IMO 1974
Read the problems at the official link. The 1974 IMO problems are Β© the International Mathematical Olympiad, so we don't reproduce them here. Open the 1974 problems (PDF) from the IMO's official site.
The topic labels, hints and solutions below are our own, written from scratch: our solutions, not official ones. If you spot a mistake, please use βReport a problemβ at the bottom of the page.
P1 Card-dealing puzzle: recover the first round from totals
- Hint 1. The grand total of counters is $39$. It is the number of rounds times $p+q+r$.
- Hint 2. Use the totals of $A$ and $B$ to pin down $r$, then $p$ and $q$.
Our solution (not official)
Answer: Player $C$ (the cards were $1,4,8$, dealt three times)
Key idea: the grand total fixes the number of rounds, and the two extreme players fix the cards.
Rounds. All counters together: $20+10+9=39=N(p+q+r)$, where $N\ge2$ is the number of rounds. Since $p+q+r\ge1+2+3=6$, the only option is $N=3$ and $p+q+r=13$.
The cards. Player $A$ got $20$ in three rounds, so $3r\ge20$ and $r\ge7$. Player $B$ got $r$ in the last round and at least $p$ in each of the others, so $r+2p\le10$ and $r\le8$.
- If $r=7$: then $p+q=6$, and $B$’s other two cards add up to $3$. With $p<q$ and $p+q=6$, the possible sums of two cards from $\{p,q\}$ are $2p$, $p+q=6$ and $2q$. None of these is $3$ ($2p=3$ is impossible). So $r\ne7$.
- So $r=8$, $p+q=5$, and $B$’s other two cards add up to $2$, so both are $p=1$. Then $q=4$.
Who got what. With cards $1,4,8$: $B$ got $1,1,8$ (the $8$ in round $3$). Player $A$’s $20$ must be $8+8+4$, so $A$ took the $8$ in rounds $1$ and $2$ and the $4$ in round $3$. Then $C$ got $4,4,1$, which is $9$, as required. In round $1$, $A$ had $8$, $B$ had $1$, so $C$ received $q$.
How we checked it: script: brute force over all cards and deals finds only p, q, r = 1, 4, 8 with C getting q first
P2 Cevian equal to the geometric mean of its base pieces
- Hint 1. Extend $CD$ to meet the circumcircle again at $E$. The power of $D$ gives $AD\cdot DB=CD\cdot DE$.
- Hint 2. So the condition says $D$ is the midpoint of $CE$. How far can a point $E$ of the arc be from line $AB$?
Our solution (not official)
Key idea: by the power of a point, the condition says that $D$ is the midpoint of a chord from $C$. That happens when the far arc reaches as far from $AB$ as $C$ does.
Let $R$ be the circumradius. For $D$ strictly inside segment $AB$, extend $CD$ to meet the circumcircle again at $E$, on the arc $AB$ not containing $C$. The power of $D$ gives $AD\cdot DB=CD\cdot DE$. So $$CD^2=AD\cdot DB\iff CD=DE\iff D\text{ is the midpoint of }CE .$$
Reformulate. The midpoint of $CE$ lies on line $AB$ exactly when $E$ and $C$ are at the same distance from that line (on opposite sides). And when it does, it lies inside segment $AB$, since chord $CE$ crosses chord $AB$ inside the circle. So a suitable $D$ exists if and only if some point $E$ of the far arc has distance $h_C$ from $AB$, where $h_C$ is the altitude from $C$.
Distances. Along the far arc (ends excluded), the distance to $AB$ takes every value in $(0,d]$, where $d$ is attained at the arc’s midpoint. The center $O$ is at signed distance $R\cos C$ from $AB$ on the side of $C$, so $d=R-R\cos C=2R\sin^2\frac C2$. And $h_C=b\sin A=2R\sin A\sin B$. So $D$ exists if and only if $$2R\sin A\sin B\le2R\sin^2\frac C2,\quad\text{that is,}\quad \sin A\sin B\le\sin^2\frac C2 .$$
How we checked it: script: 3,000 random triangles; a numeric search for D agrees with the criterion
P3 Binomial sum with powers of eight avoids multiples of five
- Hint 1. The sum is the coefficient of $\sqrt8$ in $(1+\sqrt8)^{2n+1}$.
- Hint 2. Multiply by the conjugate $(1-\sqrt8)^{2n+1}$ and look at the result modulo $5$.
Our solution (not official)
Key idea: the sum is the irrational part of $(1+\sqrt8)^{2n+1}$. A norm identity then turns divisibility by $5$ into a square congruence that is impossible.
Expand $(1+\sqrt8)^{2n+1}=X+Y\sqrt8$ with integers $X,Y$. The terms with odd powers of $\sqrt8$ give $$Y=\sum_{k=0}^n\binom{2n+1}{2k+1}8^k=\sum_{k=0}^n\binom{2n+1}{2k+1}2^{3k},$$ our number. Also $(1-\sqrt8)^{2n+1}=X-Y\sqrt8$. Multiplying, $$X^2-8Y^2=(1-8)^{2n+1}=-7^{2n+1}.$$ Suppose $5\mid Y$. Then modulo $5$, using $7\equiv2$ and $4\equiv-1$, $$X^2\equiv-7^{2n+1}\equiv-2\cdot4^n\equiv\mp2\pmod 5 .$$ But squares modulo $5$ are $0,1,4$, and $\pm2\equiv2,3$. Contradiction, so $5\nmid Y$.
How we checked it: script: the sum is not divisible by 5 for n = 0 to 499
P4 Chessboard cut into balanced rectangles of increasing size
- Hint 1. A rectangle has as many white as black squares exactly when its area is even. The numbers $a_i$ are distinct positive integers adding up to $32$.
- Hint 2. List the sets of seven distinct positive integers with sum $32$. Which value cannot be the half-area of a rectangle on the board?
Our solution (not official)
Answer: $p=7$, with $(1,2,3,4,5,7,10)$, $(1,2,3,4,5,8,9)$, $(1,2,3,4,6,7,9)$ or $(1,2,3,5,6,7,8)$
Key idea: each $a_i$ is half the area of a rectangle, the $a_i$ are distinct and add up to $32$, so there are at most seven. Of the five candidate lists, one needs a rectangle too long for the board, and the other four can be drawn.
Balance means even area. A rectangle with an even side splits into dominoes, each with one square of each color. A rectangle with both sides odd has odd area, so it cannot be balanced. So rectangle $i$ is balanced exactly when its area $2a_i$ is even, and the condition is that the areas $2a_1<\dots<2a_p$ add up to $64$, that is, $a_1+\dots+a_p=32$.
At most seven. Eight distinct positive integers add up to at least $1+2+\dots+8=36>32$. So $p\le7$.
Seven pieces. Seven distinct positive integers add up to at least $28$, so we add $4$ in total to $1,2,\dots,7$ while keeping them distinct. That gives exactly five lists: $$(1,2,3,4,5,6,11),\ (1,2,3,4,5,7,10),\ (1,2,3,4,5,8,9),\ (1,2,3,4,6,7,9),\ (1,2,3,5,6,7,8).$$ The first is impossible: a rectangle of area $22=2\cdot11$ would need a side of length $11$ or $22$. The other four lists all occur. The animation shows a cutting of the board for each one. All needed areas fit on the board: $2\times1$, $2\times2$, $2\times3$, $2\times4$, $2\times5$, $3\times4$, $2\times7$, $4\times4$, $3\times6$, $4\times5$.
How we checked it: script: listed all seven-element sets with sum 32, and found a tiling for each of the four valid lists by backtracking
P5 Range of a cyclic sum of four fractions
- Hint 1. For the lower bound, replace each denominator by $a+b+c+d$. For the upper bound, pair the first two terms and the last two.
- Hint 2. Find positive values where the sum is close to $1$ and others where it is close to $2$, then use continuity.
Our solution (not official)
Answer: Every number strictly between $1$ and $2$
Key idea: simple comparisons trap the sum strictly between $1$ and $2$, and continuity fills in everything between.
Call the sum $S$ and let $T=a+b+c+d$.
$S>1$: each denominator is less than $T$, so $S>\frac{a+b+c+d}{T}=1$.
$S<2$: shrinking a denominator increases a fraction, so $$\frac a{a+b+d}+\frac b{a+b+c}<\frac a{a+b}+\frac b{a+b}=1,\qquad \frac c{b+c+d}+\frac d{a+c+d}<\frac c{c+d}+\frac d{c+d}=1 .$$
Every value in between. With $(a,b,c,d)=(1,1,\varepsilon,\varepsilon)$, $$S=\frac{2}{2+\varepsilon}+\frac{2\varepsilon}{1+2\varepsilon}\to1\quad(\varepsilon\to0).$$ With $(a,b,c,d)=(1,\varepsilon,1,\varepsilon)$, $$S=\frac{2}{1+2\varepsilon}+\frac{2\varepsilon}{2+\varepsilon}\to2 .$$ $S$ is a continuous function on the connected set of positive quadruples, so its values form an interval. That interval lies in $(1,2)$ and comes arbitrarily close to both ends, so it is all of $(1,2)$.
How we checked it: script: 100,000 random quadruples give values from 1.0000004 to 1.9999993
P6 Integer polynomial hitting plus or minus one
- Hint 1. If $P(u)=1$ and $P(v)=-1$ for integers $u,v$, then $u-v$ divides $P(u)-P(v)=2$.
- Hint 2. So when both values occur, all the integers involved lie in a short interval. Treat degree $1$ and $2$ separately.
Our solution (not official)
Key idea: an integer where $P=1$ and one where $P=-1$ are at most $2$ apart, so if both kinds occur, at most five integers are involved. Small degrees are handled by counting roots.
Let $d=\deg P\ge1$. Let $U$ be the set of integers $k$ with $P(k)=1$ and $W$ the set with $P(k)=-1$. Then $n(P)=|U|+|W|$. The polynomials $P-1$ and $P+1$ have degree $d$, so $|U|\le d$ and $|W|\le d$.
If $U$ or $W$ is empty, then $n(P)\le d$.
If $d\le2$, then $n(P)\le2d\le d+2$.
If both are nonempty and $d\ge3$. For $u\in U$ and $w\in W$, the difference $u-w$ divides $P(u)-P(w)=2$ (because $P$ has integer coefficients), so $|u-w|\le2$. Let $m$ be the smallest element of $U\cup W$; by symmetry (replace $P$ by $-P$) say $m\in U$. Then every $w\in W$ satisfies $m<w\le m+2$. And every $u\in U$ is within $2$ of some $w$, so $u\le m+4$. So $U\cup W\subseteq\{m,\dots,m+4\}$, which gives $n(P)\le5\le d+2$.
In every case $n(P)-\deg P\le2$.
How we checked it: algebra only: divisibility of integer polynomials and root counts