You are on page 1of 5

I recently found two new "magic tricks" which are really just riddles; they are of similar structure,

content and cuteness, and though I cheated on the first one I was able to solve the more challenging
second one. The riddle I heard first is as follows:

Two magicians are to preform a magic trick with a deck of cards. One mathematician leaves the
room while the other has a member from the audience select five cards from the deck at random. He
then removes one of the cards, gives it back to the audience member, and sets the remaining four in
a pile face down on a table in an order of his choosing. The first mathematician then returns to the
room, looks at the cards, and from their order is able to deduce the identity of the card in the audi-
ence (suit and value, Ace, 2, 3, ... King). How is this trick done?

The second is a problem I found in a Russian mathematical Olympiad from 2007 (quoted here):

A conjurer Arutyun and his assistant Amayak are going to show following super-trick. A circle is
drawn on the board in the room. Spectators mark 2007 distinct points on this circle, after that
Amayak removes one of them. Then Arutyun comes to the room and shows a semicircle, to which
the removed point belonged. Explain, how Arutyun and Amayak may show this super-trick.

Laughable and unintentionally comical translators aside, this second problem proved quite interest-
ing. My instinct was naturally to start with 2 points on the circle and hope to either 1) develop a
recursive solution or 2) see a general one that would work for any number of points. I ended up
doing the latter, though I think there is an equivalent recursive way of looking at it. Before we get to
that, I'll go over the solution to the first problem, as well as my numerous failed attempts at solving
it.

My first guess was to give the suit of the unknown card by the top card in the pile: this can be done
because there are only 4 suits to go among 5 cards, and consequently at least two of the cards must
have the same suit. Then we just have to determine the value of the card, from among 13 possible
values, with the remaining 3 cards. This seems, at first glance, impossible; I quickly and mistakenly
assumed that not making use of the value of the first card was too great a loss of possible coding,
and thus abandoned that strategy. I went on to create elaborate codes using the 4 cards, noting some
useless points as well as one important one: if the mathematicians pre-agree on a certain ordering of
the 52 cards of the deck then they can both see the same "code" for any 4 cards in the following
manner. The one that is first in the ordering is given a 1, and so on, so that any of the 24 combina-
tions of the digits 1,2,3 and 4 may be communicated. I attempted to make use of the special cases of
the cards, i.e. where there are 2 of a kind, 3 of a kind, and so on, and use that to indicate the value.
That kind of information proved difficult and sometimes impossible to translate, making the solution
a sub 100% one.

The 100% solution is this: First, use the first card to indicate the suit. Then, using the same coding
technique among the remaining 3 cards, we can create 6 different codes: {{1,2,3}, {1,3,2}, ...
{3,2,1}}. The way we "double" that 6 to create 12 codes is by picking which of the two cards that
My first guess was to give the suit of the unknown card by the top card in the pile: this can be done
because there are only 4 suits to go among 5 cards, and consequently at least two of the cards must
have
2
the same suit. Then we just have to determine the value of the card, from among 13 possible
Magic Tricks.nb
values, with the remaining 3 cards. This seems, at first glance, impossible; I quickly and mistakenly
assumed that not making use of the value of the first card was too great a loss of possible coding,
and thus abandoned that strategy. I went on to create elaborate codes using the 4 cards, noting some
useless points as well as one important one: if the mathematicians pre-agree on a certain ordering of
the 52 cards of the deck then they can both see the same "code" for any 4 cards in the following
manner. The one that is first in the ordering is given a 1, and so on, so that any of the 24 combina-
tions of the digits 1,2,3 and 4 may be communicated. I attempted to make use of the special cases of
the cards, i.e. where there are 2 of a kind, 3 of a kind, and so on, and use that to indicate the value.
That kind of information proved difficult and sometimes impossible to translate, making the solution
a sub 100% one.

The 100% solution is this: First, use the first card to indicate the suit. Then, using the same coding
technique among the remaining 3 cards, we can create 6 different codes: {{1,2,3}, {1,3,2}, ...
{3,2,1}}. The way we "double" that 6 to create 12 codes is by picking which of the two cards that
had the same suit to keep (and thus making use of that card's value). We pick the one that is less than
or equal to 6 "less" than the other one, working mod 13. For example, given the 8 and 3 of hearts,
we keep the 3, since it is 5 less than 8; given the 8 and Ace (1) of hearts, we keep the 8 because it is 6
less than 1 (1-6 = -5 = 8 mod 13). We could have also picked the one that is less than or equal to 6
greater than the other one, it doesn't matter; all that does is that both mathematicians know which
direction they are working. Finally, we code the number our chosen card is less than the other one
with the remaining 3 cards. Once that is done, we hand over the pile; the other mathematician sees
the suit of the card, and knows to add some number to its value to obtain that of the unknown card.

The similar principle between this problem and the next is the crux of both: that there always exists
that "ordering" of the two numbers between 1 and 13 so that their difference has absolute value less
than or equal to 6. (The second problem uses the geometric analog of this discrete integer principle.)

So, how can Amayak and Arutyun preform this super trick? Similarly, it has many (even more, in
fact) very-close solutions: note first that just drawing a random semi-circle gives you a 50% chance,
and it isn't particularly hard to do better. However, the 100% solution is just that; it does require a
short lemma, to prove the existence of the "chain" we'll be using (equivalent to the pair of #'s in the
card trick).

Lemma: For any arrangement of N points on a circle, $ a series of points of size N, i.e. an ordering
8P1 , P2 , ... PN < so that each Pi is adjacent on the circumference to Pi-1 and Pi+1 (mod N), and the
distance between Pi and Pi+1 clockwise along the circumference of the circle is less than or equal to
Π, assuming a unit circle, " i Ε @1, ND. (Note: This does not require PN be less than or equal to Π
units from P1 .)

Proof: Assume that such an ordering doesn't exist. Then there must be some troublesome point, i.e.
one that is more than Π from the point "behind" it in our sequencing. Then we simply have to recre-
ate the ordering again, starting at this "troublesome point;" since it was more than Π from the point
behind it, it must be less than Π from the point in front of it because the circle has circumference 2Π
and all the points are distinct, and so the ordering should work. After doing so, we realize that there
cannot be any more troublesome points, because the first troublesome point's existence implies that
So, how can Amayak and Arutyun preform this super trick? Similarly, it has many (even more, in
fact) very-close solutions: note first that just drawing a random semi-circle gives you a 50% chance,
and it isn't particularly hard to do better. However, the 100% solution is just that; it does require a 3
Magic Tricks.nb
short lemma, to prove the existence of the "chain" we'll be using (equivalent to the pair of #'s in the
card trick).

Lemma: For any arrangement of N points on a circle, $ a series of points of size N, i.e. an ordering
8P1 , P2 , ... PN < so that each Pi is adjacent on the circumference to Pi-1 and Pi+1 (mod N), and the
distance between Pi and Pi+1 clockwise along the circumference of the circle is less than or equal to
Π, assuming a unit circle, " i Ε @1, ND. (Note: This does not require PN be less than or equal to Π
units from P1 .)

Proof: Assume that such an ordering doesn't exist. Then there must be some troublesome point, i.e.
one that is more than Π from the point "behind" it in our sequencing. Then we simply have to recre-
ate the ordering again, starting at this "troublesome point;" since it was more than Π from the point
behind it, it must be less than Π from the point in front of it because the circle has circumference 2Π
and all the points are distinct, and so the ordering should work. After doing so, we realize that there
cannot be any more troublesome points, because the first troublesome point's existence implies that
there are no points at all on 12 of the circumference, and thus the cycle just has to start at the first
point in the other 12 of the circle, and the resulting series can never violate our conditions. ƒ

Let's take a look at some examples:


4 Magic Tricks.nb

In the first diagram, (upper left), the lone point is a troublesome point: we begin at 12 O'clock, and
move clockwise (similar to the card trick, the direction and starting point are ambiguous, but must be
agreed upon beforehand). We find the four points to be in keeping with our regulations; however,
the fifth is not. Thus, we simply start over at that point, and see that it does work. The second dia-
gram (upper right) is a slightly different example, where the first point we cross is the troublesome
one.

How does this help Amayak and Arutyun complete their super trick? Amayak simply has to find this
series, and remove the last point in it (e.g., for the first diagram, the clockwise-most point in the
group of 4 points). Then when Arutyun comes back in, he will certainly find the same series of points
(excluding the last one), since the placement of the final point in the sequence doesn't actually effect
the rest of the sequence (if it did effect the chain, it would have to be less than Π away from the point
in front of it, placing it not at the end but in the middle, a contradiction). Thus, he simply has to
draw his semicircle from the last point in the series he comes across: by definiton, the last point must
be within Π of where the semicircle (of length Π is drawn), which is the desired result.

Finally, an interesting theoretical question arises from the second trick: how many different such
series are there per an arrangement of k points on a circle? For the first and second diagrams, the
move clockwise (similar to the card trick, the direction and starting point are ambiguous, but must be
agreed upon beforehand). We find the four points to be in keeping with our regulations; however,
the fifth is not. Thus, we simply start over at that point, and see that it does work. The Magic
second dia- 5
Tricks.nb
gram (upper right) is a slightly different example, where the first point we cross is the troublesome
one.

How does this help Amayak and Arutyun complete their super trick? Amayak simply has to find this
series, and remove the last point in it (e.g., for the first diagram, the clockwise-most point in the
group of 4 points). Then when Arutyun comes back in, he will certainly find the same series of points
(excluding the last one), since the placement of the final point in the sequence doesn't actually effect
the rest of the sequence (if it did effect the chain, it would have to be less than Π away from the point
in front of it, placing it not at the end but in the middle, a contradiction). Thus, he simply has to
draw his semicircle from the last point in the series he comes across: by definiton, the last point must
be within Π of where the semicircle (of length Π is drawn), which is the desired result.

Finally, an interesting theoretical question arises from the second trick: how many different such
series are there per an arrangement of k points on a circle? For the first and second diagrams, the
points are all confined on 12 of the circle, and thus there is only one. In the third diagram, the maxi-
mum number of series is obtained, with one for each point on the circle (as each can start the cycle).
In general, what function approximates the expected value of the number of such cycles for a given
k?

You might also like