1
$\begingroup$

I need help with the following puzzle:

Consider a round table which hosts $2n$ knights. At each seat, there is a namecard of one knight. The knights don't necessarily sit in front of their own namecard. Prove that we can rotate the table such that at least 2 knights sit in front of their own nameplate.

Using the pigeon hole principle, it's quite easy to see that this is equivalent with:

  • Proving that there exists a rotation where no knights sit in front of their own namecard.
  • Proving that it is impossible that in every rotation, exactly one knight sits in front of his own namecard.

After going over some examples, I noted that if for knight number $k$, the angle needed to get this knight in front of his nameplate is $\frac{a_k \pi}{n}$, then the sum $a_1 + ... + a_{2n}$ is a multiple of n. If one could prove this, then this would solve the problem: if in every rotation, there is exactly one person sitting in front of his namecard, then $$a_1 + .. . + a_{2n} = 0+1+...+ (2n-1) = n(2n - 1),$$ which is not a multiple of $2n$.

Another method is also welcome of course. Thank you in advance!

$\endgroup$
4
  • $\begingroup$ To clarify, what are you looking for? $\quad$ You've shared the standard solution to this puzzle. It can also be interpreted as an "expected value argument". $\endgroup$
    – Calvin Lin
    Commented Feb 28 at 15:49
  • $\begingroup$ Also, being very nitpicky about the presentation, you don't need the preamble of PHP. Your solution directly shows that the set of $ a_i $ contains duplicates, hence we could rotate by $-a_I$ to get the duplicate knights. $\endgroup$
    – Calvin Lin
    Commented Feb 28 at 15:53
  • $\begingroup$ I did not solve the puzzle since I did not prove yet that $a_1+...+a_{2n}$ is a multiple of $2n$. I just observed this phenomenon in examples. $\endgroup$
    – Steve
    Commented Feb 28 at 23:40
  • $\begingroup$ Ah, then just FYI that you had all of the elements needed to prove the problem, and just needed to put it together. I glossed over your "if one could prove this" and treated it as "based on what we've been discussing, the following holds". $\endgroup$
    – Calvin Lin
    Commented Feb 29 at 0:03

2 Answers 2

3
$\begingroup$

Let $1,2,..,,2n$ be the numbers of the knights clockwise. Let $a_i$ be the number of places we have to rotate a table counter-clockwise so that $i$-th knight sits in front of his namecard.

If $a_i=a_j$ for some $i\neq j$ then we’re done. Otherwise $$\sum a_i = 0+1+2+…+(2n-1)=n\cdot(2n-1).$$

Note that $i+a_i$ or $i+a_i-2n$ is the number of place where the namecard of the $i$-th knight is. It means that for some integer $k$ holds $$\sum (i+a_i)=1+2+…+2n+(2n)\cdot k.$$

So $$\sum a_i=(2n)\cdot k=n\cdot(2k).$$

But $n\cdot(2n-1)= n\cdot(2k)$ never holds. Contradiction.

$\endgroup$
3
  • 2
    $\begingroup$ You can simply the solution by stating that you're working in modulo $2n$. EG that gets rid of the "not necessarily true" part. $\endgroup$
    – Calvin Lin
    Commented Feb 28 at 15:58
  • $\begingroup$ I think you meant $i - a_i \mod 2n$ is the number of the knight sitting g at the nameplate of knight $i$ $\endgroup$
    – Steve
    Commented Feb 28 at 23:47
  • $\begingroup$ @Steve You’re right, but I’ve just changed the direction of table rotation from clockwise to counter-clockwise instead. $\endgroup$
    – Aig
    Commented Feb 29 at 0:53
1
$\begingroup$

Another way to present the same argument (I solved it like this when I saw it).


Label the cards around the table in clockwise order as $0,1,2 \ldots 2n-1$ and name the knights with their respective numbers. Let $f: \Bbb Z/(2n)\Bbb Z \to \Bbb Z/(2n)\Bbb Z$ be such that $f(k)$ is the position (i.e., the name card at his current position) of knight $k$.

Define $g: \Bbb Z/(2n)\Bbb Z \to \Bbb Z/(2n)\Bbb Z$ as $g(k) = (f(k) - k) \text{ mod } 2n$. FTSOC assume that $g$ is injective (and hence, bijective).

Next, note that the orbit of $f(k)$ is a cycle of length $l_k$, where $1\le l_k \le 2n$. Then

$$f(0) - 0 = g(0)$$ $$f^2(0) - f(0) = g(f(0))$$ $$\cdot \\ \cdot \\ \cdot$$ $$f^{l_0}(0) - f^{l_0-1}(0) = g(f^{l_0-1}(0))$$ Since $g$ is injective, the values in the RHS's must be distinct. Sum this, and note the LHS telescopes to give $0$, so the RHS must be $0 \text{ mod } 2n$. Do this for all the cycles, to get $$\sum_{k=0}^{2n-1} k \equiv \sum_{k=0}^{2n-1} g(k) \equiv 0 \pmod{2n}$$ a contradiction.


The uselessly technical terminology aside, note that the orbits of $f$ will always have such properties; it is the injectivity of $g$ that allows us to achieve a contradiction, and as such, all other methods must boil down to the injectivity of $g$.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .