Estonian Math Competitions

2020/2021

University of Tartu Youth Academy

Tartu 2021
 WE THANK:

Estonian Ministry of Education and Research

University of Tartu

Jaan Tallinn

Robert Kitt

Kadi Liis Saar & Alexey Morgunov

Problem authors: Artur Avameri, Maksim Ivanov, Kaarel Hänni,

Jaan Kristjan Kaasik, Urve Kangro, Oleg Koshik, Urmas Luhaäär,
Richard Luhtaru, Härmel Nestra, Markus Rene Pae, Erik Paemurru,
Ago-Erik Riet, Kaur Aare Saar, Toomas Tennisberg,
Tähvend Uustalu, Triinu Veeorg, Hendrik Vija
Translator: Härmel Nestra
Editor: Härmel Nestra

Estonian Mathematical Olympiad

http://www.math.olympiaadid.ut.ee/
 Mathematics Contests in Estonia
The Estonian Mathematical Olympiad is held annually in three rounds: at
the school, town/regional and national levels. The best students of each
round (except the final) are invited to participate in the next round. Every
year, about 110 students altogether reach the final round.
In each round of the Olympiad, separate problem sets are given to the
students of each grade. Students of grade 9 to 12 compete in all rounds,
students of grade 7 to 8 participate at school and regional levels only. Some
towns, regions and schools also organize olympiads for even younger stu-
dents. The school round usually takes place in December, the regional
round in January or February and the final round in March or April in
Tartu. The problems for every grade are usually in compliance with the
school curriculum of that grade but, in the final round, also problems re-
quiring additional knowledge may be given.
The first problem solving contest in Estonia took place in 1950. The next
one, which was held in 1954, is considered as the first Estonian Mathemati-
cal Olympiad.
Apart from the Olympiad, open contests take place in September and
in December. In addition to students of Estonian middle and secondary
schools, Estonian citizens who are studying abroad may also participate in
these contests. The participants must have never enrolled in a university
or other higher educational institution. The contestants compete in two
categories: Juniors and Seniors. In the former category, only students up
to the 10th grade may participate. Being successful in the open contests
generally assumes knowledge outside the school curriculum.
Based on the results of all competitions during the year, about 20 IMO
team candidates are selected. IMO team selection contest for them is held
in April or May in two rounds. Each round is an IMO-style two-day com-
petition with 4.5 hours to solve 3 problems on both days. Some problems
in our selection contest are at the level of difficulty of the IMO but easier
problems are usually also included.
The problems of previous competitions can be downloaded at the Esto-
nian Mathematical Olympiads website.

Problems Listed by Topic

Number theory: O1, O7, O11, O13, O16, O18, F5, F12, S2
Algebra: O3, O8, O12, O19, F4, F8, F13, S4, S5
Geometry: O4, O9, O14, O20, F2, F6, F9, F14, F16, S3
Discrete mathematics: O2, O5, O6, O10, O21, O17, O21, F1, F3, F7, F10, F11,
F15, S1

3
Problems
 Selected Problems from Open Contests
O1 (Juniors.) Juku claims that if the sum of the squares of all digits of a
natural number is divisible by 3 then the number itself is divisible by 3. Is
Juku’s claim always true?
O2 (Juniors.) Xavier and Olivier are playing tic-tac-toe in the rectangular
grid of size 3 × 3 with modified rules. On every move, a player chooses an
empty square and writes his token into it. Players take turns alternately,
with Xavier starting. The player who is the first to occupy any three squares
that either are all in the same row or column or lie in pairwise distinct rows
and columns wins. Does either of the players have a winning strategy and
if yes then who?
O3 (Juniors.) Do there exist numbers a, b, c that satisfy the equation
2a(c − a) − b(2a + b) + c(2b − c) = 2020?
O4 (Juniors.) The bisector of the angle on vertex A of triangle ABC inter-
sects the circumcircle of triangle ABC at point F (F 6= A). Points D and E
are chosen on the sides AB and AC, respectively, in such a way that the
lines DE and BC are parallel. Let G and H be the points of intersection
of the rays FD and FE, respectively, with the circumcircle of triangle ABC
(G 6= F, H 6= F). The circumcircles of triangles AGD and AHE intersect at
point P (P 6= A). Prove that point P lies on the line AF.
O5 (Juniors.) Find all integers n > 3 such that one can write a number
(not necessarily an integer) into each vertex of a regular n-gon in such a
way that both following conditions are met:
(1) Whenever three consecutive vertices of the n-gon, taken clockwise,
contain numbers x, y and z, respectively, the equality x = |y − z| holds;
(2) The sum of the numbers in all vertices of the n-gon is 1.
O6 (Juniors.) A set of 8 dominoes is given, each consisting of two unit
squares:

Is it possible to completely cover a rectangular grid of size 4 × 4 with these

dominoes in such a way that all rows and columns of the grid contain the
same number of pips?
O7 (Juniors.) Find all quadruples ( p, q, r, s) of primes that satisfy the fol-
lowing system of equations:

6p + 5q + 5r + 3s = 130
3p + 3q + 5r + 6s = 130

5
 O8 (Juniors.) In a math period, the teacher asks pupils to solve quadratic
equations of the form x2 + px + q = 0 where p and q are some integers. The
teacher obtains every new equation by either increasing by 1 or decreasing
by 1 the value of either p or q in the equation just solved. In the initial
equation, p = 2020 and q = 2010, whereas in the last equation, p = 2010
and q = 2020. Is it definitely true that both solutions of at least one equation
solved during the period are integers?
O9 (Juniors.) Let ABC be a triangle such that AB = AC. Point K lies on
the altitude drawn from vertex A and point L is chosen on the line BK in
such a way that AL k BC. Prove that if KC ⊥ CL then point L lies on the
bisector of the external angle on vertex C of triangle ABC.
O10 (Juniors.) Let n be a natural number, n > 2. There are n lamps on a
circle. The lamps are labeled clockwise by natural numbers from 1 kuni n.
Each lamp can be either on or off. A switch between every two adjacient
lamps enables one to change the state of both lamps simultaneously. In the
beginning, all lamps are off. How many distinct configurations of states of
lamps is it possible to achieve using these switches?
O11 (Seniors.)
(a) Find the largest number expressible as the difference of two two-digit
numbers obtained from each other by changing the order of digits.
(b) The same question with three-digit instead of two-digit numbers.
O12 (Seniors.) Do there exist real numbers x, y, z, t that meet the following
system of equations?
 1 + x 3 + y2 = 0


 1 + y3 + z2 = 0



1 + z3 + t2 = 0
 1 + t3 + x 2 = 0



x+y+z+t = 0


O13 (Seniors.) Let n be a fixed positive integer. Find all triples ( a, b, c) of

integers satisfying the following system of equations:
 n +3
a + b n +2 c + c n +1 a 2 + a n b 3 = 0
b n + 3 + c n +2 a + a n +1 b 2 + b n c 3 = 0
 n +3
c + a n +2 b + b n +1 c 2 + c n a 3 = 0
O14 (Seniors.) The incircle of triangle ABC touches the sides AB and AC
at points K and L, respectively. The line BL intersects the incircle of triangle
ABC at point M (M 6= L). A circle passing through point M touches the
lines AB and BC at points P and Q, respectively, and intersects the incircle
of triangle ABC at point N (N 6= M). Prove that if KM k AC then points P,
N and L are collinear.
O15 (Seniors.) There is a rectangular grid with 3 rows and n columns on
a blackboard.
6
 (a) Find the number of ways to write exactly one of numbers 1, 2, . . . , 3n
into each square in such a way that all the following conditions are met:
(1) Different squares contain different numbers.
(2) For each i = 1, 2, . . . , 3n − 1, the numbers i and i + 1 are written
into adjacent squares (i.e., squares having a common side).
(3) The numbers 1 and 3n are written into adjacent squares.
(b) The same question, but as the 3rd condition, the number 1 must be
written into the leftmost and the number 3n into the rightmost column.
√ √
√O16 (Seniors.) Find the least positive integer n such that 5 5n, 6 6n and
7
7n are integers.
O17 (Seniors.) How many positive integers, where the only allowed digits
are 0 and 1, are less than 11111100100?
O18 (Seniors.) Prove that a2020 + 10a1010 + 1001 is prime for no integers a.
O19 (Seniors.) There are n lists of candidates taking part in elections. Let
hi be the total number of votes given for the candidates of the ith list. There
are M seats in the representative assembly.
Anna proposes the following system for delivering mandates: For each list,
one computes a reference number vi = a h+i 1 where ai is the number of man-
i
dates already given to the ith list (initially ai = 0, i.e., the reference number
of each list equals its number of votes). On every step (M times in total),
one chooses the list with the greatest reference number (if several lists share
the first place, one of them is chosen randomly) and adds one mandate to
this list, after which the reference number of this list is recomputed.
Bert’s idea for delivering mandates is to multiply the number of votes of
every list by M/K where K = h1 + . . . + hn , whereby fractional results are
rounded downwards. As rounding may cause some seats to be undeliv-
ered, he proposes multiplying all numbers of votes of the lists by some
j β, kso that the number of mandates given to the ith list
suitable coefficient
βhi M
would be mi = K where m1 + . . . + mn = M.
Prove that if such coefficient β exists then Anna’s and Bert’s methods lead
to the same distribution of mandates.
O20 (Seniors.) The lines tangent to the circumcircle of triangle ABC at
points B and C intersect at point D. The circumcircle of triangle BCD inter-
sects the lines AB and AC the second time at points K and L, respectively.
Prove that the line AD bisects the line segment KL.
O21 (Seniors.) Let n be a positive integer. Find the largest
• •
number of knights that can be placed on a board of size n × n • •
in such a way that no two knights attack each other. A knight N
attacks precisely the squares that are located either horizon- • • • •
tally by one square and vertically by two squares away or
horizontally by two squares and vertically by one square away.

7
 Selected Problems from the Final Round of
National Olympiad
F1 (Grade 9.) Can natural numbers 1 through 12 be written
into the squares of the grid shown in the figure in such a way
that all following conditions are met?
(1) Each square contains exactly one number.
(2) The sums of numbers in both rows consisting of 4 squares, in both
columns consisting of 4 squares, and in 4 middle squares are all equal.
(3) The numbers in any two squares with a common side or vertex differ
from each other by at least 2.
F2 (Grade 9.) Let D be the foot of the altitude drawn to the hypotenuse
AB of a right triangle ABC. The inradii of the triangles ABC, CAD and
CBD are r, r1 and r2 , respectively. Prove that CD = r + r1 + r2 .
F3 (Grade 9.) Players A, B and C are playing the following game. Ini-
tially, the number 1 is written on a blackboard. On their move, each player
replaces the number n currently on the blackboard with either n + 1, 7n + 7,
or 4n3 + 3n + 4 at their own choice, under the condition that the new num-
ber must not be larger than 109 . The player A makes the first move, then
B takes turn, then C, after him A again etc., until some player cannot make
a legal move. The player who makes the last move wins. Can any of the
players win the game against every legal play by the opponents and if yes
then who?
F4 (Grade 10.) Let a, b, c, d be positive real numbers satistfying the sys-
tem of equations

2 1 1
 a + b2 = 2 ,


 b2 + 4 = 8,

c2


 c2 + 16
d2
= 2,
 2 4

d + a2 = 32.
Determine the product abcd.
F5 (Grade 10.) Find all positive integers k for which there√is a right trian-
gle with legs of integral lengths and hypotenuse of length 88 . . . 822 . . . 2,
where the number under the root consists of exactly k eights and exactly k
twos.
F6 (Grade 10.) Let D and E be the midpoints of sides AB and AC, re-
spectively, of a triangle ABC. Prove that the line AB is tangent to the cir-
cumcircle of the triangle BEC if and only if the line AC is tangent to the
circumcircle of the triangle BED.
F7 (Grade 10.) Natural numbers 1 through n are written on a black-
board. On each move, one erases from the blackboard 2 or more numbers
8
whose sum is divisible by any of the chosen numbers and writes their sum
on the blackboard. Two players make moves by turns and the player who
cannot move loses the game. Which player can win the game against any
play by the opponent, if:
(a) n = 6;
(b) n = 11?
F8 (Grade 11.) Prove that
1
sin 10◦ · cos 20◦ · sin 30◦ · cos 40◦ · sin 50◦ · cos 60◦ · sin 70◦ · cos 80◦ = .
256
F9 (Grade 11.) The bisector of the internal angle on vertex A of a triangle
ABC intersects the side BC at point D. The line tangent to the circumcircle
of the triangle ABC at point A intersects the line BC at point K. Prove that
KA = KD.
F10 (Grade 11.) Prove that in every solved sudoku X X XX
the squares marked “X” in the figure contain precisely X X XX
YYYYY
the same digits as the squares marked “Y”, counting Y Y
repetitions. (For instance, if the squares marked “X” Y Y
contain three digits 3 in total then the squares marked Y Y
YYYYY
“Y” also contain three digits 3 in total.) XX XX
XX XX

Remark: A solved sudoku is a board of size 9 × 9 whose every row, every

column and every part of size 3 × 3 bounded by bold rules contains every
digit from 1 to 9 exactly once.
F11 (Grade 11.) Every term of the sequence a1 , a2 , a3 , . . . is either 0 or 1.
It is known that both 0 and 1 occur at least 1010 times among every 2021
consecutive terms of the sequence. May one be sure that the sequence is
periodic from some place on, i.e., there exist positive integers n and p such
that an+i = an+i+ p for every natural number i?
F12 (Grade 12.) Find all pairs ( a, b) of positive integers such that a > b
and
1 1 1
+ = .
a b 2021
F13 (Grade 12.) Anna, Anne and Anni seek for real solutions ( x, y) to the
system of equations
 3
4x y − x4 − 3x2 y2 = 2021,
4y3 x − y4 − 3y2 x2 = 2021.
Anna claims that the system of equations has a solution. Anne claims that
the system of equations has no solution but at least one of the two equations
has solutions. Anni claims that the system of equations has no solution and,
even worse, neither of the two equations alone has a solution. Who is right?

9
 F14 (Grade 12.) Point D inside an acute triangle ABC satisfies
∠ ADC = ∠ BDA = 180◦ − ∠CAB.
Prove that the point symmetric to point A w.r.t. point D lies on the circum-
circle of the triangle ABC.
F15 (Grade 12.) There are 6 distinct lines chosen in the space. Find the
largest possible number of points where at least 3 chosen lines intersect.
F16 (Grade 12.) Two regular polygons have a common circumcircle. The
sum of the areas of the incircles of these polygons equals the area of their
common circumcircle. Find all possibilities of how many vertices can the
two polygons have.

Selected Problems from the IMO Team Selection

Contests
S1 Juku has the first 100 volumes of the Harrie Totter book series at his
home. For every i and j, where 1 6 i < j 6 100, call the pair (i, j) reversed if
volume No j is before volume No i on Juku’s shelf. Juku wants to arrange
all volumes of the series to one row on his shelf in such a way that there
does not exist numbers i, j, k, where 1 6 i < j < k 6 100, such that pairs
(i, j) and ( j, k) are both reversed. Find the largest number of reversed pairs
that can occur under this condition.
S2 Find all polynomials P( x ) with integral coefficients whose values at
points x = 1, 2, . . . , 2021 are numbers 1, 2, . . . , 2021 in some order.
S3 (a) There are 2n rays marked in a plane, with n being a natural
number. Given that no two marked rays have the same direction and no
two marked rays have a common initial point, prove that there exists a
line that passes through none of the initial points of the marked rays and
intersects with exactly n marked rays.
(b) Would the claim still hold if the assumption that no two marked rays
have a common initial point was dropped?
S4 Positive real numbers a, b, c satisfy abc = 1. Prove that
a b c 3
+ + > .
1+b 1+c 1+a 2
S5 Find all polynomials P( x, y) with real coefficients which for all real
numbers x and y satisfy P( x + y, x − y) = 2P( x, y).

10
Problems with Solutions
 Selected Problems from Open Contests
O1 (Juniors.) Juku claims that if the sum of the squares of all digits of a
natural number is divisible by 3 then the number itself is divisible by 3. Is
Juku’s claim always true?
Answer: No.
Solution: The sum of the squares of the digits of the number 112 is 6 which
is divisible by 3, while the number 112 is not divisible by 3.
O2 (Juniors.) Xavier and Olivier are playing tic-tac-toe in the rectangular
grid of size 3 × 3 with modified rules. On every move, a player chooses an
empty square and writes his token into it. Players take turns alternately,
with Xavier starting. The player who is the first to occupy any three squares
that either are all in the same row or column or lie in pairwise distinct rows
and columns wins. Does either of the players have a winning strategy and
if yes then who?
Answer: Yes, Xavier.
Solution: Note that every two unit squares uniquely determine a unit square
that constitutes a winning triple together with the given two squares. La-
bel all unit squares except the middle square clockwise with numbers 1
through 8 (Fig. 1); let Xavier initially play into the middle square and if
Olivier then moves into the square No. i then let Xavier make his second
move into the square No. i + 1 (square No. 1 if i = 8; all possible po-
sitions after Xavier’s second move, modulo rotation of the table, are de-
picted in Figures 2 and 3). In order to prevent loss of the game, Olivier
must occupy the unit square that constitutes a winning triple together with
the two squares occupied by Xavier. Let Xavier then move into the square
that constitutes a winning triple together with the two squares previously
occupied by Olivier (all possible positions after Xavier’s third move are de-
picted in Figures 4 and 5). Then Olivier can’t win on his next move. But the
three squares occupied by Xavier contain three pairs, out of which only one
pair has its corresponding winning third square occupied by Olivier. Thus
Olivier can’t block them all, whence Xavier can win on his next move.
Remark: The strategy described in the solution is not the only winning strat-
egy that Xavier has.

1 2 3 O X O X O X O X
8 4 X X X X X X
7 6 5 O O

Fig. 1 Fig. 2 Fig. 3 Fig. 4 Fig. 5

O3 (Juniors.) Do there exist numbers a, b, c that satisfy the equation

2a(c − a) − b(2a + b) + c(2b − c) = 2020?
12
Answer: No.
Solution: Transforming the l.h.s. of the equation gives
2a(c − a) − b(2a + b) + c(2b − c) = 2ac − 2a2 − 2ab − b2 + 2bc − c2
= − a2 − ( a + b − c )2 .
The equality − a2 − ( a + b − c)2 = 2020 cannot be valid since all terms of its
l.h.s. are non-positive whereas the r.h.s. is positive.
O4 (Juniors.) The bisector of the angle on vertex A of triangle ABC inter-
sects the circumcircle of triangle ABC at point F (F 6= A). Points D and E
are chosen on the sides AB and AC, respectively, in such a way that the
lines DE and BC are parallel. Let G and H be the points of intersection
of the rays FD and FE, respectively, with the circumcircle of triangle ABC
(G 6= F, H 6= F). The circumcircles of triangles AGD and AHE intersect at
point P (P 6= A). Prove that point P lies on the line AF.
Solution 1: Let K and L be the points of intersection of the line AF with lines
BC and DE, respectively (Fig. 6). Then
∠ AGD = ∠ AGF = ∠ AGC + ∠CGF = ∠ ABC + ∠CAF
= ∠ ABK + ∠KAB = ∠CKA = ∠KLD = 180◦ − ∠ ALD.
Hence the quadrilateral AGDL is cyclic. Interchanging the roles of points B
and C, points D and E and also points G and H, we can similarly prove that
the quadrilateral AHEL is cyclic. Thus the circumcircles of triangles ADG
and AEH meet at point L, i.e., P = L. Point L was chosen on the line AF.
Solution 2: Choose a point Z on the same side of the line AF as point B such
that ZF k BC k DE (Fig. 7). Then ∠ZFB = ∠ FBC = ∠ FAC = ∠ FAB =
∠ FCB. Hence ZF is tangent to the circumcircle of the triangle ABC at F,
implying that ∠GFZ = ∠GHF = ∠GHE. As ∠GFZ = ∠ DFZ = ∠ FDE =
180◦ − ∠GDE, we altogether have ∠GHE = 180◦ − ∠GDE. Consequently,
the quadrilateral DEHG is cyclic. Hence | FD | · | FG | = | FE| · | FH |, implying
that F lies on the radical axis of the circumcircles of triangles ADG and
AEH. The desired result follows.
A A

H H
G G
L
D E D E

K
B C B C

F Z F
Fig. 6 Fig. 7

13
 O5 (Juniors.) Find all integers n > 3 such that one can write a number
(not necessarily an integer) into each vertex of a regular n-gon in such a
way that both following conditions are met:
(1) Whenever three consecutive vertices of the n-gon, taken clockwise,
contain numbers x, y and z, respectively, the equality x = |y − z| holds;
(2) The sum of the numbers in all vertices of the n-gon is 1.
Answer: All positive multiples of 3.
Solution: Let the vertices of an n-gon A0 A1 . . . An−1 be labeled with num-
bers satisfying the conditions. Let a be the least among these numbers.
W.l.o.g., assume that A0 contains a and the indices of vertices are increasing
counterclockwise. Let b and c be the numbers at vertices An−1 and An−2 ,
respectively (Fig. 8). By condition (1), a = |b − c| > 0. By the choice of a, we
must have b > a, whence by condition (1), A1 contains b − a. As b − a > a by
the choice of a, the condition (1) also implies that A2 contains b − 2a. Hence
A3 contains (b − a) − (b − 2a) = a by condition (1) and non-negativity of a.
Since the vertices can be renumerated without changing the direction in
such a way that A3 becomes A0 , we can conclude that A6 also contains a.
Similarly, every third vertex contains a. If the number n of vertices is not
divisible by 3 then either An−1 or A1 must contain a and, as we can repeat
this argument, also An−2 or A2 , respectively, contains a. Thus three consec-
utive vertices contain a. Applying the condition (1) to these three vertices,
we obtain a = | a − a| = 0. But 0 being in two consecutive vertices implies
that 0 is in all vertices. Then the sum of all labels is 0, contradicting the
condition (2).
This shows that n must be divisible by 3. Let n = 3k where k is a positive
integer. For every 3k-gon, the conditions of the problem can be satisfied by
1
writing 0 into every third vertex and 2k into all other vertices (Fig. 9 depicts
the situation for k = 4).

A6
A7 A5
1
0 1
A8 8 8 A4
1 1
8 8
A9 0 0 A3
1 1
8 8
c
A n−2 A2 A10 1 1 A2
b a
8
0
8

A n−1 A1 A11 A1
A0 A0
Fig. 8 Fig. 9

O6 (Juniors.) A set of 8 dominoes is given, each consisting of two unit

squares:

14
Is it possible to completely cover a rectangular grid of size 4 × 4 with these
dominoes in such a way that all rows and columns of the grid contain the
same number of pips?
Answer: Yes.
Solution: Figure 10 shows one possibility.

1 2 8 9
4 5 5 6
8 7 3 2
7 6 4 3

Fig. 10

O7 (Juniors.) Find all quadruples ( p, q, r, s) of primes that satisfy the fol-

lowing system of equations:

6p + 5q + 5r + 3s = 130
3p + 3q + 5r + 6s = 130
Answer: (11, 3, 2, 13).
Solution 1: Subtracting the second equation from the first one gives 3p +
2q − 3s = 0 which implies 2q = 3(s − p). Thus 2q is divisible by 3. As both
2 and q are primes, this implies q = 3. Substituting q = 3 into the initial
system of equation and simplifying gives

6p + 5r + 3s = 115,
(1)
3p + 5r + 6s = 121.
If r and s were both odd then also 5r and 3s would be odd, whence 5r +
3s would be even. As 6p is even, too, the l.h.s. of the first equation of
system (1) would be even and could not equal the r.h.s. 115. Thus one of r
and s is even, i.e., r = 2 or s = 2. Analogously if p and r were both odd
then the second equation of system (1) would give a contradiction; hence
p = 2 or r = 2. Consequently, if r 6= 2 then p = s = 2, but substituting
p = s into system (1) and subtracting the second equation from the first one
gives 0 = −6. The contradiction shows that r = 2. Substituting r = 2 into
system (1) and simplifying gives

6p + 3s = 105,
3p + 6s = 111.
Solving this equation gives p = 11 and s = 13.
Solution 2: As in Solution 1 we find q = 3 and obtain system (1). Subtracting
the first equation from the second one in system (1) gives 3s − 3p = 6,
implying s = p + 2. The least pairs of prime numbers with difference 2 are
(3, 5), (5, 7) and (11, 13). The next such pair (17, 19) gives 6p + 3s > 115,
15
implying that no more pairs are suitable. Substituting the three pairs one
by one into system (1) and simpligying each time gives 5r = 82, 5r = 64
and 5r = 10, respectively. Only the last alternative leads to an integral
solution r = 2, which is a prime, too. Consequently, the only possibility is
( p, q, r, s) = (11, 3, 2, 13).
O8 (Juniors.) In a math period, the teacher asks pupils to solve quadratic
equations of the form x2 + px + q = 0 where p and q are some integers. The
teacher obtains every new equation by either increasing by 1 or decreasing
by 1 the value of either p or q in the equation just solved. In the initial
equation, p = 2020 and q = 2010, whereas in the last equation, p = 2010
and q = 2020. Is it definitely true that both solutions of at least one equation
solved during the period are integers?
Answer: Yes.
Solution: In the first equation, one has p − q = 10, while in the last equation,
one has p − q = −10. At each step, either p or q changes exactly by 1,
whence also p − q changes exactly by 1. Thus at some step one must have
an equation where p − q = 1, or equivalently, p = q + 1. The equation
x2 + (q + 1) x + q = 0 has integral solutions −1 and −q.
O9 (Juniors.) Let ABC be a triangle such that AB = AC. Point K lies on
the altitude drawn from vertex A and point L is chosen on the line BK in
such a way that AL k BC. Prove that if KC ⊥ CL then point L lies on the
bisector of the external angle on vertex C of triangle ABC.
Solution 1: As AL k BC and AK ⊥ BC (Fig. 11), we have ∠KAL = 90◦ .
As KAL and KCL are both right angles, points A and C lie on circle with
diameter KL. By inscribed angles, ∠KCA = ∠KLA. On the other hand,
∠KLA = ∠KBC and, by symmetry of isosceles triangle, ∠KBC = ∠KCB.
Consequently, ∠KCA = ∠KCB, implying that KC bisects the internal angle
on vertex C of the triangle ABC. As KCL is right angle, CL bisects the
corresponding external angle.
Solution 2: The altituded drawn from the vertex angle of the isosceles trian-
gle ABC is the perpendicular bisector of its base. All points of the perpen-
dicular bisector of a line segment lie at equal distance from the endpoints
of the line segment, implying that |KB| = |KC |. As AL k BC and AK ⊥ BC,
we conclude AK ⊥ AL.
A L

B C
Fig. 11

16
 A L
Q A L

K K

B C P B C
Fig. 12 Fig. 13

Let P be a point on line BC such that C is between B and P, and let lines CK
and AL intersect at Q (Fig. 12). As ∠ BKC = ∠ LKQ and ∠KBC = ∠KLQ,
triangles KBC and KLQ are similar. Together with the equality |KB| = |KC |,
it implies |KL| = |KQ|. Thus KA is the altitude drawn from the vertex angle
of the isosceles triangle KLQ, implying also | AL| = | AQ|. Hence A is the
midpoint of the hypotenuse of the right triangle CLQ, implying that A is the
center of the circumcircle of the triangle CLQ. Consequently, | AC | = | AL|,
implying that ∠ ACL = ∠ ALC = ∠ LCP, i.e., CL bisects the external angle
on vertex C of the triangle ABC.
Solution 3: When point K moves along the altitude drawn from vertex A
of the triangle ABC farther from point A, the angle KCB decreases, while
point L also moves farther from point A, causing the angle LCB to increase.
Thus the difference ∠ LCB − ∠KCB also increases in this process and can
equal 90◦ in the case of exactly one location of point K. Hence it suffices
to show that if L lies at the bisector of the external angle on vertex C of the
triangle ABC then ∠ LCK = 90◦ .
So let L lie at the bisector of the external angle on vertex C of the triangle
ABC (Fig. 13). As AL k BC and AK ⊥ BC, we have AK ⊥ AL; as AK
is the altitude drawn from the vertex angle of the triangle ABC, it bisects
the vertex angle, whence also L lies on the bisector of the external angle on
vertex A of the triangle ABC. Thus L is the center of the excircle tanglent
to side AC of the triangle ABC and lies also on the bisector of the internal
angle on vertex B of the triangle ABC. Hence both AK and BK are angle
bisectors, implying that K is the intersection point of angle bisectors of the
triangle ABC and CK is the bisector of the internal angle on vertex C of the
triangle ABC. Consequently, CK and CL are perpendicular.
O10 (Juniors.) Let n be a natural number, n > 2. There are n lamps on a
circle. The lamps are labeled clockwise by natural numbers from 1 kuni n.
Each lamp can be either on or off. A switch between every two adjacient
lamps enables one to change the state of both lamps simultaneously. In the
beginning, all lamps are off. How many distinct configurations of states of
lamps is it possible to achieve using these switches?
Answer: 2n−1 .
Solution 1: As the state of each lamp is determined by the parity of the

17
number of switchings that influence this lamp, the result of every sequence
of switchings is determined by the set of switches that have been touched
an odd number of times. Hence all possible states can be achieved by
sequences of switchings that touch some switches once and do not touch
other switches at all, whereby the order of switchings is irrelevant.
Note that every two set of switches, one of which contains precisely the
switches that the other one does not, lead to equal final states, because
touching these two sets of switches in a row one touches each switch ex-
actly once and thus changes the state of each lamp exactly twice. As there
are 2n sets of switches and these sets can be divided into pairs that give
equal final states, there can be at most 2n−1 final states.
If two sets of switches are neither equal nor complementary (in the sense
of previous paragraph) then touching these two sets of switches in a row
does not give back the initial state, because this sequence of switchings is
equivalent to touching a set of switches that contains some switch and does
not contain some other switch, but then there must exist a lamp, the switch
in only one side of which is touched. Hence every pair of sets of switches
defined in the previous paragraph gives a unique final state, i.e., there are
2n−1 possible distinct final states.
Solution 2: After every switching, either two more lamps are burning, two
less lamps are burning, or the number of burning lamps does not change.
As the initial number of burning lamps is even, the number of burning
lamps stays even.
We show that every state with an even number of burning lamps can be
achieved. As the switchings are invertible (repeating the same touch re-
verts the previous change), it suffices to show that, starting from any state
with an even number of burning lamps, one can achieve the state where no
lamp is burning. Indeed, while there exist two consecutive burning lamps,
we can switch off both these lamps. If there do not exist such lamps, but
some lamps are still burning, one can shift a lonely burning lamp clock-
wise by repeatedly touching the clockwise nearest switch until two consec-
utive lamps are burning. Switching these two lamps off decreases the total
number of burning lamps by 2. This way, we can decrease the number of
burning lamps until all lamps are switched off.
2b n c
Thus the number of all achievable states is Cn0 + Cn2 + . . . + Cn 2 which
equals 2n−1 .
O11 (Seniors.)
(a) Find the largest number expressible as the difference of two two-digit
numbers obtained from each other by changing the order of digits.
(b) The same question with three-digit instead of two-digit numbers.
Answer: (a) 72; (b) 801.

18
Solution:
(a) Let the given two-digit number be ab. The only number that can be
obtained by changing the order of digits is ba. The difference of these num-
bers is (10a + b) − (10b + a) = 9( a − b). To obtain the largest difference,
a must be as large as possible and b as small as possible. Since b = 0 is
impossible, we must have a = 9 and b = 1 giving a − b = 8. Hence the
desired largest difference is 72.
(b) Let the given three-digit number be abc. Interchanging the last two
digits can change it by less than 100. Interchanging the first two digits can
change the number by at most 720 by part a) of the problem. It remains to
study cases where changing the order of digits results in cab, bca or cba.
• The difference of numbers abc and cab is (100a + 10b + c) − (100c +
10a + b) = 9(10a + b − 11c). To obtain the largest difference, a and b
must be as large as possible and c as small as possible. Since c is the first
digit of the number, c = 0 is impossible, whence the largest difference
is obtained if a = b = 9 and c = 1. This difference is 991 − 199 = 792.
• The difference of numbers abc and bca is (100a + 10b + c) − (100b +
10c + a) = 9(11a − 10b − c). To obtain the largest difference, a must be
as large as possible and both b and c as small as possible, i.e., a = 9,
b = 1 and c = 0. This difference is 910 − 109 = 801.
• The difference of numbers abc and cba is (100a + 10b + c) − (100c +
10b + a) = 99( a − c). To obtain the largest difference, a must be as
large as possible and c as small as possible. Since c 6= 0, the largest
difference is obtained if a = 9 and c = 1. Then the difference of the
three-digit numbers is 99 · 8 = 792.
Consequently, the desired largest difference is 801.
O12 (Seniors.) Do there exist real numbers x, y, z, t that meet the following
system of equations?
1 + x 3 + y2 = 0


 1 + y3 + z2 = 0



1 + z3 + t2 = 0
1 + t3 + x 2 = 0




x+y+z+t = 0


Answer: No.
Solution 1: The first equation implies x3 = −y2 − 1. Thus x3 < 0, implying
also x < 0. Similarly from the second, third and fourth equations we obtain
y < 0, z < 0 and t < 0, respectively. The sum of negative numbers x, y, z, t
is negative, contradicting the fifth equation.
Solution 2: Suppose that the system has a solution. W.l.o.g., let x be variable
with the largest value. Then 4x > x + y + z + t, which by the last equation
implies x > 0. Consequently, also x3 > 0. As y2 > 0, this implies 1 + x3 +
y2 > 1, contradicting the first equation. Hence no solution can exist.

19
 O13 (Seniors.) Let n be a fixed positive integer. Find all triples ( a, b, c) of
integers satisfying the following system of equations:
 n +3
a + b n +2 c + c n +1 a 2 + a n b 3 = 0
b n + 3 + c n +2 a + a n +1 b 2 + b n c 3 = 0
 n +3
c + a n +2 b + b n +1 c 2 + c n a 3 = 0
Answer: (0, 0, 0).
Solution: If a = 0 then the first equation implies bn+2 c = 0. Hence b = 0
or c = 0; w.l.o.g., b = 0. Then the last equation reduces to cn+3 = 0 which
implies c = 0. Thus if a = 0 then a = b = c = 0. Analogously we can prove
that if b = 0 or c = 0 then a = b = c = 0. The triple (0, 0, 0) satisfies the
equation.
It remains to consider triples ( a, b, c) whose all terms are different from 0.
Let p be an arbitrary prime number. If p | a then the first equation implies
p | bn+2 c. Thus p | b or p | c; w.l.o.g., p | b. Then the last equation gives
p | cn+1 which implies p | c. Thus if p | a then p | a, p | b, p | c. Analogously
we can prove that if p | b or p | c then p | a, p | b, p | c.
But rewriting a = pa0 , b = pb0 , c = pc0 enables us to divide both sides of
all equations by pn+3 , giving a similar system of equations having a0 , b0 ,
c0 in the role of a, b, c. Hence the triple ( a0 , b0 , c0 ) also satisfies the system
of equations. As the numbers other than 0 cannot be infinitely divided in
integers, a finite number of divisions should give us a solution whose terms
have no common prime factors. By the previous paragraph, this is possible
only if the values of variables have no prime factors, i.e., a = ±1, b = ±1,
c = ±1. We show now that such solutions do not exist. Indeed, if n is even
then the system of equations reduces to

 a + c + c + b = 0,
b + a + a + c = 0,
c + b + b + a = 0.


Adding all equations results in 4( a + b + c) = 0, implying a + b + c = 0;

but the sum of three odd numbers cannot equal the even number 0. If n is
odd then the system of equations reduces to

 1 + bc + 1 + ab = 0,
1 + ca + 1 + bc = 0,
1 + ab + 1 + ca = 0.


Adding all equations results in 2( ab + bc + ca) + 6 = 0, implying ab + bc +

ca = −3. As | ab| = |bc| = |ca| = 1, the only possibility is ab = bc =
ca = −1. This demands that a, b, c have pairwise opposite signs which is
impossible. Consequently, no solutions except (0, 0, 0) exist.
O14 (Seniors.) The incircle of triangle ABC touches the sides AB and AC
at points K and L, respectively. The line BL intersects the incircle of triangle
ABC at point M (M 6= L). A circle passing through point M touches the

20
lines AB and BC at points P and Q, respectively, and intersects the incircle
of triangle ABC at point N (N 6= M). Prove that if KM k AC then points P,
N and L are collinear.
Solution: Note that the circles of the problem can be obtained from each
other by homothetic transformation with center B since both are tangent
to sides BA and BC. Let X be the other intersection point of the line BL
with the circumcircle of the triangle MPQ (Fig. 14). The aforementioned
homothety takes point P to point K, point M to point L, and point X to
point M. By the homothety, ∠ PXM = ∠KML. Hence ∠KML = ∠KLA =
∠ LKM. Finally,
∠ PN M + ∠ LN M = (180◦ − ∠ PXM) + ∠ LKM
= (180◦ − ∠KML) + ∠KML = 180◦ .
This proves the desired claim that points P, N and L are collinear.
Remark: The converse is also true: if points P, N and L are collinear then
lines KM and AC are parallel. The proof is analogous.

P
X
N Q
K
M

A L C

Fig. 14

O15 (Seniors.) There is a rectangular grid with 3 rows and n columns on

a blackboard.
(a) Find the number of ways to write exactly one of numbers 1, 2, . . . , 3n
into each square in such a way that all the following conditions are met:
(1) Different squares contain different numbers.
(2) For each i = 1, 2, . . . , 3n − 1, the numbers i and i + 1 are written
into adjacent squares (i.e., squares having a common side).
(3) The numbers 1 and 3n are written into adjacent squares.
(b) The same question, but as the 3rd condition, the number 1 must be
written into the leftmost and the number 3n into the rightmost column.
n
Answer: (a) 3n · 2 2 for even n and 0 for odd n; (b) 2n .
21
Solution:
(a) Note that all three squares in the leftmost and rightmost columns must
be traversed consecutively because along with both corner squares of either
of the columns one has to traverse the middle square of the same column
and this cannot be done twice. In no other column can one traverse all three
squares consecutively because such column would divide the grid into two
parts between which one could not move without visiting a square twice.
Note also that one cannot traverse three or more consecutive squares of
the middle row consecutively because such trajectory would also divide
the grid into two isolated parts (Figures 15 and 16). For similar reasons,
when visiting two consecutive squares of the middle row consecutively, the
previous and the next square should be taken from one and the same row
(either top or bottom).
Hence apart from the first and the last column, every legal trajectory divides
the middle row into pairs of squares that are visited consecutively. There-
fore the required enumeration is impossible for odd n. For even n, one can
choose for every pair of squares in the middle row the row where the pre-
vious and the next square are located (either top or bottom); the number of
n −2
such pairs of squares is n− 2
2 , whence there are 2
2 possibilities to choose
the direction of continuation for every such pair of squares. The whole tra-
jectory is determined by these choices. For writing numbers in the case of a
fixed trajectory, there are 6n possibilites (2 possibilities to choose the direc-
tion and 3n possibilities to choose the initial square). Consequently, there
n −2 n
are 6n · 2 2 = 3n · 2 2 possibilities of the required enumeration for even n.

Fig. 15 Fig. 16

(b) One cannot start from the middle square of the leftmost column since
either of the corner squares would become a dead end. Hence suppose
w.l.o.g. that one starts from the top left corner and visits k squares in the
top row (inclusive of the initial square), where 1 6 k 6 n (Fig. 17). Going
directly to the bottom row would generally divide the grid into two isolated
parts; also turning to the right along the middle row would make it impos-
sible to visit squares in the left and afterwards reach the rightmost column.
Hence one should turn to the left along the middle row. Similarly we find
that the leftmost part of size 3 × k of the grid should be traversed along
Z-shape (Fig. 18). After that, one should follow similar sules to traverse the
remaining part of size 3 × (n − k).
Repeating the argument, we can see that every legal trajectory divides the
grid into Z-shapes and S-shapes, whereby Z-shapes are immediately fol-

22
 1 2 ... k 1 2 ... k

Fig. 17 Fig. 18

lowed by S-shapes and vice versa. In every column but the last one we can
choose whether to start a new letter or make the current letter wider. In the
first column, we also can choose which corner to start in. Hence there are
2n possibilities altogether.
√ √
√O16 (Seniors.) Find the least positive integer n such that 5 5n, 6 6n and
7
7n are integers.
Answer: 235 · 335 · 584 · 790 .
Solution: Let n = 2α · 3β · 5γ · 7δ · s, where s is not divisible by 2, 3, 5 or
7; then 5n = 2α · 3β · 5γ+1 · 7δ · s, 6n = 2α+1 · 3β+1 · 5γ · 7δ · s and 7n =
2α · 3β · 5γ · 7δ+1 · s. Consequently:
√
• For 5 5n to be an integer, α, β, γ + 1 and δ must be divisible by 5;
√
• For 6 6n to be an integer, α + 1, β + 1, γ and δ must be divisible by 6;
√
• For 7 7n to be an integer, α, β, γ and δ + 1 must be divisible by 7.
Hence α and β must be divisible by 35, γ must be divisible by 42 and δ must
be divisible by 30. The least suitable value for α and β is 35 since 35 is the
least positive multiple of 35 and the next integer is divisible by 6. Studying
the multiples of 42 and 30 similarly shows that the least suitable value for γ
is 84 and the least suitable value for δ is 90. For the least suitable value for n,
take s = 1. Hence the desired number is 235 · 335 · 584 · 790 .
O17 (Seniors.) How many positive integers, where the only allowed digits
are 0 and 1, are less than 11111100100?
Answer: 2019.
Solution 1: The given number has 11 digits. There are 211 − 1 = 2047 posi-
tive integers with at most 11 digits each of which is either 0 or 1. We solve
the problem by subtracting the number of positive integers that are not less
than the given number.
The 11-digit numbers larger than the given number are all of the form
111111abcde. There are 25 = 32 numbers in this form. Among them, 4 num-
bers 11111100000, 11111100001, 11111100010 and 11111100011 are less than
the given number. Thus the number of positive integers consisting of zeros
and ones and being less than 11111100100 is 2047 − 32 + 4 = 2019.
Solution 2: Ordering any two digit sequences that constitute a positional
representation on different bases does not depend on the base. This means
that if a number is larger than another number on base 2 then the first
number is larger than the second number also on base 10. The number
23
11111100100 on base 2 equals 210 + 29 + 28 + 27 + 26 + 25 + 22 = 2020 on
base 10. Hence, to solve the problem, it suffices to count all positive integers
less than 2020. The result is obviously 2019.
O18 (Seniors.) Prove that a2020 + 10a1010 + 1001 is prime for no integers a.
Solution: Note that 1001 = 11 · 91. If 11 | a then also 11 | a2020 + 10a1010 +
1001. Otherwise, consider a5 modulo 11. A case study shows that if a is
congruent to 1, 3, 4, 5, or 9, then a5 is congruent to 1, and in all other cases, a5
is congruent to −1. Hence a10 ≡ 1 (mod 11) whenever 11 - a. This implies

101
that a1010 ≡ 1 (mod 11) and a2020 ≡ 1 (mod 11) because a1010 = a10
202
ja a2020 = a10 . Consequently, 10a1010 ≡ 10 (mod 11), implying that

a 2020 + 10a 1010 + 1001 ≡ 0 (mod 11). Thus a2020 + 10a1010 + 1001 cannot be
prime since obviously a2020 + 10a1010 + 1001 > 1001 > 11.
Remark: The fact a10 ≡ 1 (mod 11) directly follows from Fermat’s theorem.
O19 (Seniors.) There are n lists of candidates taking part in elections. Let
hi be the total number of votes given for the candidates of the ith list. There
are M seats in the representative assembly.
Anna proposes the following system for delivering mandates: For each list,
one computes a reference number vi = a h+i 1 where ai is the number of man-
i
dates already given to the ith list (initially ai = 0, i.e., the reference number
of each list equals its number of votes). On every step (M times in total),
one chooses the list with the greatest reference number (if several lists share
the first place, one of them is chosen randomly) and adds one mandate to
this list, after which the reference number of this list is recomputed.
Bert’s idea for delivering mandates is to multiply the number of votes of
every list by M/K where K = h1 + . . . + hn , whereby fractional results are
rounded downwards. As rounding may cause some seats to be undeliv-
ered, he proposes multiplying all numbers of votes of the lists by some
j β, kso that the number of mandates given to the ith list
suitable coefficient
βhi M
would be mi = K where m1 + . . . + mn = M.
Prove that if such coefficient β exists then Anna’s and Bert’s methods lead
to the same distribution of mandates.
Solution: Let the total number of mandates given to the ith list be mi in the
case of Bert’s method and mi0 in the case of Anna’s method. Suppose that
these methods result in different distribution of mandates. As the sum of
numbers of mandates must be the same, we must have mi0 > mi for some
i = 1, 2, . . . , n and m0j < m j for some j = 1, 2, . . . , n. As the numbers of
mandates are integers, we have mi0 > mi + 1 and m0j + 1 6 m j .
Consider the situation in the case of Anna’s method immediately after the
ith list having obtained its last mandate. Before obtaining the last mandate,
the ith list had mi0 − 1 mandates and reference number vi = (m0 −h1i )+1 = mhi0 ,
i i

24
while the jth list had at most m0j mandates, implying that a j 6 m0j and v j =
hj hj hi hj
a j +1 > m0j +1
. As the mandate was given to the ith list, mi0
> m0j +1
. Thus
hi mi0
hj > m0j +1
, implying that

hi m0 m +1
mj · > mj · 0 i > mj · i > mi + 1.
hj mj + 1 mj
On the other hand, from the specification of Bert’s method we know that
βh j M hi βh j M
   
hi hi βhi M βhi M
mj · = · 6 · = < + 1 = mi + 1.
hj hj K hj K K K
This inequation contradicts the previous inequation. Hence both methods
indeed lead to the same distribution of mandates.
Remark: The first method is called D’Hondt’s method after Victor D’Hondt,
a Belgian lawyer. The second method is called Jefferson’s method after the
president of the USA Thomas Jefferson.
O20 (Seniors.) The lines tangent to the circumcircle of triangle ABC at
points B and C intersect at point D. The circumcircle of triangle BCD inter-
sects the lines AB and AC the second time at points K and L, respectively.
Prove that the line AD bisects the line segment KL.
Solution 1: We prove that AKDL is a parallelogram; this implies the desired
claim since AD and KL are diagonals of this quadrilateral. We prove at first
that KD k AL. If K lies between A and B (Fig. 19) then by inscribed angles
∠ BAC = ∠ BCD and ∠ BCD = ∠ BKD. Consequently, ∠ BAL = ∠ BAC =
∠ BKD, implying KD k AL. If B lies between A and K (Fig. 20) then by in-
scribed angles ∠ BAC = ∠ BCD and ∠ BCD = 180◦ − ∠ BKD. Consequently,
∠ BAL + ∠ BKD = ∠ BAC + ∠ BKD = 180◦ , implying KD k AL. If A lies
between K and B (Fig. 21) then by inscribed angles ∠ BAC = 180◦ − ∠ BCD
and ∠ BCD = ∠ BKD. Consequently, ∠ BAL = 180◦ − ∠ BAC = ∠ BKD,
implying KD k AL again. Analogously, we can show that LD k AK. Alto-
gether, this establishes that AKDL is a parallelogram.
Solution 2: By inscribed angles, ∠ ABC = ∠ ALK, implying that the trian-
gles ABC and ALK are similar. Let the lines AD and KL intersect at point N
A D
A L K
C
A
B L B
K L K
C C
B

D D
Fig. 19 Fig. 20 Fig. 21

25
and let M be the midpoint of the side BC. It is known that the symmedian
line drawn through a vertex of a triangle and the lines tangent to the cir-
cumcircle of the triangle at the other two vertices meet in one point; hence
N is the point of intersection of the symmedian line drawn through ver-
tex A of the triangle ABC with line KL. By the definition of symmedian,
∠ BAM = ∠CAN = ∠ LAN, whence M and N are corresponding points in
similar triangles ABC and ALK. Thus N bisects the line segment KL.
Remark: The claim of the problem is true if the circumcircle of the triangle
BCD is replaced with any circle that passes through vertices B and C.
O21 (Seniors.) Let n be a positive integer. Find the largest
• •
number of knights that can be placed on a board of size n × n • •
in such a way that no two knights attack each other. A knight N
attacks precisely the squares that are located either horizon- • • • •
tally by one square and vertically by two squares away or
horizontally by two squares and vertically by one square away.
l 2m
Answer: n2 if n 6= 2; 4 if n = 2.
Solution:
l 2m Clearly one can place only 1 knight on an 1 × 1 board, which is
1
2 , and at most 4 knights on an 2 × 2 board. In the rest, assume n > 3.
If a set of unit squares is divided into pairs in such a way that a knight on
one square of any pair attacks the other square of that pair then there cannot
be more knights than half of the total number of unit squares in this set
without some knights attacking each other. In Figures 22, 23 and 24, all unit
squares of boards of size 2 × 4, 3 × 4 and 3 × 6 are divided into pairs whose
members are located at one knight move from each other. In Figures 25 and
26, all unit squares of boards of size 3 × 3 and 5 × 5 except the middle square
are divided into pairs whose members are located at one knight move from
each other. By the above, it is impossible to place knights to more than half
of unit squares of boards of size 2 × 4, 3 × 4 and 3 × 6. Taking into account
that an additional knight mayl mbe on the middle square, it also followslthat m
2 2
there cannot be more than 32 knights on a 3 × 3 board or more than 52
knights on a 5 × 5 board. As a 4 × 4 board can be formed from two 2 × 4
boards and a 6 × 6 board from two 3 × 6 boards, there cannot be more than
42 62
2 knights on a 4 × 4 board or more than 2 knights on a 6 × 6 board.

9 7 8
8 6 9 2 10 6 7 3
4 3 5 3 4 5 4 7 6 9 2 10 11
2 1 6 1 5 2 3 6 2 1 3 5 1 3 7
3 4 2 4 3 4 5 1 4 2 9 12 4 11 8
1 2 1 6 2 1 2 3 1 3 4 1 5 8 12 4

Fig. 22 Fig. 23 Fig. 24 Fig. 25 Fig. 26

26
If n > 7 then an n × n board can be divided into pieces of size 4 × 4, r × 4
and r × r, where 3 6 r 6 6 and 4 | n − r. By the above, neither 4 × 4
nor r × 4 board can contain more knights than half of the number of unit
squares (5 × 4 and 6 × 4 are divisible into rectangles of size 2 × 4 and 3 × 4).
The same holds for an r × r piece, if the middle square in the case of odd l mr
2
is not taken into account. Thus for no n > 3 can one place more than n2
knights on an n × n board.
On the other hand, coloring the unit squares black and white chesswise, one
can place a knight on all squares of one and the same color since a knight
attacks only squares of the opposite color. The number of unit squares of a
2
fixed color is n2 if n is even. In the case of odd n, the number oflunit
m squares
2
whose color coincides with the color of the middle square is n2 . Hence
l 2m
one can place n2 pairwise non-attacking knights on an n × n board.

Selected Problems from the Final Round of

National Olympiad
F1 (Grade 9.) Can natural numbers 1 through 12 be written
into the squares of the grid shown in the figure in such a way
that all following conditions are met?
(1) Each square contains exactly one number.
(2) The sums of numbers in both rows consisting of 4 squares, in both
columns consisting of 4 squares, and in 4 middle squares are all equal.
(3) The numbers in any two squares with a common side or vertex differ
from each other by at least 2.
Answer: Yes.
Solution: A configuration that meets all conditions is shown in Fig. 27.

11 9
6 3 5 12
1 8 10 7
4 2

Fig. 27

F2 (Grade 9.) Let D be the foot of the altitude drawn to the hypotenuse
AB of a right triangle ABC. The inradii of the triangles ABC, CAD and
CBD are r, r1 and r2 , respectively. Prove that CD = r + r1 + r2 .
Solution 1: Let a = | BC |, b = |CA|, c = | AB| and h = |CD |; then ab = ch =
( a + b + c)r.

27
 C

A D B

Fig. 28

Note that the triangles ABC, ACD and CBD are similar by two equal angles
(Fig. 28). As the similarity ratio of triangles ACD and ABC is bc and that of
triangles CBD and ABC is ac , we have r1 = bc · r and r2 = ac · r, whence
r + r1 + r2 = a+bc +c · r. As the equality ab = ( a + b + c)r implies r = a+ab
b+c ,
we obtain r + r1 + r2 = ab c . On the other hand, the equality ab = ch implies
h = abc . Consequently, r + r1 + r2 = h.
Solution 2: Let the incenters of the triangles ABC, ACD and BCD be I, I1 and
I2 . Let the points of tangency of the incircle of the triangle ABC with sides
BC, CA and AB be K, L and M, respectively; let the points of tangency of the
incircle of the triangle ACD with sides DC, CA and AD be K1 , L1 and M1 ,
respectively; let the points of tangency of the incircle of the triangle BCD
with sides DC, CB and BD be K2 , L2 and M2 , respectively (Fig. 29). As a
tangent line is perpendicular to the radius drawn to the point of tangency,
we have ∠ IKC = 90◦ = ∠ ILC. Thus IKCL is a square by three right angles
and the equality IK = IL. Hence CK = CL = r. Similarly we see that
C

K
L L2

L1 I
K1 K2 I2
I1

A M1 D M M2 B

Fig. 29

28
I1 K1 DM1 and I2 K2 DM2 are squares, whereby DK1 = DM1 = r1 ja DK2 =
DM2 = r2 . By property of tangent line segments, we have CK1 = CL1 ,
AL = AM and AL1 = AM1 , whence
CD = CK1 + K1 D = CL1 + r1 = CL + LL1 + r1
= r + AL − AL1 + r1 = r + AM − AM1 + r1 = r + MM1 + r1 .
By symmetry, we also get CD = r + MM2 + r2 . After adding up these two
equalities and taking into account that MM1 + MM2 = DM1 + DM2 =
r1 + r2 , we obtain
2CD = 2(r + r1 + r2 ).
Consequently, CD = r + r1 + r2 .
F3 (Grade 9.) Players A, B and C are playing the following game. Ini-
tially, the number 1 is written on a blackboard. On their move, each player
replaces the number n currently on the blackboard with either n + 1, 7n + 7,
or 4n3 + 3n + 4 at their own choice, under the condition that the new num-
ber must not be larger than 109 . The player A makes the first move, then
B takes turn, then C, after him A again etc., until some player cannot make
a legal move. The player who makes the last move wins. Can any of the
players win the game against every legal play by the opponents and if yes
then who?
Answer: Yes, C.
Solution: The game lasts while the number on the blackboard stays less
than 109 , because it is possible to make a move of the first kind (replace n
with n + 1). As the number on the blackboard is increased by at least 1 by
every move, it cannot stay less than 109 infinitely. When the number on the
blackboard equals 109 , there is no legal move. As numbers larger than 109
cannot appear on the blackboard, the last move is made by the player who
writes the number 109 on the blackboard.
Let a number n be written on the blackboard. Note that 7n + 7 ≡ n + 1
(mod 3) and 4n3 + 3n + 4 ≡ n3 + 1 ≡ n + 1 (mod 3). Hence all numbers
that can appear on the blackboard on the next move are congruent to n + 1
modulo 3, whence the residues repeat in cycle 1, 2, 0, 1, 2, 0, . . .. As 109 ≡ 1
(mod 3), the number 109 is written by the third player C.
Remark: The solution shows that C can win with whatever strategy.
F4 (Grade 10.) Let a, b, c, d be positive real numbers satistfying the sys-
tem of equations

 a2 + 12 = 12 ,

 b
 b2 + 4 = 8,

c2


 c2 + 16
d2
= 2,
d + a42 = 32.
 2


Determine the product abcd.

Answer: 4.

29
Solution 1: Multiplying all equations gives
    
2 1 2 4 2 16 4
a + 2 b + 2 c + 2 d + 2 = 28 .
2
b c d a
By AM-GM, a2 + b12 > 2 · ba , where the equality holds if and only if a = 1b .
Similarly b2 + c42 > 4 · bc (equality if and only if b = 2c ), c2 + 16d2
> 8 · dc
4 2 4 d
(equality if and only if c = d ) and d + a2 > 4 · a (equality if and only if
d = 2a ). Multiplying the obtained four inequalities gives
    
2 1 2 4 2 16 4 a b c d
a + 2 b + 2 c + 2 d + 2 > 2 · · 4 · · 8 · · 4 · = 28 .
2
b c d a b c d a
The resulting inequality must hold as equality by the first step of the solu-
tion. This is possible only if all four inequalities hold as equalities, whence


 a = 1b ,

 b = 2,

c
4


 c = d,
2

d = a.


16
By multiplying the equations of this system, we get abcd = abcd , whence
( abcd)2 = 16. As a, b, c and d are positive, the only possibility is abcd = 4.
Solution 2: By introducing a = 2x , b = 2y, c = z, d = 4t, rewrite the system
as  x2
 4 + 4y12 = 12 ,


4y2 + z42 = 8,


16


 z2 + 16t 2 = 2,
4·4

 2
16t + x2 = 32.
Multiplying the first equation by 4, the second equation by 14 , and the fourth
1
equation by 16 , we obtain an equivalent system
x2 + y12 = 2,




y + z12 = 2,
 2


 z2 + t12 = 2,



t + x12 = 2.
 2

Adding the equations of this system gives

1 1 1 1
x2 + 2 + y2 + 2 + z2 + 2 + t2 + 2 = 8.
y z t x
But for every real number u, u + u1 > 2, where equality holds only if u =
1. By adding up inequalities x2 + x12 > 2, y2 + y12 > 2, z2 + z12 > 2 and

30
 1
t2 + t2
> 2, we get
1 1 1 1
x2 + 2
+ y2 + 2 + z2 + 2 + t2 + 2 > 8.
x y z t
This inequality must hold as equality by the above; hence all four previous
inequalities must hold as equalities, too, i.e., x2 = y2 = z2 = t2 = 1. As
the numbers are positive, the only possibility is x = y = z = t = 1. Hence
abcd = 2x · 2y · z · 4t = 4xyzt = 4.
Remark: This problem can also be solved by finding the values of a, b, c,
d using standard techniques (substituting variables from one equation to
another) and directly computing abcd.
F5 (Grade 10.) Find all positive integers k for which there√is a right trian-
gle with legs of integral lengths and hypotenuse of length 88 . . . 822 . . . 2,
where the number under the root consists of exactly k eights and exactly k
twos.
Answer: 1.
Solution 1: Let the lengths of legs be a and b. By the Pythagorean theorem,
a2 + b2 = 88 . . . 822 . . . 2.
If k = 1 then one can choose a = 9 and b = 1 as 92 + 12 = 82. If k > 2 then
88 . . . 822 . . . 2 ≡ 6 (mod 8) since 822 ≡ 6 (mod 8) and 222 ≡ 6 (mod 8).
On the other hand, the residues of a2 and b2 modulo 8 can be 0, 1, or 4.
However, the sum of such two numbers can have residue 0, 1, 2, 4, or 5
modulo 8. Consequently, a2 + b2 = 88 . . . 822 . . . 2 cannot hold for k > 2.
Solution 2: Let the lengths of legs be a and b. By the Pythagorean theorem,
a2 + b2 = 88 . . . 822 . . . 2.
If k = 1 then one can choose a = 9 and b = 1 as 92 + 12 = 82. Assume
in the rest that k > 2. It is known that a positive integer is expressible as
the sum of two squares of integers if and only if its canonical representa-
tion contains primes congruent to 3 modulo 4 only with even exponents.
Note that 88 . . . 822 . . . 2 = 2 · 44 . . . 411 . . . 1, where only the odd second
factor can be divisible by primes congruent to 3 modulo 4. If all such
primes would occur with even exponents in the canonical representation
of 44 . . . 411 . . . 1, this number itself would be congruent to 1 modulo 4, but
actually 44 . . . 411 . . . 1 ≡ 3 (mod 4). This shows that right triangles with
the required property cannot exist in the case k > 2.
F6 (Grade 10.) Let D and E be the midpoints of sides AB and AC, re-
spectively, of a triangle ABC. Prove that the line AB is tangent to the cir-
cumcircle of the triangle BEC if and only if the line AC is tangent to the
circumcircle of the triangle BED.
Solution 1: The conditions of the problem imply that DE is the midsegment
parallel to the side BC of the triangle ABC (Fig. 30). By properties of in-
scribed angle, the line AB is tangent to the circumcircle of the triangle BEC
31
 A

D
E

B
C

Fig. 30

if and only if ∠ECB = ∠ DBE, and the line AC is tangent to the circumcircle
of the triangle BED if and only if ∠ DBE = ∠ AED. But ∠ECB = ∠ AED
since the lines DE and BC are parallel, whence validity of either of these
two equalities implies validity of the other one.
Solution 2: By powers, the line AB is tangent to the circumcircle of the tri-
angle BEC if and only if AB2 = AE · AC, and the line AC is tangent to
the circumcircle of the triangle BED if and only if AE2 = AD · AB. As
AB = 2AD and AC = 2AE, the equality AB2 = AE · AC is equivalent to
the equality 2| AD |2 = | AE|2 , as well as the equality AE2 = AD · AB is
equivalent to the equality AE2 = 2AD2 . Hence both conditions reduce to
the same equality, which solves the problem.
F7 (Grade 10.) Natural numbers 1 through n are written on a black-
board. On each move, one erases from the blackboard 2 or more numbers
whose sum is divisible by any of the chosen numbers and writes their sum
on the blackboard. Two players make moves by turns and the player who
cannot move loses the game. Which player can win the game against any
play by the opponent, if:
(a) n = 6;
(b) n = 11?
Answer: (a) The first player; (b) The first player.
Solution:
(a) The first player can replace numbers 1, 2, 3, 6 with 12. After that,
the blackboard contains numbers 4, 5, 12. In this state, the sum of no two
or three numbers on the blackboard is divisible by all the added numbers.
Thus the second player cannot move and the first player wins immediately.
(b) The first player can replace numbers 1, 2, 3, 4, 6, 8 with 24. After
that, the blackboard contains numbers 5, 7, 9, 10, 11, 24, which sum up
to 66. Among numbers 7, 9, 10, 11, 24, the l.c.m. of any two numbers is
greater than 66. Thus when choosing two or more numbers from among
the mentioned numbers, and perhaps also the number 5, the sum of the
chosen numbers is less than their l.c.m. and cannot be divisible by all of
them. Also when choosing one of the mentioned numbers together with 5,
the sum of the chosen numbers is not divisible by the larger one. Hence the

32
second player cannot move and the first player wins immediately.
F8 (Grade 11.) Prove that
1
sin 10◦ · cos 20◦ · sin 30◦ · cos 40◦ · sin 50◦ · cos 60◦ · sin 70◦ · cos 80◦ =
.
256
Solution 1: As sin 10◦ = cos 80◦ , sin 30◦ = cos 60◦ , sin 50◦ = cos 40◦ , and
sin 70◦ = cos 20◦ , the desired equality is equivalent to
1
(cos 20◦ · cos 40◦ · cos 60◦ · cos 80◦ )2 = .
256
It is known that cos 60◦ = 12 . Concerning the other factors, we obtain
8 sin 20◦ cos 20◦ ·cos 40◦ ·cos 80◦
cos 20◦ · cos 40◦ · cos 80◦ = 8 sin 20◦
4 sin 40◦ cos 40◦ ·cos 80◦
= 8 sin 20◦
2 sin 80◦ cos 80◦
= 8 sin 20◦
sin 160◦ sin 20◦ 1
= 8 sin 20◦ = 8 sin 20◦ = 8 .
Consequently,
 2  2
1 1 1 1
(cos 20◦ · cos 40◦ · cos 60◦ · cos 80◦ )2 = · = = .
2 8 16 256
Solution 2: As sin 10◦ = cos 80◦ , sin 30◦ = cos 60◦ , sin 50◦ = cos 40◦ , and
sin 70◦ = cos 20◦ , the l.h.s. of the desired equality can be rewritten as
(cos 20◦ cos 40◦ cos 60◦ cos 80◦ )2 . Hence the desired equality is equivalent
to
1
cos 20◦ cos 40◦ cos 60◦ cos 80◦ = . (2)
√ 16
As cos 60◦ = 21 and sin 60◦ = 23 , we have
cos 40◦ cos 80◦
= cos (60◦ − 20◦ ) cos (60◦ + 20◦ )
= (cos 60◦ cos 20◦ + sin 60◦ sin 20◦ ) (cos 60◦ cos 20◦ − sin 60◦ sin 20◦ )
= cos2 60◦ cos2 20◦ − sin2 60◦ sin2 20◦
= 14 cos2 20◦ − 34 sin2 20◦ = 14 cos2 20◦ − 3 sin2 20◦ .

Thus
1 3 ◦ 
cos 20◦ cos 40◦ cos 60◦ cos 80◦ = cos 20 − 3 sin2 20◦ cos 20◦ . (3)
8
By applying the formula cos 3x = cos3 x − 3 sin2 x cos x to x = 20◦ , we get
cos3 20◦ − 3 sin2 20◦ cos 20◦ = cos 60◦ = 21 . Consequently, (3) reduces to (2),
completing the solution of the problem.
F9 (Grade 11.) The bisector of the internal angle on vertex A of a triangle
ABC intersects the side BC at point D. The line tangent to the circumcircle
of the triangle ABC at point A intersects the line BC at point K. Prove that
KA = KD.
33
 A
A
Q
K B M
K C
B D
C O
D

X
Fig. 31 Fig. 32
Solution 1: Assume w.l.o.g. that ∠ ABC > ∠ ACB (Fig. 31; otherwise change
the roles of points B and C). Note that
∠ ADK = 180◦ − ∠CDA = ∠ DAC + ∠ ACD = ∠ BAD + ∠ ACB.
By inscribed angle property, ∠KAB = ∠ ACB, whence
∠ BAD + ∠ ACB = ∠ BAD + ∠KAB = ∠KAD.
Consequently, ∠ ADK = ∠KAD, which implies KA = KD.
Solution 2: Let O be the circumcenter of the triangle ABC, M be the mid-
point of the side BC, and X be the point of intersection of the ray OM with
the circumcircle of the triangle ABC. Moreover, let Q be the point of in-
tersection of the line perpendicular to the side BC and passing through
point D with the line AO (Fig. 32). As OM is the perpendicular bisector
of the side BC, point X bisects the arc BC of the circumcircle of the triangle
ABC, whence the line AD also passes through X. As OA = OX, we have
∠XAO = ∠OXA. On the other hand, OX ⊥ BC and QD ⊥ BC together
imply OX k QD, whence ∠OXA = ∠QDA. Thus ∠ DAQ = ∠XAO =
∠QDA, implying AQ = QD. Consequently, there exists a circle with cen-
ter Q passing through both points A and D. As KA ⊥ AQ and KD ⊥ DQ,
the lines KA and KD are tangent to this circle. By the property of tangent
line segments, KA = KD.
F10 (Grade 11.) Prove that in every solved sudoku X X XX
the squares marked “X” in the figure contain precisely X X XX
YYYYY
the same digits as the squares marked “Y”, counting Y Y
repetitions. (For instance, if the squares marked “X” Y Y
contain three digits 3 in total then the squares marked Y Y
YYYYY
“Y” also contain three digits 3 in total.) XX XX
XX XX

Remark: A solved sudoku is a board of size 9 × 9 whose every row, every

column and every part of size 3 × 3 bounded by bold rules contains every
digit from 1 to 9 exactly once.
Solution: As digits 1 through 9 occur exactly once in every row, every col-
34
 Fig. 33 Fig. 34 Fig. 35

umn and every part of size 3 × 3 bounded by bold rules, each of these digits
occurs exactly 4 times in the squares colored in Figures 33 and 34, if digits
in green squares are counted twice.
Removing from both (multi)sets of digits those digits that occur in squares
colored in Figure 35, we obtain precisely the sets of digits that occur in
squares marked “X” and “Y”, respectively. Thus these sets must also equal.
F11 (Grade 11.) Every term of the sequence a1 , a2 , a3 , . . . is either 0 or 1.
It is known that both 0 and 1 occur at least 1010 times among every 2021
consecutive terms of the sequence. May one be sure that the sequence is
periodic from some place on, i.e., there exist positive integers n and p such
that an+i = an+i+ p for every natural number i?
Answer: No.
Solution: Consider the tuples 11 . . . 1} 00
| {z . . . 0} and 11
| {z . . . 1} 00
| {z . . . 0} . Concate-
| {z
1011 times 1010 times 1010 times 1011 times
nating infinitely many instances of these tuples in any order produces a
sequence that satisfies the conditions of the problem. Indeed, consider any
segment of 2021 consecutive terms of such sequence. As any two consec-
utive full blocks of zeros or ones contain at least 2020 terms in total, the
segment under consideration contains terms of at most three such blocks.
If it contains terms of three blocks then it contains the middle block fully.
If the segment contained only two partial blocks of zeros or ones, it could
contain at most 2020 terms in total. Hence the segment under consideration
must contain one full block even if it contains only terms of two consecu-
tive blocks. In any case, the digit of the full block occurs either 1010 or 1011
times and the other digit must occur either 1011 or 1010 times, respectively.
Combining the tuples defined at the beginning of the solution according
to some non-periodic pattern (e.g., xyxxyxxxyxxxxy . . . ), the resulting se-
quence is not periodic from any place on. Indeed, suppose the contrary; let
the length of the period be p. We can find a pattern of length lcm( p, 2021)
starting from the place where periodicity starts, containing the tuples de-
fined at the beginning of the solution a full number of times. But the middle
term of these tuples is not repeating periodically, implying that the entire
pattern also cannot be periodic. Consequently, there exist non-periodic se-
quences that satisfy the conditions of the problem.

35
 F12 (Grade 12.) Find all pairs ( a, b) of positive integers such that a > b
and
1 1 1
+ = .
a b 2021
Answer: (2021 · 2022, 2022), (2021 · 48, 43 · 48), (2021 · 44, 47 · 44), (2021 ·
2, 2021 · 2), (47 · 90, 43 · 90).
Solution 1: Let d = gcd( a, b), a = da0 , b = db0 . The given equation reduces
to da1 0 + db1 0 = 2021
1
which is equivalent to
2021( a0 + b0 ) = da0 b0 . (4)
As a0 and b0are relatively prime to each other, they both are relatively prime
to a0 + b0 , implying that a0 b0 and a0 + b0 are relatively prime. By (4), a0 b0 |
2021( a0 + b0 ), implying a0 b0 | 2021. As 2021 = 43 · 47 where the factors are
prime, the number 2021 has exactly four positive factors 1, 43, 47 and 2021.
Taking into account that a > b holds if and only if a0 > b0 , consider all cases:
• If a0 = 2021 and b0 = 1 then (4) implies d = 2022. Consequently,
( a, b) = (2021 · 2022, 2022).
• If a0 = 47 and b0 = 43 then (4) implies d = 90. Consequently, ( a, b) =
(47 · 90, 43 · 90).
• If a0 = 47 and b0 = 1 then (4) implies d = 43 · 48. Consequently,
( a, b) = (2021 · 48, 43 · 48).
• If a0 = 43 and b0 = 1 then (4) implies d = 47 · 44. Consequently,
( a, b) = (2021 · 44, 47 · 44).
• If a0 = 1 and b0 = 1 then (4) implies d = 2021 · 2. Thus ( a, b) =
(2021 · 2, 2021 · 2).
Solution 2: The given equation is equivalent to aab +b = 1 which reduces to
2021
ab − 2021a − 2021b = 0. After adding 20212 to both sides and factorizing in
the l.h.s, we get
( a − 2021)(b − 2021) = 20212 . (5)
As both a and b are positive, the factors in the l.h.s. of (5) are greater than
−2021. Thus if both factors were negative then the absolute value of their
product would be less than 20212 and (5) could not hold. Hence both fac-
tors are positive. Since a > b, we must have a − 2021 > b − 2021. As
2021 = 43 · 47 with factors being prime, we obtain variants ( a − 2021, b −
2021) = (20212 , 1), ( a − 2021, b − 2021) = (2021 · 47, 43), ( a − 2021, b −
2021) = (2021 · 43, 47), ( a − 2021, b − 2021) = (472 , 432 ) and ( a − 2021, b −
2021) = (2021, 2021). Consequently, ( a, b) = (2021 · 2022, 2022), ( a, b) =
(2021 · 48, 43 · 48), ( a, b) = (2021 · 44, 47 · 44), ( a, b) = (47 · 90, 43 · 90), or
( a, b) = (2021 · 2, 2021 · 2).
F13 (Grade 12.) Anna, Anne and Anni seek for real solutions ( x, y) to the

36
system of equations
4x3 y − x4 − 3x2 y2 = 2021,


4y3 x − y4 − 3y2 x2 = 2021.

Anna claims that the system of equations has a solution. Anne claims that
the system of equations has no solution but at least one of the two equations
has solutions. Anni claims that the system of equations has no solution and,
even worse, neither of the two equations alone has a solution. Who is right?
Answer: Anne.
Solution: The system does not have a solution since adding the equation
gives −( x − y)4 = 4042 whose l.h.s. is non-positive but r.h.s. is positive.
We show that the first equation 4x3 y − x4 − 3x2 y2 = 2021 has solutions (the
same could be done for the second equation by symmetry). Dividing the
equation by x4 and reordering the terms in the l.h.s. results in
 y 2 y 2021
−3 +4 −1 = 4 . (6)
x x x
As the discriminant of the quadratic equation −3t2 + 4t − 1 = 0 is positive,
there exists a real number
q t such that −3t2 + 4t − 1 equals a positive num-
ber ε. Define x = 4 2021ε and y = xt; then x and y satisfy (6) and also the
first equation of the given system of equations. Hence Anne is right.
F14 (Grade 12.) Point D inside an acute triangle ABC satisfies
∠ ADC = ∠ BDA = 180◦ − ∠CAB.
Prove that the point symmetric to point A w.r.t. point D lies on the circum-
circle of the triangle ABC.
Solution 1: Let the line AD intersect the circumcircle of the triangle ABC
second time at point D 0 ; by the assumptions, ∠CDD 0 = ∠ D 0 DB = ∠CAB
(Fig. 36). We show that AD = DD 0 .
As the quadrilateral ABD 0 C is cyclic, ∠ ABC = ∠ AD 0 C = ∠ DD 0 C and
∠ BCA = ∠ BD 0 A = ∠ BD 0 D. Thus the triangled ABC, DD 0 C and DBD 0 are
| DD 0 | | DC |
similar. Hence | DB|
= | DD 0 |
, implying | DD 0 |2 = | BD | · |CD |.

D
A D′

C
Fig. 36

37
 B
B

D
A D′
O
A A′
D D′

C C
Fig. 37 Fig. 38

By similarity of the triangles ABC and DD 0 C, we obtain ∠ BCA = ∠ D 0 CD,

which implies ∠ DCA = ∠ D 0 CB = ∠ D 0 AB = ∠ DAB. By similarity of
the triangles ABC and DBD 0 we analogously get ∠ ABD = ∠CBD 0 =
∠CAD 0 = ∠CAD. Hence the triangles ABD and CAD are similar. Con-
AD BD
sequently, CD = AD , which implies AD2 = BD · CD.
Altogether, we have proven AD2 = DD 02 , which implies the desired result.
Solution 2: Let the line AD intersect the circumcircle of the triangle ABC
second time at point D 0 ; by the assumptions, ∠CDD 0 = ∠ D 0 DB = ∠CAB.
We show that AD = DD 0 .
Let O be the circumcenter of the triangle ABC. Let the line AD intersect the
circumcircle of the triangle BCD second time at point A0 (Fig. 37). By the
equality ∠CDA0 = ∠ A0 DB, the arcs BA0 and A0 C of the circumcircle of the
triangle BCD are equal.
By the conditions of the problem, ∠CDB = 2∠CAB. Hence point O lies on
the circumcircle of the triangle BCD. Since OC = OB, the arcs CO and OB
of the circumcircle of the triangle BCD are equal.
Consequently, OA0 is a diameter of the circumcircle of the triangle BCD.
If O 6= D then, by Thales’ theorem, ∠ A0 DO = 90◦ . Thus OD ⊥ AD 0 ,
which implies AD = DD 0 since a radius perpendicular to a chord bisects
the chord. If O = D (Fig. 38) then AD 0 is a diameter of the circumcircle of
the triangle ABC, obviously bisected by the center O.
F15 (Grade 12.) There are 6 distinct lines chosen in the space. Find the
largest possible number of points where at least 3 chosen lines intersect.
Answer: 4.
Solution 1: Call a point rich if at least 3 chosen lines intersect there. Let the
number of rich points be n. We count pairs ( P, s) where P is a rich point
and s is a chosen line that passes through P. The number of these pairs is at
least 3n since each of the n rich points occurs in at least 3 pairs. On the other
hand, after fixing a chosen line, the other lines passing through distinct rich
points on the fixed line are distinct (two lines
j cannot
k intersect in more than
6−1
one point). Hence there can be at most 2 = 2 rich points on every

38
 Fig. 39 Fig. 40

chosen line. This implies that the number of pairs under consideration does
not exceed 6 · 2, i.e., 12. Altogether, we have established 3n 6 12, which
implies n 6 4.
A configuration with 4 rich points can be obtained by choosing 6 lines de-
termined by the 6 edges of a tetrahedron; those lines intersect at 4 vertices
of the tetrahedron, 3 in each one (Figure 39).
Solution 2: Call a point rich if at least 3 chosen lines intersect there. Let the
number of rich points be n. We count pairs ( P, p) where P is a rich point
and p is a pair of distinct chosen lines both passing through P (the order
of lines in the pair is irrelevant). The number of such pairs is at most 15
since C62 = 15 and every two lines intersect in at most one rich point. On
the other hand, the number of these pairs must be at least 3n, as at least 3
chosen lines intersect at every rich point and these give rise to C32 = 3 pairs.
Consequently, 3n 6 15 which implies n 6 5. In the case n = 5, all inequali-
ties in the argument above should hold as equalities, i.e., every two chosen
lines intersect in a rich point and in every rich point exactly 3 chosen lines
intersect. Thus every chosen line must intersect at rich points with exactly 5
chosen lines whereby at every rich point with exactly 2 of them, which is
impossible as 5 is not divisible by 2. Consequently, n 6 4.
To find a configuration with 4 rich points, choose arbitrarily 4 points, no 3 of
which lie on one line. Pairs of these points determine 6 distinct lines. Every
chosen point belongs to 3 distinct pairs, whence 3 such lines pass through
every chosen point (in Fig. 40, rich points are marked with bubbles).
Solution 3: Call a point rich if at least 3 chosen lines intersect there. As in
Solution 1 or Solution 2, we show that 4 rich points is possible. To prove
that there can be no more rich points, let n be the number of rich points in
a configuration that satisfies the conditions of the problem with the largest
possible number of rich points; by the above, we may assume that n > 4.
Consider arbitrary two distinct rich points in this configuration and let A
and B be the corresponding sets of chosen lines that meet in these points.

39
The sets A and B cannot have two or more common members as two lines
cannot intersect in two distinct points. Suppose that the sets A and B are
disjoint. As both A and B contain at least 3 chosen lines and the total num-
ber of chosen lines is 6, every chosen line must belong to either A or B.
Let C be the set of chosen lines intersecting in a third rich point. As the
set C contains at least 3 lines and each of them belongs to either A or B, at
least 2 lines of the set C belong to one of A and B. But then C would have 2
common members with this set which we saw is impossible. Consequently,
every two rich points are connected with a chosen line. As 3 rich points
cannot lie on one line since this would require 3 · 2 + 1 = 7 chosen lines,
lines that connect different pairs of rich points are distinct. Thus there must
be at least Cn2 chosen lines, i.e., Cn2 6 6. Consequently, n 6 4.
F16 (Grade 12.) Two regular polygons have a common circumcircle. The
sum of the areas of the incircles of these polygons equals the area of their
common circumcircle. Find all possibilities of how many vertices can the
two polygons have.
Answer: 3 and 6 or 4 and 4.
Solution: The ratio of the inradius and the circumradius of a regular n-gon
◦
is cos 180
n . Hence the ratio of the areas of the incircle and the circumcircle
◦
of a regular n-gon is cos2 180
n .
Let a regular n-gon and a regular m-gon with a common circumcircle be
given. W.l.o.g., let n 6 m and the area of the common circumcircle be 1.
◦
By the above, the areas of the incircles of these polygons are cos2 180n and
◦
cos2 180
m , respectively. As their sum must equal the area of the common cir-
◦ 2 180◦ = 1 which is equivalent
cumcircle, we get the equation cos2 180
n + cos m
◦ 2 180◦ 180◦
to cos2 180n = sin m . As both n and m are larger than 2, both n and
180 ◦ ◦ 180 ◦ 180◦
m are less than 90 , whence the equation reduces to cos n = sin m .
◦ 180◦ ◦
Thus 180 1 1 1
n + m = 90 , implying that n + m = 2 . If n = 3 then m = 6; if
n = 4 then m = 4; if n > 4 then m < 4, contradicting the assumption n 6 m.

Selected Problems from the IMO Team Selection

Contests
S1 Juku has the first 100 volumes of the Harrie Totter book series at his
home. For every i and j, where 1 6 i < j 6 100, call the pair (i, j) reversed if
volume No j is before volume No i on Juku’s shelf. Juku wants to arrange
all volumes of the series to one row on his shelf in such a way that there
does not exist numbers i, j, k, where 1 6 i < j < k 6 100, such that pairs
(i, j) and ( j, k) are both reversed. Find the largest number of reversed pairs
that can occur under this condition.
Answer: 2500.
40
Solution 1: Let all 100 volumes be placed to a shelf a in some order. For
every i = 1, 2, . . . , 100, let ai be the number of the volume that occurs as
the ith in this order. In the order the volumes occur on shelf a, we start
relocating the volumes to two new shelves b and c. We place a volume
to the end of shelf b if, as the intermediate result, all volumes on shelf b
would be increasingly sorted by volume numbers. Otherwise, we place the
volume to the end of shelf c if, as the intermediate result, all volumes on
shelf c would be increasingly sorted by volume numbers. Continuing this
way, we either can relocate all volumes onto two shelves in such a way that
all volumes on either shelf are increasingly sorted by volume numbers or
get stuck on some step k of the process because ak is less than the number of
the last volume on both new shelves. In the last case, let the number of the
last volume on shelf c be a j ; then, by assumptions, j < k and a j > ak . Since
volume No a j has been relocated to shelf c, some volume with number ai
must occur on shelf b such that i < j and ai > a j . This means that pairs (i, j)
and ( j, k) were initially reversed. Hence we can conclude that, in the case
of Juku’s favourite orderings, all volumes can be relocated to two shelves,
i.e, there exist two tuples ai1 , ai2 , . . . , ais and a j1 , a j2 , . . . , a jt with s + t = 100,
where i1 < . . . < is , ai1 < . . . < ais and j1 < . . . < jt , a j1 < . . . < a jt .
The number of pairs (i, j) such that i < j and the numbers ai and a j are in
distinct tuples is exactly st. Each reversed pair (i, j) must be one of these
st pairs. Thus the number of reversed pairs does not exceed st. As st 6
s+t 2
= 502 = 2500, the number of reversed pairs cannot exceed 2500.


2
The number 2500 is achieved by the order 51, 52, . . . , 100, 1, 2, . . . , 50.
Solution 2: Let all 100 volumes be placed to the shelf in some order. Con-
sider the graph whose vertices are numbers 1, 2, . . . , 100 and an edge oc-
curs between vertices i and j if and only if either (i, j) or ( j, i ) (depend-
ing on whether i < j or j < i) is reversed. Suppose that the graph con-
tains an odd cycle. Such cycle must contain three consecutive vertices i,
j and k that are in either increasing or decreasing order; w.l.o.g., assume
that i < j < k. Then pairs (i, j) and ( j, k ) are both reversed. Thus un-
der the conditions of the problem, the graph cannot contain an odd cy-
cle. Hence the graph is bipartite. Let the numbers of vertices in the two
parts be s and t, respectively; then the maximal number of edges is st.
t 2
By AM-GM, st 6 s+ = 502 = 2500. If the volumes are in the order


2
51, 52, . . . , 100, 1, 2, . . . , 50, exactly 2500 reversed pairs arise indeed.
Remark: The problem can be solved even more immediately with Mantel’s
theorem that states that the maximum j k number of edges in a graph that does
n2
not contain cycles of length 3 is 4 , where n is the number of vertices.
S2 Find all polynomials P( x ) with integral coefficients whose values at
points x = 1, 2, . . . , 2021 are numbers 1, 2, . . . , 2021 in some order.
Answer: All polynomials of the form P( x ) = x + R( x )( x − 1)( x − 2) . . . ( x −

41
2021) and the form P( x ) = 2022 − x + R( x )( x − 1)( x − 2) . . . ( x − 2021),
where R( x ) is an arbitrary polynomial with integral coefficients.
Solution: Since P( x ) has integral coefficients, it satisfies
a − b | P( a) − P(b) (7)
for arbitrary integers a and b. By substituting a = 2021 and b = 1 into (7),
we get 2020 | P(2021) − P(1). As P(1), . . . , P(2021) must be 2021 distict
numbers, P(1) and P(2021) cannot be equal, whence their difference must
be at least 2020. The only possibility is P(1) and P(2021) being 1 and 2021
in some order. Consider these two cases separately.
• If P(1) = 1 and P(2021) = 2021 then substituting a = 2020 and
b = 1 into (7) gives 2019 | P(2020) − 1 that can be satisfied only if
P(2020) = 2020. Similarly by substituting a = k and b = 1 into (7) for
k = 2019, 2018, . . . , 2, we each time get P(k) = k as the only possibil-
ity, since all larger values of the polynomial that could satisfy k − 1 |
P(k) − 1 are in use already. Hence P( x ) = x for every x = 1, . . . , 2021.
As 1, . . . , 2021 are roots of the polynomial P( x ) − x, we must have
P( x ) − x = ( x − 1) . . . ( x − 2021) R( x ), where R( x ) can be any poly-
nomial with integral coefficients. From here, we obtain solutions of
the form P( x ) = x + R( x )( x − 1)( x − 2) . . . ( x − 2021).
• If P(1) = 2021 and P(2021) = 1 then by substituting a = k and b =
2021 into (7) for k = 2, . . . , 2020, we each time get 2021 − k | P(k ) − 1
that can be satisfied only if P(k) = 2022 − k analogously to the pre-
vious case. Hence P( x ) = 2022 − x for every x = 1, . . . , 2021. As
1, . . . , 2021 are roots of the polynomial P( x ) − (2022 − x ), we must
have P( x ) − (2022 − x ) = ( x − 1) . . . ( x − 2021) R( x ), where R( x ) is
any polynomial with integral coefficients. This leads to solutions of
the form P( x ) = 2022 − x + R( x )( x − 1)( x − 2) . . . ( x − 2021).
Both families of solutions meet the required conditions: P(1), . . . , P(2021)
are 1, . . . , 2021, respectively, in the first case and 2021, . . . , 1, respectively, in
the second case.
S3 (a) There are 2n rays marked in a plane, with n being a natural
number. Given that no two marked rays have the same direction and no
two marked rays have a common initial point, prove that there exists a
line that passes through none of the initial points of the marked rays and
intersects with exactly n marked rays.
(b) Would the claim still hold if the assumption that no two marked rays
have a common initial point was dropped?
Answer: (b) Yes.
Solution: Consider any circle such that all of the endpoints of the marked
rays are inside the circle. Choose a tangent line of the circle that is not par-
allel to any of the marked rays. Let this tangent line be l0 and let lα be the
tangent line we get by rotating l0 counterclockwise by angle α w.r.t. the
42
center of the circle. For any α, let f (α) be the number of marked rays that
intersect with lα . Since l0 and lπ are two parallel lines and all of the end-
points of the marked rays are between them, we know from the definition
of l0 that f (0) + f (π ) = 2n. W.l.o.g., let f (0) 6 n and f (π ) > n. Because no
two rays have the same direction, when α increases continuously f (α) can
at any time instance only change by at most one. Thus f (α) ranges over all
integral values between f (0) and f (π ). Hence f (α) = n for some α.
The argument above does not use the assumption that the initial points of
the marked rays are distinct, whence it solves both parts of the problem.
Remark: Part a) of the problem can be solved otherwise. Choose an arbitrary
line l that is not parallel to any of the marked rays or any line connecting
the initial points of two marked rays and from which the initial points of
all marked rays lie on one side. Let l 0 be a line parallel to l such that the
initial points of all marked rays lie between l and l 0 . Then every marked
ray intersects exactly one of the lines l and l 0 , whence the numbers of the
intersection points always sum up to exactly 2n. W.l.o.g., let the number
of intersection points on the line l be less than or equal to the number of
intersection points on the line l 0 . When shifting the line l towards the line l 0
while keeping its direction unchanged, the number of intersection points
on it can change by at most one at any time instance. Thus there exists a
position where the number of intersection points equals n.
S4 Positive real numbers a, b, c satisfy abc = 1. Prove that
a b c 3
+ + > .
1+b 1+c 1+a 2
Solution 1: The given inequality is equivalent to
2a(1 + c)(1 + a) + 2b(1 + a)(1 + b) + 2c(1 + b)(1 + c) > 3(1+a)(1+b)(1+c).
By removing parentheses and collecting similar terms, it is reduced to
2a2 + 2b2 + 2c2 + 2a2 c + 2b2 a + 2c2 b > 3 + a + b + c + ab + bc + ca + 3abc. (8)
Using AM-GM, we can make the following observations:
2 2 2 c2 √ √ √
c2 + a2
a2 + b2 + c2 = a +2 b + b +2 + 2 > a2 b2 + b2 c2 + c2 a2 = ab + bc + ca;
a2 + a2 + a2 + a2 + b2 + c2 b2 + b2 + b2 + b2 + c2 + a2 c2 + c2 + c2 + c2 + a2 + b2
a2 + b2 + c2 = 6 + 6 + 6
√
6
√
6
√
6 4 1 1 4 1 1 4 1 1
> a8 b2 c2 + b8 c2 a2 + c8 a2 b2 = a 3 b 3 c 3 + b 3 c 3 a 3 + c 3 a 3 b 3
1
= ( abc) 3 ( a + b + c);
√
a2 c + b2 a + c2 b 3
a2 c + b2 a + c2 b = 3 · 3 > 3 a3 b3 c3 = 3abc.
Taking into account that abc = 1, we obtain
2a2 + 2b2 + 2c2 +
2a2 c + 2b2 a + 2c 2b

= a2 + b2 + c2 + a2 + b2 + c2 + a2 c + b2 a + c2 b + a2 c + b2 a + c2 b




1
> ( ab + bc + ca) + ( abc) 3 ( a + b + c) + 3abc + 3abc
= 3 + a + b + c + ab + bc + ca + 3abc.
43
This proves (8), whence the desired result follows.
Remark: Instead of the second out of the three observations made in So-
lution 1, one can prove the inequality a2 + b2 + c2 > a + b + c using the
assumption abc = 1 as follows: Apply AM-GM 2 2 2
√ qto obtain a + b + c =
2 2 2 3 2 2 2
3 · a +b3 +c > 3 a2 b2 c2 = 3 and AM-HM to get a +b3 +c > a+3b+c ; then
q q
a2 + b2 + c2 = ( a2 + b2 + c2 ) ( a2 + b2 + c2 ) > 3 ( a2 + b2 + c2 ) > a + b + c.
The rest is analogous to Solution 1.
a2 2 2
Solution 2: The given inequality is equivalent to a(1+b)
+ b(1b+c) + c(1c+a) > 32 .

Applying the so-called Titu’s lemma to ( a, b, c) and a(11+b) , b(11+c) , c(11+a)
2 2 2 ( a + b + c )2 ( a + b + c )2
gives a(1a+b) + b(1b+c) + c(1c+ a) > a(1+b)+b(1+c)+c(1+ a) = (a+b+c)+(ab+bc+ca) .
( a + b + c )2
Hence it suffices to prove the inequality (a+b+c)+(ab+bc+ca) > 32 ; we prove
the equivalent inequality a+1b+c + ab +bc+ca 6 2 . By AM-GM, a + b + c >
( a + b + c )2 3
√3 1 1
3 abc = 3, implying a+b+c 6 3 . Hence it suffices to prove the inequality
ab+bc+ca
( a + b + c )2
6 13 . By removing parentheses and collecting similar terms, the
latter reduces to a2 + b2 + c2 > ab + bc + ca, which can be established by
applying AM-GM repeatedly (similarly to Solution 1).
S5 Find all polynomials P( x, y) with real coefficients which for all real
numbers x and y satisfy P( x + y, x − y) = 2P( x, y).
Answer: P( x, y) = ( a + b) x2 + axy + by2 , where a, b are any real numbers.
Solution: By the given equality,
P(2x,2y) = P(( x+y)+( x−y),( x+y)−( x−y)) = 2P( x+y,x−y) = 4P( x,y). (9)
As (9) holds for arbitrary real numbers x and y, we can consider it as an
equality of polynomials. Hence the corresponding coefficients are equal.
Let i and j be arbitrary non-negative integers. The coefficient of xi y j in
the l.h.s. of (9) is ai,j 2i 2 j or, equivalently, 2i+ j ai,j , whereas the coefficient of
the xi y j in the r.h.s. of (9) is 4ai,j . Hence 2i+ j ai,j = 4ai,j or, equivalently,
2i+ j − 4 ai,j = 0, which provides us two possibilities: either i + j = 2 or

ai,j = 0. Consequently, the only terms in the polynomial P( x, y) have expo-

nents of variables that sum up to 2. In other words, P( x, y) = cx2 + axy +
by2 for some coefficients a, b, c. After substituting into the initial equality
and collecting similar terms in both sides, we obtain
(c + a + b) x2 + 2(c − b) xy + (c − a + b)y2 = 2cx2 + 2axy + 2by2 .
Again, the coefficients in the l.h.s. and r.h.s. must be equal. From the coef-
ficients of x2 , we get c + a + b = 2c which implies c = a + b. This makes
also the corresponding coefficients of the other terms equal. Consequently,
all polynomials of the form P( x, y) = ( a + b) x2 + axy + by2 where a and b
are arbitrary real numbers satisfy the conditions of the problem.

44