2018 Canadian IMO Training Grids and Related Problems Matthew Brennan

Grids and Related Problems

Matthew Brennan
brennanm@mit.edu

1 Colouring and Weights

Problems relating to grids appear frequently and can require techniques from all corners of combi-
natorics, including graph theory, induction, algorithmic ideas, pigeonhole/probabilistic method and
the extremal principle/greedy. Each of these ideas appear in more than one problem in these notes.
One technique especially useful in grid problems is introducing a colouring or, often equivalently,
assigning weights to grid squares. We begin with the classic intro problem to colourings.
Example 1. A pair of diagonally opposite corners are removed from an 8×8 grid. Can the resulting
figure be tiled with dominos?
Solution. No it cannot. Colour the squares of the 8×8 grid black and white as in a chessboard. The
two removed squares are both the same colour and thus the resulting figure has unequal numbers
of black and white squares. Each domino covers exactly one black and one white square. This
implies that any figure that can be tiled with dominos must have an equal number of black and
white squares, implying this figure cannot.

These types of problems are quite common – introduce a colouring so that when you throw
down some shape, it intersects this colouring in a predictable way. This usually gives impossibility
results i.e. that certain figures cannot be tiled with certain shapes. However, colourings are good
for more than that! In the next example, we see that they can give tight extremal bounds.
Example 2. Consider an n × n staircase, which consists of the squares on or below the main
diagonal of an n × n grid. A path is a sequence of distinct squares, every two consecutive of which
share an edge. What is the minimum number of paths that an n × n staircase can be partitioned
into?
Solution. We will show that dn/2e paths are the minimum necessary. This can be achieved using
L-shaped paths between the first and last squares of the diagonal, between the second and second
last squares, etc. Colour the squares of the staircase like a chessboard so that its diagonal is all
black. Each path in the previous decomposition contains exactly one more black square than white
square. Thus the staircase contains dn/2e more black squares than white squares. Since any path
contains at most one more black square than white square, at least dn/2e paths are necessary.

1 of 15
2018 Canadian IMO Training Grids and Related Problems Matthew Brennan

Sometimes more than two colours can be useful! In many of the problems at the end of these
notes, three colours is the right number. In the next example, we use a variable number.

Example 3. (ToT 2009) We place 2009 n × n cardboard pieces each covering exactly n2 squares
on an infinite chessboard. Prove that the number of cells of the chessboard which are covered by odd
numbers of cardboard pieces is at least n2 .

Solution. Let (0, 0), (0, 1), . . . , (n − 1, n − 1) be labels for n2 colours. Colour the square (x, y) with
colour (x (mod n), y (mod n)) and note that every n × n grid contains exactly one copy of each
colour. Each colour (a, b) is covered by exactly once by each of the 2009 cardboard pieces, implying
that there is some square of that colour covered an odd number of times. Thus there is at least
one square of each of the n2 colours covered by an odd number of cardboard pieces, proving the
result.

The next example is a classical result with applications beyond grids and tiling. There are many
alternative proofs including integrating various function or finding paths of rectangles. We give a
simple colouring proof which illustrates that colouring can be useful even when rectangles are not
grid-aligned.

Example 4. The interior of a rectangle R is partitioned into rectangles with sides parallel to R
such that each rectangle has at least one side which has an integer length. Prove that R has a side
which has an integer length.

Solution. Scale R and the rectangles inside of it up by a factor of two, place the origin (0, 0) at the
bottom left corner of R and its bottom and left sides along the x and y axes. Colour the squares
of the plane like a chessboard. We now make two observations:

• Claim 1. Any axis-parallel rectangle R0 with a side with an even-integer side length contains
an equal area of black and white.
Proof. Suppose w.l.o.g. that the horizontal side of R0 has even length 2m and lower left corner
(x, y). Note that (a, b) and (a + 2m, b) have the same colour for all (a, b). Thus R0 contains
the same amount of each colour as the rectangle R00 with lower left corner (0, y) congruent
to R0 . Reflecting R00 about its axis of symmetry x = m brings points to points of opposite
colours. This implies R00 contains equal amounts of black and white.

• Claim 2. Any axis-parallel rectangle with lower left corner (0, 0) that contains equal amounts
of black and white has an even-integer side length.

2 of 15
2018 Canadian IMO Training Grids and Related Problems Matthew Brennan

Proof. Assume for contradiction that it does not. Let its upper right corner be (x, y). Let
0 < a, b < 2 be the remainders when its horizontal and vertical side lengths are divided by 2.
Applying Claim 1 twice yields that the rectangle R0 with lower left corner (0, 0) and upper
right corner (a, b) must contain equal amounts of black and white. Suppose the lattice square
with lower left corner (0, 0) is black. Then if min(a, b) ≤ 1, then R0 clearly contains more
black than white. Otherwise, it contains (a − 1)(b − 1) + 1 black area and a + b − 2 white
area. The difference between these is (2 − a)(2 − b) > 0, which is a contradiction.

Now each rectangle in the partition of R contains equal amount of black and white by Claim 1. By
Claim 2 this implies that R has an even-integer side length after being scaled by a factor of two.
Thus R originally had an integer side length.

We now give a simple corollary of Example 4. When k is prime, this result follows from the
fact that the area k of a 1 × k must divide the area mn of the grid, which implies that k divides
one of m or n. However, when k is composite, area arguments do not suffice but the result is still
true. We also give two other proofs using a colouring and complex number weights. As is often
the case, introducing complex numbers yields an algebraically slick way of stating what is morally
equivalent to a colouring argument. Even though they often aren’t doing something you couldn’t
do with a colouring argument, they can make a problem easier to reason about.

Example 5. Suppose an m × n grid can be tiled with 1 × k dominos, then k divides one of m or n.

Proof 1. Scale both axes by a factor of 1/k. Each 1 × k domino now has a side length of 1. Thus
the scaled m × n grid must have an integer side-length by Example 4, implying the result.

Proof 2. Colour the m × n grid with k colours {0, 1, . . . , k − 1} such that the square (x, y) has the
colour x + y (mod k). Each 1 × k and k × 1 domino covers one square of each colour so it suffices to
show that any grid with m, n not divisible by k does not have an equal number of each colour. If
m or n is greater than k, we can remove k rows or k columns since any k rows or columns contain
an equal amount of each colour. Thus we may assume that 0 < m, n < k. Assume w.l.o.g. that
m ≤ n. Consider the diagonal containing the bottom right corner of the grid. All m squares on
this diagonal have the same colour. Thus a fraction of m/mn = 1/n > 1/k of the squares have the
same colour, making it impossible for the grid to contain equal amounts of each colour.

Proof 3. Let ω be the kth root of unity ω = e2πi/k . Place the weight ω i+j−2 in cell (i, j). Note that
in any 1 × k or k × 1 subgrid, there is exactly one square with each of the weights 1, ω, . . . , ω k−1 .
Thus the sum of the weights in any domino is zero, implying that if the grid can be tiled then the
sum of its weights is zero. Now note that the total sum of weights in the grid is

m−1
! n−1 
X n−1
X m−1
X X (ω m − 1)(ω n − 1)
ω i+j = ωi  ωj  =
(ω − 1)2
i=0 j=0 i=0 j=0

which implies that either ω m = 1 or ω n = 1. Thus k divides one of m or n.

One beauty of introducing weights is that they often reduce various conditions to linear equa-
tions over the complex numbers. Bijections and double counting arguments can boil down to
manipulating a linear system with weights. We will illustrate this with the next problem from IMO
2016. Many grid problems ultimately come down to a linear system of equations, sometimes over

3 of 15
2018 Canadian IMO Training Grids and Related Problems Matthew Brennan

F2 . Many slick parity arguments can be thought of as manipulating a system of equations! We will
demonstrate this later.
Example 6. (IMO 2016) Find all integers n for which each cell of n × n table can be filled with
one of the letters I, M and O in such a way that:
• in each row and each column, one third of the entries are I, one third are M and one third
are O; and
• in any diagonal, if the number of entries on the diagonal is a multiple of three, then one third
of the entries are I, one third are M and one third are O.
Solution. The answer is all n divisible by 9. These n can be achieved by stacking copies of the
following 9 × 9 table:
I I I M M M O O O
M M M O O O I I I
O O O I I I M M M
I I I M M M O O O
M M M O O O I I I
O O O I I I M M M
I I I M M M O O O
M M M O O O I I I
O O O I I I M M M
We will use weights to show that 9 divides n by simply manipulating equations. Let ω = e2πi/3 be
a third root of unity. Given such a table, write 1 for an I, ω for an M and ω 2 for an O. Given
three real numbers a, b, c, it holds that a + bω + cω 2 = 0 if and only if a = b = c. Therefore the
sum of the weights in every column, row and every third diagonal is zero. Note that since each row
consists of a multiple of three entries, we have that 3 divides n.
We have a lot of equations (in total 8n/3). Searching directly for an informative linear com-
bination of these equations could get really messy, so it seems natural to start with the simplest
linear combinations possible. A first idea is to sum them all and observe that the sum of all entries
is zero. This isn’t very helpful and doesn’t even require all of the equations to be true. So we try
the next most coarse approach, grouping every third equation together. Manipulating these groups
seems as though it may allow us to use all of the given equations. Formally, we group variables by
letting aij be the sum of the weights in all cells (i0 , j 0 ) with i ≡ i0 (mod 3) and j ≡ j 0 (mod 3) for
each 1 ≤ i, j ≤ 3. Our equations now imply that

a11 + a12 + a13 = 0 (1)

a21 + a22 + a23 = 0 (2)
a31 + a32 + a33 = 0 (3)
a11 + a21 + a31 = 0 (4)
a12 + a22 + a32 = 0 (5)
a13 + a23 + a33 = 0 (6)
a11 + a22 + a33 = 0 (7)
a31 + a22 + a13 = 0 (8)

4 of 15
2018 Canadian IMO Training Grids and Related Problems Matthew Brennan

Let’s try to isolate a variable if we can. Observe that

0 = (1) + (3) + (4) + (6) − 2 · (7) − 2 · (8) − (2) − (5) = −6a22

So a22 = 0 but this implies that the squares with indices congruent to 2 modulo 3 must contain an
equal number of I’s, M ’s and O’s. So n2 /9 is divisible by 3, implying 9 divides n.

After realizing that the key is to show that these squares contain an equal number of I’s, M ’s
and O’s, we can also show this by double counting.

2 Graph Theory and Grids

Many ideas in the previous section can be thought of graph-theoretically. A domino tiling is a
perfect matching of the squares of a grid where two squares are adjacent if they share an edge. The
first example realizes that a chessboard is bipartite with the two colours corresponding to the two
parts of the graph and that a perfect matching in a bipartite graph is possible only if the two parts
have the same size. The next example requires a different bipartite graph representation of grids
where squares correspond to edges between their row and column. Here we essentially use the fact
that any graph on n vertices with at least n edges contains a cycle.
Example 7. In an n × n grid, at least 2n squares are marked. Prove that there is a sequence
P1 , P2 , . . . , Pk of centeres of marked squares such that the segments Pi Pi+1 alternate between hori-
zontal and vertical for all 1 ≤ i ≤ k where Pk+1 = P1 .

Solution. Consider the bipartite graph G with 2n vertices u1 , u2 , . . . , un , v1 , v2 , . . . , vn such that

ui vj is an edge if and only if the square (i, j) is marked. If G does not contain a cycle, it must be
a forest and contains 2n − c edges where c ≥ 1 is the number of connected components of the G.
However this contradicts the fact that G contains at least 2n edges. Thus G contains a cycle. The
edges of this cycle correspond to cells that alternate being in the same column and being in the
same row, proving the desired result.

Depending on the specific problem, other graph representations of grids may be useful. In the
next problem, we use the fact that any undirected graph where every vertex has degree 2 is a
disjoint union of cycles. We will use the geometry of the grid to infer a structural property of its
graph representation – namely that all cycles in the graph have even length.
Example 8. (Gabriel Carroll) A company wants to build a 2001 × 2001 building with doors con-
necting pairs of adjacent rooms. Two rooms are adjacent if they share an edge. Is it possible for
every room to have exactly 2 doors?

5 of 15
2018 Canadian IMO Training Grids and Related Problems Matthew Brennan

Solution. No it is not possible. Consider the graph G with 20012 vertices, each corresponding to
room, such that two rooms are connected by an edge if there is a door between them. If every
room has exactly two doors, then every vertex in G has degree two, implying that G is a disjoint
union of cycles. Now colour the grid as in a chessboard. Adjacent rooms have different colours
and therefore any cycle alternates colours, implying it is even in length. Therefore G must have an
even number of vertices if it is a disjoint union of cycles, which is a contradiction.

Note that the above solution essentially observes that G is bipartite with the two parts given
by the colours in a chessboard colouring, as in Example 1. The fact that any finite graph in which
every vertex has degree two is a disjoint union of cycles can be proven by strong induction. Begin
at an arbitrary vertex and take an edge from it. Then take an edge from the new vertex other
than the one just taken to get there. Continue in this way until you see a vertex for the second
time. This vertex must be the vertex you started at, otherwise it would have degree at least three.
Furthermore, we have exhausted the degrees of the vertices we have seen with this cycle. Applying
the induction hypothesis to the rest of the graph yields the result.
The next example uses deeper properties of the geometry of the grid to infer something about
a related graph. It also illustrates the usefulness of considering grid vertices and edges as tools in
problems about grid squares.

Example 9. (St. Petersburg 2000) On an infinite checkerboard are placed 111 non-overlapping
corners, L-shaped figures made of 3 unit squares. Suppose that for any corner, the 2 × 2 square
containing it is entirely covered by the corners. Prove that one can remove each number between 1
and 110 of the corners so that the property will be preserved.

Solution. Consider the directed graph G with 111 vertices, each corresponding to an L-shaped
figure. Draw a directed edge from each L-shape u to the L-shape v such that a square of v occupies
the missing square of the 2 × 2 box containing u. As given, every vertex in G has out-degree exactly
one. We are looking for a subset with between 1 and 110 vertices of G such that every vertex has
out-degree exactly one on the subset.
If there are two vertices that both point to each other, they must together form a 2 × 3 or
3 × 2 box, that we can remove. If there is a vertex with in-degree zero, we can just remove that
vertex. Now note that if there is a vertex with in-degree other than one, there must be a vertex
with in-degree zero since the sum of the in-degrees equals the sum of the out-degrees. Thus we
can assume that every vertex of G has in-degree one and no two vertices point to each other. By
a directed analogue of the argument above, G must be the disjoint union of cycles, each of which
has length at least three. Now consider the centers of the 2 × 2 boxes containing any two L-shapes
u and v with u → v. If u and v do not point to each other, some casework shows that the x
coordinates of these centers have different parity. This implies that any cycle in G is even and thus
G contains an even number of vertices. This is a contradiction.

6 of 15
2018 Canadian IMO Training Grids and Related Problems Matthew Brennan

Note that the proof above actually shows the stronger result that we can remove exactly one
corner while preserving the property.

3 Some Other Techniques

There are many other ideas that appear in grid problems. Some of these include:
• Algorithms to show existence

• Extremal arguments

• Induction

• Parity arguments and double-counting

The next example finds a desired substructure algorithmically. As with many algorithmic existence
proofs, it can be written as an induction, thought of as generalizing the problem (e.g. here to
removing a single row from m × n grids) or as dynamic programming. It also illustrates the power
of making use of parity, which is the reason behind the entire result. If zeros were allowed in the
grid as well, it would be far from true.
Example 10. (David Arthur) Numbers 1 and −1 are written in the cells of a board 1500 × 1500.
It is known that the sum of all the numbers in the board is positive. Show that one can select 1000
rows and 1000 columns such that the sum of numbers written in their intersection cells is at least
1000.
Solution. First we prove the following general lemma.
Lemma 1. The numbers 1 and −1 are written in a m × n grid with sum S. If n is odd, we can
delete a row so that the remaining grid has sum at least min(S + 1, m − 1). If n is even, then we
can delete a row so that the remaining grid has sum at least min(S, m − 1)

Proof. If there is a row with sum at most 0, deleting this yields a sum of at least S. If n is odd,
then every row has odd sum. Thus if there is a row with sum at most 0, it must have sum at most
−1. Deleting this row yields a sum of at least S + 1. If there is no row with sum at most 0, then
deleting any row yields a sum of at least m − 1. This proves the lemma.

Note that the sum of the grid must initially be at least 2 since it is even. Now delete one row
to yield a 1499 × 1500 grid with sum at least 2. Delete 499 columns to yield a 1499 × 1001 grid
with sum at least 501. Delete 499 rows to yield a 1000 × 1001 grid with sum at least 1000. Now
delete one column to yield a 1000 × 1000 grid with sum at least 1000.

The next problem uses a very tricky extremal argument, the basic idea behind which is geomet-
rically intuitive. The key is to use the fact that any set of n − 1 consecutive values geometrically
separates the squares corresponding to the values above and below it in the grid. However, n − 1
squares seems like too few to separate sufficiently large sets of squares. Making this argument
rigorous ends up requiring some clever ideas.
Example 11. (ISL 1988) The numbers 1, 2, . . . , n2 are written in an n × n square grid. Prove that
there is some pair of adjacent squares with difference at least n.

7 of 15
2018 Canadian IMO Training Grids and Related Problems Matthew Brennan

Ak ck

Bk
Ck

rk

Solution. Assume for contradiction that every pair of adjacent squares has difference at most n − 1.
For each k = 1, 2, . . . , n2 − n, define
• Ak to be the set of squares with values 1, 2, . . . , k

• Bk to be the set of squares with values k + 1, . . . , k + n − 1

• Ck to be the set of squares with values k + n, . . . , n2

No square in Ak can be adjacent to a square in Ck i.e. Bk separates Ak from Ck . Since |Bk | = n−1,
there is some row rk and some column ck that does not intersect Bk . Since rk and ck intersect
in a square, it either holds that rk , ck ⊆ Ak or rk , ck ⊆ Ck for each k. Note that r1 , c1 ⊆ C1 and
rn2 −n , cn2 −n ⊆ An2 −n since |A1 |, |Cn2 −n | = 1 < n. Now let t > 1 be the first t such that rt , ct ⊆ At .
Since rt−1 , ct−1 ⊆ Ct−1 , it follows that At and Ct−1 must intersect in the two squares rt−1 ∩ ct and
rt ∩ ct−1 . However, At and Ct−1 intersect in exactly one square, which is a contradiction.

The next example is an induction with a non-standard reduction to smaller cases. In this
problem, we observe that two good objects can be composed in a natural way to yield another
good object (e.g. in this case taking the element-wise product of two successful arrangements). So
given a good object A, we face the question: how do we generate another object B to compose
with A in a useful way? Here, we generate B by applying a natural symmetry to A. One very hard
problem at the end of these notes uses a similar idea!
Example 12. (Russia 1998) Let n ≥ 2 be a positive integer. Each square of a (2n − 1) × (2n − 1)
board contains either 1 or −1. Such an arrangement is called successful if each number is the
product of its neighbors. Find the number of successful arrangements.

n−1
1
succ.

1 1 1

Solution. We prove by induction on n that the only possible arrangement contains only 1’s for
n ≥ 2. We first deal with the base case n = 2. Let aij where 1 ≤ i, j ≤ 3 be any successful arrange-
ment. First observe that a22 = a21 a23 a12 a32 = a422 a211 a213 a231 a233 = 1. Also a11 a13 = a212 a21 a23 =

8 of 15
2018 Canadian IMO Training Grids and Related Problems Matthew Brennan

(a22 a11 a31 )(a22 a13 a33 ) which implies that a31 a33 = 1. Therefore a32 = a22 a31 a33 = 1. Similarly
a12 = a32 = a23 = a21 = 1. This implies that all of the entries of aij are 1, proving the base case.
Assume for contradiction that there is some arrangement a0ij for 1 ≤ i, j ≤ 2n − 1 satisfying the
given conditions with some a0ij = −1. We will show that there is some arrangement aij such that
the 2n−1 th row and column both consist of all 1’s with some a0ij = −1. Note that the element-
wise product of any two successful arrangements is also successful. If a0ij is not symmetric about its
vertical axis of symmetry, multiply it by its reflection in this axis. Since it is not symmetric, it must
still contain a −1. Do the same for the horizontal axis of symmetry of the new arrangement. Call
the resulting arrangement aij , which is both horizontally and vertically symmetric, and contains
a −1. Let k = 2n−1 and consider the elements in the kth column of aij . It follows that a1,k =
a1,k−1 a1,k+1 a2,k = a2,k since a1,k−1 = a1,k+1 by symmetry. Similarly, for all 1 ≤ t ≤ k − 1, we have
that at+1,k = at,k at+2,k and a2n −1,k = a2n −2,k . If a1,k = a2,k = 1, then these equations imply that
at,k = 1 for all t. If a1,k = a2,k = −1, then a3,k = 1 and in general

1 if t ≡ 1, 2 (mod 3)
at,k =
−1 if t ≡ 0 (mod 3)

However, 2n − 1 6≡ 2 (mod 3) which implies that the sequence a1,k , . . . , a2n −1,k is not symmetric if
it takes this form. Therefore it must follow that at,k = 1 for all t. By the same reasoning the kth
row must also consist of all 1’s, which proves the claim.
Since aij is horizontally and vertically symmetric, any of its four (2n−1 − 1) × (2n−1 − 1) corner
subgrids contains a −1. Furthermore, since the 2n−1 th row and column consist of all 1’s, these four
subgrids must each be successful for n − 1. The induction hypothesis yields a contradiction.

If we let aij = 1 if there is a −1 in square (i, j) and aij = 0 otherwise, then every aij is equal to
the sum of its neighboring values modulo 2. In the above proof, we show that any solution to these
equations over F2 must be all zeros. In linear algebra terms, that the linear system is invertible.
However, to actually show this we need to use the geometry of the grid as in the above solution.
The next example gives several disallowed local structures which together yield a contradiction.
A common trick in these problems is to look locally for a “nice” characterization e.g. some count
takes on specific values. This often leads to a double counting argument as in the example below.
Example 13. (Russia 2017) Each cell of 100 × 100 table is coloured black or white. Every cell
sharing an edge with the boundary of the table is coloured black. Suppose that in every 2 × 2 square
there are cells of both colors. Prove that there exists a 2 × 2 square that is coloured like a 2 × 2
subgrid of a chessboard.

Solution. Assume for contradiction that there is no 2 × 2 square coloured as in a chessboard.

Call an edge between two squares good if it separates two squares of opposite colours. The key
observation is that the allowed 2 × 2 patterns – those with three squares of the same colour or two
pairs of adjacent squares of the same colour – each contain exactly two good edges. This is the
nice characterization of local structure that we need. There are 992 total 2 × 2 subgrids, each of
which contains two good edges. Since no good edges lie in the first or last row or column, we have
counted each good edge in exactly two 2 × 2 subgrids. Therefore there are exactly 992 good edges.

9 of 15
2018 Canadian IMO Training Grids and Related Problems Matthew Brennan

However, each row or column begins and ends with the colour black, implying it changes colour
at an even number of good edges. Therefore the total number of good edges is even, which is a
contradiction.

Problems
We adopt the David Arthur style of dividing the following problems into several difficulty classes
A, B, C and D. They are (hopefully) roughly ordered by difficulty.

A1. 100 queens are placed on a 100 × 100 chessboard so that no two attack each other. Prove
that each of the four 50 × 50 corners of the board contains at least one queen.

A2. Is it possible to tile 2003 × 2003 board by 1 × 2 dominoes placed horizontally and 1 × 3
rectangles placed vertically?

A3. James and Alex play a game on a n × n chessboard. At the beginning, all squares are white
apart from one black corner square containing a rook. Players take turns to move the rook
to a white square and recolour the square black. The player who can not move loses. James
goes first. Who has a winning strategy?

A4. A square has been removed from a 2n × 2n grid. Prove that the remaining figure can be tiled
with L-trominos.

A5. Some of the fields on a rectangular m × n board (m, n ≥ 2) are painted black, while the
remaining fields are white. There is a frog sitting outside the board. At a certain moment the
frog jumps onto one of the fields on the edge of the board, and then makes a sequence of jumps.
In every jump, it jumps from a field to its neighbouring field (two fields are neighbouring if
they have a side in common). Every time the frog jumps to a field, the colour of that field
changes either from black to white or vice-versa. Is there a path the frog could take so that,
when it leaves the board by jumping from a field on the edge, all the fields are black?

A6. Each edge of an m × n rectangular grid is oriented with an arrow such that (a) the border
is oriented clockwise, and (b) each interior vertex has two arrows coming out of it, and two
arrows going into it. Prove that there is at least one square whose edges are oriented clockwise.

A7. Prove that it is not possible colour the squares of a 11 × 11 grid using three colours, such that
no four squares whose centres form the vertices of a rectangle with sides parallel to the sides
of the grid, have the same colour.

A8. On a (4n + 2) × (4n + 2) square grid, a turtle can move between squares sharing a side.The
turtle begins in a corner square of the grid and enters each square exactly once, ending in
the square where she started. In terms of n, what is the largest positive integer k such that
there must be a row or column that the turtle has entered at least k distinct times?

A9. On an n×n table real numbers are put in the unit squares such that no two rows are identically
filled. Prove that one can remove a column of the table such that the new table has no two
rows identically filled.

10 of 15
2018 Canadian IMO Training Grids and Related Problems Matthew Brennan

A10. In an n × n array, each of the numbers 1, 2, . . . , n appears exactly n times. Show that there
√
is a row or a column in the array with at least n distinct numbers.

A11. There are k rooks on a 10 × 10 chessboard. We mark all the squares that at least one rook
can capture (we consider the square where the rook stands as captured by the rook). What
is the maximum value of k so that the following holds for some arrangement of k rooks: after
removing any rook from the chessboard, there is at least one marked square not captured by
any of the remaining rooks.

A12. Can a 5 × 7 checkerboard be covered by L’s, not crossing its borders, in several layers so that
each square of the board is covered by the same number of L’s?

A13. A maze is an 8 × 8 board with some adjacent squares separated by walls, so that any two
squares can be connected by a path not meeting any wall. Given a command LEFT, RIGHT,
UP, DOWN, a pawn makes a step in the corresponding direction unless it encounters a wall or
an edge of the chessboard. Eddy writes a program consisting of a finite sequence of commands
and gives it to the Calvin, who then constructs a maze and places the pawn on one of the
squares. Can Eddy write a program which guarantees the pawn will visit every square despite
the Calvin’s efforts?

B1. Consider a (2m − 1) × (2n − 1) rectangular region, where m and n are integers such that
m, n ≥ 4. The region is to be tiled using tiles of the two types shown:

The tiles may be rotated and reflected, as long as their sides are parallel to the sides of the
rectangular region. They must all fit within the region, and they must cover it completely
without overlapping. What is the minimum number of tiles required to tile the region?

B2. A number of robots are placed on the squares of a finite, rectangular grid of squares. A square
can hold any number of robots. Every edge of each square of the grid is classified as either
passable or impassable. All edges on the boundary of the grid are impassable. You can give
any of the commands up, down, left, or right. All of the robots then simultaneously try to
move in the specified direction. If the edge adjacent to a robot in that direction is passable,
the robot moves across the edge and into the next square. Otherwise, the robot remains on
its current square. You can then give another command of up, down, left, or right, then
another, for as long as you want. Suppose that for any individual robot, and any square on
the grid, there is a finite sequence of commands that will move that robot to that square.
Prove that you can also give a finite sequence of commands such that all of the robots end
up on the same square at the same time.

B3. A 6 × 6 rectangle is tiled by 2 × 1 dominoes. Prove that it has always at least one fault-line,
i.e., a line cutting the rectangle without cutting any domino.

B4. A chessboard is tiled with 32 dominoes. Each domino covers two adjacent squares, a white
and a black square. Show that the number of horizontal dominoes with the white square on

11 of 15
2018 Canadian IMO Training Grids and Related Problems Matthew Brennan

the left of the black square equals the number of horizontal dominoes with the white square
on the right of the black square.

B5. Every vertex of the unit squares on an m × n grid is coloured either blue, green, or red, such
that all the vertices on the boundary of the board are coloured red. We say that a unit square
on the board is properly coloured if exactly one pair of adjacent vertices of the square are the
same colour. Show that the number of properly coloured squares is even.

B6. On an n × n chart, where n ≥ 4, stand + signs in the cells of the main diagonal and − signs
in all the other cells. You can change all the signs in one row or in one column, from − to
+ or from + to −. Prove that you will always have n or more + signs after finitely many
operations.

B7. Construct a tetromino by attaching two 2 × 1 dominoes along their longer sides such that the
midpoint of the longer side of one domino is a corner of the other domino. This construction
yields two kinds of tetrominoes with opposite orientations. Let us call them S- and Z-
tetrominoes, respectively. Assume that a lattice polygon P can be tiled with S-tetrominoes.
Prove that no matter how we tile P using only S- and Z-tetrominoes, we always use an even
number of Z-tetrominoes.

B8. Several fleas sit on the squares of a 10 × 10 chessboard (at most one flea per square). Every
minute, all fleas simultaneously jump to adjacent squares. Each flea begins jumping in one of
four directions (up, down, left, right), and keeps jumping in this direction while it is possible;
otherwise, it reverses direction on the opposite. It happened that during one hour, no two
fleas ever occupied the same square. Find the maximal possible number of fleas on the board.

B9. A square is divided into congruent rectangles with sides of integer lengths. A rectangle is
important if it has at least one point in common with a given diagonal of the square. Prove
that this diagonal bisects the total area of the important rectangles.

B10. A rectangle is divided into 2×1 and 1×2 dominoes. In each domino, a diagonal is drawn, and
no two diagonals have common endpoints. Prove that exactly two corners of the rectangle
are endpoints of these diagonals.

B11. On the infinite chessboard several rectangular pieces are placed whose sides run along the
grid lines. Each two have no squares in common, and each consists of an odd number of
squares. Prove that these pieces can be painted in four colours such that two pieces painted
in the same colour do not share any boundary points.

B12. On an n × n board, there are n2 squares, n − 1 of which are infected. Each second, any square
that is adjacent to at least two infected squares becomes infected. Show that at least one
square always remains uninfected.

B13. On a 55 × 55 square grid, 500 unit squares were cut out as well as 400 L-shaped pieces
consisting of 3 unit squares (each piece can be oriented in any way). Prove that at least two
of the cut out pieces bordered each other before they were cut out.

B14. 2500 chess kings have to be placed on a 100 × 100 chessboard so that (i) no king can capture
any other one (i.e. no two kings are placed in two squares sharing a common vertex); (ii)

12 of 15
2018 Canadian IMO Training Grids and Related Problems Matthew Brennan

each row and each column contains exactly 25 kings. Find the number of such arrangements.
(Two arrangements differing by rotation or symmetry are supposed to be different.)

B15. In a rectangular array of nonnegative real numbers with m rows and n columns, each row and
each column contains at least one positive element. Moreover, if a row and a column intersect
at a positive element, then the sums of their elements are the same. Prove that m = n.

B16. Let n be a positive integer. Denote by Sn the set of points (x, y) with integer coordinates
such that  
 1 
|x| + y +  < n.

2
A path is a sequence of distinct points (x1 , y1 ), (x2 , y2 ), . . . , (x` , y` ) in Sn such that, for i =
2, . . . , `, the distance between (xi , yi ) and (xi−1 , yi−1 ) is 1 (in other words, the points (xi , yi )
and (xi−1 , yi−1 ) are neighbors in the lattice of points with integer coordinates). Prove that
the points in Sn cannot be partitioned into fewer than n paths (a partition of Sn into m paths
is a set P of m nonempty paths such that each point in Sn appears in exactly one of the m
paths in P).

B17. A plane is coloured into black and white squares in a chessboard pattern. Then, all the white
squares are coloured red and blue such that any two initially white squares that share a corner
are different colours. (One is red and the other is blue.) Let ` be a line not parallel to the
sides of any squares. For every line segment I that is parallel to `, we can count the difference
between the length of its red and its blue areas. Prove that for every such line ` there exists
a number C that exceeds all those differences that we can calculate.

B18. An arrowgram is a finite rectangular grid with an arrow drawn in each square such that:

• Each arrow points to an adjacent square in one of the eight compass directions (and
does not point off the edge of the grid), and
• No two arrows point to the same square.

Two arrowgrams A and B are said to be similar if they are on equally sized grids, and if for
every square, the corresponding arrows in A and B either point in the same direction or in
opposite directions. For what integers N does there exist an arrowgram that is equivalent to
exactly N other arrowgrams?

C1. A solitaire game is played on an m × n rectangular board, using mn markers which are white
on one side and black on the other. Initially, each square of the board contains a marker
with its white side up, except for one corner square, which contains a marker with its black
side up. In each move, one may take away one marker with its black side up, but must then
turn over all markers which are in squares having an edge in common with the square of the
removed marker. Determine all pairs (m, n) of positive integers such that all markers can be
removed from the board.

C2. There is an n × n grid on a computer. Each of its n2 squares displays an integer from 0 to k.
For each of the n rows and each of the n columns, there is also a button that, if pressed, will
increase every number in that row or column by 1. If a number ever reaches k, it immediately
changes to 0. Initially, every square displayed 0, but then a number of buttons were pressed.

13 of 15
2018 Canadian IMO Training Grids and Related Problems Matthew Brennan

Show that after at most kn more button presses, it is possible to change every number back
to 0 again.

C3. The 52 cards in a standard deck are placed in a 13 × 4 array. If every two adjacent cards,
vertically or horizontally, have either the same suit or the same value, prove that all 13 cards
of the same suit are in the same row.

C4. Each cell of a 1000 × 1000 table contains 0 or 1. Prove that one can either cut out 990 rows
so that at least one 1 remains in each column, or cut out 990 columns so that at least one 0
remains in each row.

C5. There are some counters in some cells of 100 × 100 board. Call a cell nice if there are an even
number of counters in adjacent cells. Can exactly one cell be nice?

C6. A white plane is partitioned onto cells (in a usual way). A finite number of cells are coloured
black. Each black cell has an even (0, 2 or 4) adjacent (by the side) white cells. Prove that
one may colour each white cell in green or red such that every black cell will have equal
number of red and green adjacent cells.

C7. An m × n rectangular grid is covered by dominoes. Prove that the vertices of the grid can
be coloured using three colours so that any two vertices a distance 1 apart are colored with
different colours if and only if their segment lies on the boundary of a domino.

C8. All points in a 100 × 100 array are colored in one of four colors red, green, blue or yellow
in such a way that there are 25 points of each color in each row and in any column. Prove
that there are two rows and two columns such that their four intersection points are all in
different colors.

C9. Given positive integers m and n ≥ m, determine the largest number of dominoes (1 × 2 or
2 × 1 rectangles) that can be placed on a rectangular board with m rows and 2n columns
consisting of cells (1 × 1 squares) so that: (i) each domino covers exactly two adjacent cells of
the board; (ii) no two dominoes overlap; (iii) no two form a 2 × 2 square; and (iv) the bottom
row of the board is completely covered by n dominoes.

C10. On an infinite chessboard, a solitaire game is played as follows: at the start, we have n2 pieces
occupying a square of side n. The only allowed move is to jump over an occupied square to
an unoccupied one, and the piece which has been jumped over is removed. For which n can
the game end with only one piece remaining on the board?

C11. In an m × n rectangular grid, where m and n are odd integers, 1 × 2 dominoes are initially
placed so as to exactly cover all but one of the 1 × 1 squares at one corner of the grid.
It is permitted to slide a domino towards the empty square, thus exposing another square.
Show that by a sequence of such moves, we can move the empty square to any corner of the
rectangle.

D1. A 2n × n matrix of 1’s and −1’s is such that its 2n rows are pairwise distinct. An arbitrary
subset of the entries of the matrix are changed to 0. Prove that there is a nonempty subset
of the rows of the altered matrix that sum to the zero vector.

14 of 15
2018 Canadian IMO Training Grids and Related Problems Matthew Brennan

D2. A frog named James hops around the squares of an n × n grid always to an adjacent square
so that he visits each square exactly once and ends where he starts. Prove that there are two
adjacent squares such that if we cut the cycle James takes at these squares, the number of
squares in the two resulting pieces are each at least n2 /4.

D3. A frog is placed on each cell of a n × n square inside an infinite chessboard (so initially there
are a total of n × n frogs). Each move consists of a frog A jumping over a frog B adjacent
to it with A landing in the next
h 2 i cell and B disappearing (adjacent means two cells sharing
a side). Prove that at least n3 moves are needed to reach a configuration where no more
moves are possible.

D4. A 2010 × 2010 board is divided into corner-shaped figures of three cells. Prove that it is
possible to mark one cell in each figure such that each row and each column will have the
same number of marked cells.

D5. Let n ≥ 3 be an odd integer. Amy has coloured the squares in an n × n grid white and black.
We will call a sequence of squares S1 , S2 , . . . , Sm a “path” if all these squares are the same
colour, if Si and Si+1 share an edge for all i ∈ {1, 2, . . . , m − 1}, and if no other squares in
the sequence share an edge. Prove that if both the white squares and black squares form a
single path, then one of these paths must begin or end at the center of the grid.

D6. Given natural numbers a and b, such that a < b < 2a. Some cells on a graph are colored
such that in every rectangle with dimensions a × b or b × a, at least one cell is colored. For
which greatest α can you say that for every natural number N you can find a square N × N
in which at least α · N 2 cells are colored?

D7. Let n be a positive integer. Determine the smallest positive integer k with the following
property: it is possible to mark k cells on a 2n × 2n board so that there exists a unique
partition of the board into 1 × 2 and 2 × 1 dominoes, none of which contain two marked cells.

15 of 15