
we’ll refer back to when we provide concrete advice.

1
Evan Chen《陳誼廷》 — 13 December 2023 Remarks on English

§1.1 Introductory example: USAMO 2014 Problem 1

This example is meant to be short and more accessible.

Example 1.1 (USAMO 2014/1)

Let a, b, c, d be real numbers such that b − d ≥ 5 and all zeros x1 , x2 , x3 , and x4 of
the polynomial P (x) = x4 + ax3 + bx2 + cx + d are real. Find the smallest value the
product (x21 + 1)(x22 + 1)(x23 + 1)(x24 + 1) can take.

Here are two ways you could write the solution.1

Bad solution.
x2j + 1 = (x − i)(x + i)∀j =⇒ xj + 1 = (xj + i)(xj − i) = P (i)P (−i) so (b − d −
Q 2 Q

1)2 + (a − c)2 . ∵ xj = 1 → 16 and 42 − 1 = 5. b − d ≥ 5, so ≥ 16.

Good solution.
The answer is 16 . This can be achieved by taking x1 = x2 = x3 = x4 = 1, whence the
product is 24 = 16, and b − d = 5.
We now show the quantity is always at least 16. We prove:

Claim — We always have (x21 + 1)(x22 + 1)(x23 + 1)(x24 + 1) = (b − d − 1)2 + (a − c)2 .

√
Proof. Let i = −1. The key observation is that
4
Y 4

Y
x2j + 1 = (xj − i)(xj + i) = P (i)P (−i) = |P (i)|2 .
j=1 j=1

Since P (i) = (−1 + b − d) + (c − a)i, the claim follows.

Since b − d − 1 ≥ 4, we get the desired lower bound of 42 + 02 = 16.

Commentary.
These solutions have the same mathematical content. But notice how in the better
solution:
• The second solution makes it clear from the beginning what the answer is, and
what the equality case is. (The first solution mixes these together.)
• Moreover, the main idea (of factoring with i) is explicitly labeled, so that even if
you have never seen the problem before, you can tell at a glance what the main
idea of the solution is.
• The equations are displayed in the second solution, making them much easier to
read than in the first.
The second solution, despite being twice as “long”, is by far faster to read than the first
solution. In this case, the difference is not so bad because the problem and solution are
quite short. However, in more involved problems the “not-so-good solution” becomes the
“completely unreadable solution”.
1
Former solution worsened June 2018, with suggestions from Mitchell Lee.

2
Evan Chen《陳誼廷》 — 13 December 2023 Remarks on English

§1.2 Advanced example: USAMO 2014 Problem 4

This example is meant to show what a more involved solution could look like. (Beginners
or even intermediate students shouldn’t be discouraged if it takes a while to understand
even the “good” solution; it’s not an easy problem.)

Example 1.2 (USAMO 2014 Problem 4)

Let k be a positive integer. Two players A and B play a game on an infinite grid of
regular hexagons. Initially all the grid cells are empty. Then the players alternately
take turns with A moving first. In her move, A may choose two adjacent hexagons
in the grid which are empty and place a counter in both of them. In his move, B
may choose any counter on the board and remove it. If at any time there are k
consecutive grid cells in a line all of which contain a counter, A wins. Find the
minimum value of k for which A cannot win in a finite number of moves, or prove
that no such minimum value exists.

Bad solution.
Impose coordinates (x, y). If B always removes counters from cells (x, y) with x + y ≡ 0
(mod 3), then any line by A gets interrupted so worst case A gets up to k = 2 + 1 + 2 = 5.
For A, make an equilateral triangle on second turn and then put two more counters two
away on third turn. This makes a threat so B removes an inner counter, then A puts
this inner counter back it makes another threat each turn, so B has to keep doing this.
Eventually one of the inner counters is surrounded by a full hexagon, so A fills in the
other gap, which creates an double threat for A’s next turn so A wins, answer 6.

Good solution.
The answer is k = 6.

¶ Proof that A cannot win if k = 6. We give a strategy for B to prevent A’s victory.
Shade in every third cell, as shown in the figure below. Then A can never cover two
shaded cells simultaneously on her turn. Now suppose B always removes a counter on a
shaded cell (and otherwise does whatever he wants). Then he can prevent A from ever
getting six consecutive counters, because any six consecutive cells contain two shaded
cells.

¶ Example of a strategy for A when k = 5. We describe a winning strategy for A

explicitly. Note that after B’s first turn there is one counter, so then A may create an

3
Evan Chen《陳誼廷》 — 13 December 2023 Remarks on English

equilateral triangle, and hence after B’s second turn there are two consecutive counters.
Then, on her third turn, A places a pair of counters two spaces away on the same line.
Label the two inner cells x and y as shown below.

x y

Now it is B’s turn to move; in order to avoid losing immediately, he must remove either
x or y. Then on any subsequent turn, A can replace x or y (whichever was removed) and
add one more adjacent counter. This continues until either x or y has all its neighbors
filled (we ask A to do so in such a way that she avoids filling in the two central cells
between x and y as long as possible).
So, let’s say without loss of generality (by symmetry) that x is completely surrounded
by tokens. Again, B must choose to remove x (or A wins on her next turn). After x is
removed by B, consider the following figure.

x y

We let A play in the two marked green cells. Then, regardless of what move B plays,
one of the two choices of moves marked in red lets A win. Thus, we have described a
winning strategy when k = 5 for A.

Commentary.
The bad solution doesn’t give definitions of any of the objects it is using, such as how the
coordinates (x, y) are specified, or the meanings of any of the made-up terms like “line
by A”, “threat”, “inner counter”, “worst case”, “surrounded by a full hexagon”, “fills in
the other gap”, “double threat”. The reader is forced to guess.
Diagrams are absent, and there are no names besides the k given in the problem (which
is abused slightly as k = 2 + 1 + 2 = 5). It also never clearly states what the answer is,
and smashes everything into one paragraph.
You can read lots of solutions to this problem on AoPS at https://aops.com/
community/p3478584 if you want to see more examples of good and bad writing.

4
Evan Chen《陳誼廷》 — 13 December 2023 Remarks on English

§2 Concrete advice
Here are some tips of mine that I don’t think are stressed enough.

§2.1 Never write wrong math

This is much more of a math issue than a style issue: you can lose all of your points for
making false claims. Personally, I often stop reading a solution if it makes an egregiously
false claim: if someone claims that some fixed point is the incenter of ABC, when it’s
actually the arc midpoint, then I know the solution isn’t going to have any substantial
progress.
As a special case, don’t say something that is partially true and then say how to fix it
later. At best this will annoy the grader; at worst they may get confused and think the
solution is wrong.

§2.2 Emphasize the point where you cross the ocean

Solutions to olympiad problems often involve a few key ideas, with the rest of the solution
being checking details. You want graders to immediately see all the key ideas in the
solution: this way, they quickly have a high-level understanding of your approach.
Let me share a quote from Scott Aaronson:

Suppose your friend in Boston blindfolded you, drove you around for twenty
minutes, then took the blindfold off and claimed you were now in Beijing. Yes,
you do see Chinese signs and pagoda roofs, and no, you can’t immediately
disprove him — but based on your knowledge of both cars and geography,
isn’t it more likely you’re just in Chinatown? . . .We start in Boston, we
end up in Beijing, and at no point is anything resembling an ocean
ever crossed.

Olympiad solutions work the same way: a geometry solution might require a student to
do some angle chasing, use the incenter-excenter lemma to deduce that two triangles are
congruent, and then finish by doing a little more angle chasing. In that case, you want
to highlight the key step of proving the two triangles were congruent, so the grader sees
it immediately and can say “okay, this student is using this approach”.
Ways that you can highlight this are:

• Isolating crucial steps and claims as their own self-contained parts.2

• Using claims to say what you’re doing. Rather than doing angle chasing and writing
“blah blah blah, therefore 4MB IB M ∼ 4MC IC M ”, consider instead “We claim
4MB IB M ∼ 4MC IC M , proof”.

• Displaying important equations. For example, notice how the line

4MB IB M ∼ 4MC IC M (1)

jumps out at the reader. You can even number such claims to reference them later,
e.g. “by (1)”. This is especially useful in functional equations.

2
This is often useful for another reason: breaking the proof into individual steps. The complexity of
understanding a proof grows super-linearly in its length; therefore breaking it into smaller chunks is
often a good thing.

5
Evan Chen《陳誼廷》 — 13 December 2023 Remarks on English

• Just say it! Little hints like “the crucial claim is X” or “the main idea is Y ” are
immensely helpful. Don’t make X and Y look like another intermediate step.

In the good USAMO 2014/1 sample, we can see two examples of this. First, the critical
identity
(x21 + 1)(x22 + 1)(x23 + 1)(x24 + 1) = (b − d − 1)2 + (a − c)2
was written out as a separate independent claim, with a self-contained proof. This
divorces the “prove an identity” part of the problem from the “prove the lower bound”
part of the problem, splitting the solution into two independent halves. The reader
doesn’t need to consider b − d ≥ 5 in the first half; this compartmentalization helps make
things easier. Second, the introduction of i on its separate display, explicitly telegraphed
with “the key observation”, makes it obvious to the reader that this particular step is
the heart of the proof.

§2.3 “Find all. . .” and “find the minimum/maximum. . .”

Many problems will ask you to “find all objects satisfying some condition” (for example,
functional equations, Diophantine equations). For any solution of this form, I strongly
recommend that you structure your solution as follows:

• Start by writing “We claim the answer is . . .”.

• Then, say “We prove these satisfy the conditions”, and do so. For example, in
a functional equation with answer f (x) = x2 , you should plug this f back in and
verify the equation is satisfied. Even if this verification is trivial, you must still
explicitly include it, because it is part of the problem.

• Finally, say “Now we prove these are the only ones” and do so.

Similarly, some problems will ask you to “find the minimum/maximum value of X”.
In such situations, I strongly recommend you write your solution as follows:

• Start by writing “We claim the minimum/maximum is . . .”.

• Then, say “We prove that this is attainable”, and give the construction (or
otherwise prove existence). Even if this verification is trivial, you must still
explicitly include it, because it is part of the problem.

• Finally, say “We prove this is a lower/upper bound”, and do so.

Failing to do one of the steps mentioned above is a classic newbie mistake. Make it
abundantly clear to the grader that you know the difference between a bound and a
maximum. The good solutions to both problems on USAMO 2014 follow this structure.

§2.4 Leave space

Most people don’t leave enough space. This makes solutions hard to read.
Examples of things you can do:

• Skip a line after paragraphs. Use paragraph breaks more often than you already
do.

• If you isolate a specific lemma or claim in your proof, then it should be on its own
line, with some whitespace before and after it.

6
Evan Chen《陳誼廷》 — 13 December 2023 Remarks on English

• Any time you do casework, you should always split cases into separate paragraphs
or bullet points. Make it visually clear when each case begins and ends.

• Display important equations, rather than squeezing them into paragraphs. If you
have a long calculation, then do an aligned display3 rather than squeezing it into
a paragraph. For example, instead of writing 0 ≤ (a − b)2 = (a + b)2 − 4ab =
(10 − c)2 − 4 (25 − c(a + b)) = (10 − c)2 − 4 (25 − c(10 − c)) = c(20 − 3c), write
instead

0 ≤ (a − b)2
= (a + b)2 − 4ab
= (10 − c)2 − 4 (25 − c(a + b))
= (10 − c)2 − 4 (25 − c(10 − c))
= c(20 − 3c).

§2.5 For the love of god please include a diagram if you needed one
If you used a diagram to solve the problem, give one to the reader too. (This is not
limited to just geometry problems.)
In competition settings, you do not need to draw a new diagram for the solution.
Just turn in the one that you already drew when solving the problem, by adding the
appropriate page headers to it.
“Why can’t the reader draw their own diagram in Geogebra?” Well, yes, in theory.
But theory is not the same as practice. For example, suppose you write “let M be the
midpoint of BC” when you meant “let M be the midpoint of EF ”. Now because of that
one typo, you’re completely screwed, because you also forced the reader to draw their own
diagram, which is now different from yours. The reader has no way to fix your mistake
for you. There’s no reason to press your luck like this; just include a diagram.
If you want to understand how painful it is to read a solution with no diagram provided,
check out the AoPS thread for USAMO 2014/4 at https://aops.com/community/
p3478584 and try reading some of the text-only solutions.

§2.6 Pay special attention to definitions (extra important)

The previous illustration of how the typo “let M be the midpoint of BC” can kill you
illustrates a more general point: you have to get definitions right. For intermediate logic,
a reader can rely on truth/falsity of statements to help fill in gaps or fix typos — but
definitions cannot be intrinsically right or wrong, so the reader has to use the one you
gave them.
Thus it’s absolutely critical your definitions are solid. Whenever you define a new
object or algorithm, pay special care that your specifications are complete and correct.
Tips:

• Write in complete sentences. Do not skimp on words when you’re defining objects,
they’re too important. (If you want to be lazy, do so in the main body of your
solution, not the definitions.)

• If the definition is complicated enough, give some examples. (To be honest, I

always thought the examples were worth more than the formal definition.) Compare
for example
3
This is the align* environment, for those of you that like LATEX.

7
Evan Chen《陳誼廷》 — 13 December 2023 Remarks on English

Define a block of letters to be a maximal contiguous subsequence of

consecutive letters.
versus
Define a block of letters to be a maximal contiguous subsequence of
consecutive letters. For example, the word aabbbcaaa has four blocks,
namely aa, bbb, c, aaa, appearing in that order.

• If you are defining an object, consider whether it makes sense to name it. For
example, if you find yourself writing “the set of bad numbers” over and over,
consider naming it B instead.4 As another example, consider
Given a word A, we introduce the following notation for its m blocks:

A = A1 A2 . . . Am = a1 . . . a1 a2 . . . a2 . . . am . . . am .
| {z } | {z } | {z }
x1 x2 xm

• Never write the words “worst case” unless you really, really know what you are
doing. This term is tantamount to saying to the reader, “I’m only going to consider
this case, but I won’t explain why”, and is quite commonly accompanied by false
or circular reasoning.

The examples for USAMO 2014/4 shows a night-and-day contrast. In the bad write-up,
nothing is defined, illustrated, or named. In the good write-up, all named objects are
explicitly shown.

§2.7 Other things

Try to have nice handwriting. Leave 1-inch (or more) margins. Write your proofs forwards
even if you solved the problem backwards. If you need to cite a theorem, say clearly
how you’re doing so. Use variable names at your discretion. Strike out and cross out
unwanted parts of your solution (don’t scribble).
I’m sure someone has told you these before. If not, consider reading https://www.
aops.com/articles/how-to-write-solution.

§3 Funny quote from Steven Witten

Imagine I asked you to learn a programming language where:
• All the variable names were a single letter, and where programmers en-
joyed using foreign alphabets, glyph variation and fonts to disambiguate
their code from meaningless gibberish.
• None of the functions were documented, and instead the API docs
consisted of circular references to other pieces of similar code, often with
the same names overloaded into multiple meanings, often impossible to
Google.
• None of the sample code could be run on a typical computer; in fact,
most of it was pseudo-code lacking a definition of input and output, or
even the environment it was supposed to run.
4
One subtle side effect of this habit is that it will also remind you to check that you gave a definition of
“bad number” to begin with. If you try to name an object you never gave a full definition of, it will tend
to jump out at you.

8
Evan Chen《陳誼廷》 — 13 December 2023 Remarks on English

§A Notes specific to mathematical competitions

Up until now I’ve given my advice for how to write solutions well. But I know a lot of
you are specifically interested in olympiad grading, so here are a few quick remarks to
that end. These comments are meant for USA(J)MO in particular but should apply to
other respectable contests as well.

§A.1 Grading
Your score on an olympiad problem is a nonnegative integer at most 7. The unspoken
rubric reads something like the following:
Description
7∗ Problem was solved
6 Tiny slip (and contestant could repair)
5 Small gap or mistake, but non-central
2 Lots of genuine progress
1∗ Significant non-trivial progress
0∗ “Busy work”, special cases, lots of writing
The “default” scores are starred above. Note that, unlike high school English class or the
SAT essay, you don’t get points just because you wrote a lot!
In theory, your solutions to olympiads are graded solely based on math. In practice,
style still does play a role in some ways: the harder your solution is to understand, the
less likely the grader is to understand you, and the less likely you are to earn points you
deserve.5

§A.2 More examples of decent write-ups

I should note that on my website
https://web.evanchen.cc/problems.html
there are a very large number of solutions written by me to past problems on the USAMO,
IMO, USA TST(ST), etc. In particular, all USAMO and IMO problems since the year
2000 are present.
Not all the solutions are complete (some of them are just outlines), but I think the
majority of them are full write-ups, and these can help provide more examples of solutions
that you can compare to or model your own work after.

§A.3 How much detail to include

A common question I get is what the minimum amount of detail needed to get full marks
for a solution is. The answer is simple: enough to convince the grader you solved the
problem.
There is a myth that, sort of like your high school English or math teacher, you can
lose points for “not writing enough” or not having certain key words or leaving out details
that were obvious to everyone. This is not really how it works. USAMO graders are
interested in whether you solved the problem rather than your ability to fill pages with
ink.
Basically, you lose points if a student who did NOT solve the problem could have
written the same words as you. For example, whenever you say something like “it’s
5
In addition, poorly written solutions make the graders sad, and you wouldn’t want that, would you?

9
Evan Chen《陳誼廷》 — 13 December 2023 Remarks on English

easy to see X”, the grader has to ask whether you actually understand why X is true, or
don’t know and are just bluffing. So that’s always the criteria you should have in your
head when deciding what needs to be written out in full.

§A.4 Citing lemmas

In general it is usually okay to cite a result that is (i) named, and (ii) does not trivialize
the given problem. Anything outside this scope is a “grey area” and I don’t want to
commit to a hard set of guidelines.
However, the main thing I want to say is that if in doubt, outline a proof. You don’t
have to choose between the extremes “say absolute zero” and “prove quoted lemma in
full gory detail”. It’s better to just include a couple lines giving the overall idea of the
proof to show that you could write it out if you wanted to, but are omitting it because
the result is already known.

§A.5 Fake-solving problems

With all that said, I would say in the end, when people don’t get the points they
expect, it’s because their solution is actually wrong or incomplete, not because they
wrote it poorly. This is true something like 90% of the time, maybe more.
Some common ways to lose most or all of your points by virtue of not having solved
the problem:

• Flipping an inequality sign.

• Not understanding what the word “function” means in a functional equation.

• Making some assumption that seems intuitive, but actually requires justification
(and is the main difficulty of the problem).

• Stating key assertions with no proof (often which are equivalent to the problem).

• Making some actual logical error (for example, the so-called “pointwise trap”).

• Missing some case or possibility that the student didn’t realize existed.

• Not understanding the problem statement altogether (for example, not knowing
that “find all” problems have two parts, and only doing one direction).

Some examples of USAMO problems that are notorious for generating wrong solutions:
USAMO 2003/6, USAMO 2007/2, USAMO 2010/3, USAMO 2016/4.
I should say there is no shame in having an incorrect solution to a problem, it really
happens to everyone more often than anyone wants to admit. Just don’t delude yourself
into thinking that you lost points you deserved because the graders didn’t like your style.

§B Checking your work

§B.1 How to check your solutions during the year
When you are practicing during the year, the best way to get feedback on proofs is to
have a friend/coach who can check your work and provide suggestions. But the supply of
people willing to do this is admittedly very low, so most people are not so lucky to have
access to feedback. Almost everyone gets by instead with the following algorithm:

10
Evan Chen《陳誼廷》 — 13 December 2023 Remarks on English

Step 1. Write up your solution neatly. The more clean your write-up is, the more likely
you are to catch your own mistakes6 . As you write your solution, actively look for
ways to reorganize, consolidate, or simplify your solution (see Evan’s FAQ C-24).

Step 2. Once you have a draft of a solution written, find the problem’s thread on the
Art of Problem Solving forum. You can find almost every past contest problem via
the Contest Index at https://aops.com/community/c13.7
Compare your solution draft to others posted, particularly reputable8 users. Often,
a problem will have essentially only a few approaches, and you’ll find another user
who had more or less the same approach.9 This serves as a sanity check that what
you have does work.
If you find your solution is way shorter or simpler than everyone else, then you have
good reason to be suspicious. Look for the ocean-crossing point in other people’s
solutions. Why did they have to work so hard there, while you did not? Often,
that’s where the mistake will be.

Step 3. After any polishes or edits inspired by other’s works, post your final draft on the
forum thread too. By Cunningham’s Law, if you have a blatantly wrong solution,
someone will often point it out within a few hours.

§B.2 How to check your solutions during a contest

Of course, it is critical to eventually be able to check your own work independently
without consulting other people. The IMO does not have live feedback; by the time
someone tells you about a mistake, it is too late!
If you are a beginner it might take a while to reach this stage, but you should set this
as a goal for where you want to end up. It is easier than you might expect — as you
naturally get better at solving problems, your instincts about the correctness of proofs
will automatically develop too.
During the contest, the only advice I have is “write clearly and carefully” (which is
why developing these habits pays off later). I cannot tell you how many times I realized
only during the write-up phase that the “solution” I thought I had was actually flawed.

6
If you find that most of your solutions still have mistakes which aren’t caught at this first step, this
suggests you are either not writing well or the problems you are trying are too hard for you.
7
I do NOT recommend using the AoPS Wiki in place of the Contest Index. The solution quality in the
wiki is generally much poorer than the forum.
8
If you don’t know who is “reputable” yet, using the cleanliness of the presentation is an astonishingly good
proxy. People who typeset their LaTeX correctly, isolate claims/proofs and sections, actually include
diagrams, etc. are (not-so-coincidentally) usually the ones who know what they’re doing. As a concrete
example, if you were looking for a solution to USAMO 2014/4, post #68 by TheUltimate123 is the kind
of post you should read first: properly drawn diagram, clear headings showing the two separate parts of
the proof, proper formatting, and so on.
9
There are unfortunately some problems, like USAMO 2017/1, where so many different solutions are
possible that any two people are likely to have different approaches.

11