On The Number of Condoms at a Cheap

Safe-Sex Orgy

Yonatan Bilu, Dana Porraty, and Yoav Ya ez
De ember 11, 2002

Abstra t
Let M and F be nite sets. A straight orgy is a series of intera tions between ea h pair in M  F . Su h an intera tion is alled safe if
it is fa ilitated by a ondom omplex, a sequen e of basi units alled
ondoms. For the intera tion to be safe, the ondoms must ful ll some
dynami onditions whi h we formalize herein.
A straight orgy is alled a safe-sex straight orgy if all the intera tions
are safe. We give an exa t formula for the minimal number of ondoms
required to realize su h an orgy, up to an additive fa tor of 1. 
S hool

of Engineering and Computer S ien e, Hebrew University Jerusalem 91904

Israel, Email: johnblue s.huji.a .il

y Ele tri al

Engineering Department, Stanford University, Stanford, CA. 94305-9515

USA, Email: dporratwireless.stanford.edu

z Institute

of

Mathemati s,

Hebrew

University

yoavymath.huji.a .il

1

Jerusalem

91904

Israel,

Email:

1 Introdu tion
One of the best party riddles known to the authors is the following: Two
straight ouples are interested in having a safe-sex orgy. That is, ea h woman
would like to have sex with ea h of the two men, using a ondom. Condoms
may be used more than on e, but ea h parti ipant may only tou h a lean
side of a ondom, or one that is only stained by that parti ipant's uids.
How an they do this with but two ondoms? This is a good party riddle,
sin e as the reader may readily verify, it is not very hard to solve, and serves
as a potent i e breaker.
A vanilla riddle with the same avor, is the following: Three surgeons need
to operate on a patient (one after the other). Both the surgeons, and the
patient, may be arrying a terrible disease, so they must use surgeon gloves.
How an they operate on the patient if they have only two pairs of gloves?
Let us return to the obs ene phrasing of the problem. Denote by Con(f; m)
the minimal number of ondoms needed for a safe-sex orgy with f women
and m men. What bounds an be given for this number? It is easy to see
that:
1
(m + f )  Con(f; m)  m + f;
2
and the lever reader might also see that Con(f; m)  minfd 12 me+f; m+ 12 f g,
but it turns out that both these bounds an be improved, as stated by the
main theorem of this paper:
Theorem 1.1 Let = minfd 32 me + d 21 f e; d 23 f e + d 12 meg, then:
1  Con(m; f )  + 1
When there will be no way around that, we shall refer to a general parti ipant as \she". This in no way re e ts the authors' views on the gender of
orgy parti ipants. We would also like to larify that we have no rst-hand
knowledge of the subje t whi h inspires the mathemati al dis ussion herein,
nor wish to a quire one.

2 De nitions
Let P = F [ M be a nite set alled the parti ipants, where the subset F is
alled the women, and the subset M is alled the men. Let C be a set alled
2

the ondoms, where ea h member has two sides.

De nition 2.1 A ondom omplex is an even-length sequen e of ondom
sides, (s1 ; :::; s ), su h that for two sides in the sequen e, s ; s , there exists
a 2 C with = fs ; s g i there is some natural number i su h that
j = 2i; k = 2i 1. In su h a ondom omplex we shall say that two ondoms,
1 ; 2 , are tou hing, if there exists an i su h that s2 2 1 ; s2 +1 2 2 , or vi e
versa.
A sexual a t is a sequen e (p1 ; S; p2 ), where p1 ; p2 2 P , and S is a ondom
omplex.
An orgy is a sequen e of sexual a ts. In this ontext we shall sometimes refer
to a sexual a t as an orgy round. Also, we shall say that the two parti ipants
in that round are having sex.
A straight orgy is an orgy where M and F are disjoint, and in ea h round
one of the parti ipants is in M , and one is in F .
l

j

j

k

k

i

i

Condom sides also have a subset of P asso iated with them, alled their
tou h set. These hange as a result of sexual a ts in whi h the ondoms are

involved, in the following manner:

De nition 2.2 Let (p1 ; (s1 ; :::; s2 ); p2 ) be an orgy round, and let T 1 ; :::; T 2
be the tou h sets of s1 ; :::; s2 , respe tively, before this round. Then the respe tive tou h sets after this round, T 1 ; :::; T 2 are de ned as follows:
r

l

s

l

r

0

r

r

s l

0

s l

s 

8i = 1; :::; l 1, T 2 ; T 2 +1 = T 2 [ T 2 +1 
T 1 = T 1 [ fp1g 
T 2 = T 2 [ fp2g
r

0

s i

r

0

0

r

r

s i

s i

0

s

s

r

r

r

s i

0

s l

r

0

s l

We are now ready to de ne the kind of orgies we are interested in:

De nition 2.3 A sexual a t (p1 ; (s1 ; :::; s ); p2 ) is a safe sexual a t if l > 0,
and at its beginning the tou h set of s1 is either empty, or fp1 g, and the
tou h set of s is either empty, or fp2 g.
A straight orgy is a straight safe-sex orgy if before the rst round ea h tou h
set is empty, and ea h of the orgy rounds is a safe sexual a t.
Con(m; f ) is the minimal size of C , su h that there exists a straight safe-sex
orgy with jF j = f; jM j = m, and for ea h (p1 ; p2 ) 2 M  F there's an orgy
round where the parti ipants are p1 and p2 ..
l

l

3

3 Lower Bound
Let us rst restri t the dis ussion to Con(n; n). We'll need yet more de nitions.

De nition 3.1 A side of a ondom is lean as long as its tou h set is empty.
Otherwise, it is un lean.
A parti ipant p owns a ondom's side s, if at some point, the tou h set of s
is fpg.
For example, whenever a parti ipant tou hes a lean side, she be omes the
owner of that side.

De nition 3.2 A parti ipant is modest if she owns exa tly one side of one
ondom.
Two parti ipants are a ouple if ea h owns one side of the same ondom.
The two members are alled partners of ea h other.
A ouple is an f- ouple if both its members are female. It is an m- ouple if
both are male. Otherwise it is an h- ouple.
A ouple is a modest ouple if both its members are modest.
Note that a parti ipant may be a member of more than one ouple.

De nition 3.3 A parti ipant is a tive in an orgy round, if she has sex in
that round. A parti ipant is passive in an orgy round if she is not a tive, but
one of her partners is.
Note that during some rounds a parti ipant is neither a tive nor passive.
Understanding the a tivity pattern of the (modest) parti ipants is the key
to bounding the number of ondoms, and will also prove instrumental in
designing the orgy for the upper bound.

Lemma 3.1 On e a modest parti ipant had been a tive, and then passive,
she an not be a tive again.

4

Proof: Assume for ontradi tion that p violates the lemma. On e she is

a tive, her (unique) ondom side be omes un lean, and no other parti ipant
may tou h it. On e she is passive her ondom is in use. If it tou hes an
un lean ondom side, then obviously p an't use it anymore. If it tou hes a
lean ondom side, then that ondom side be omes un lean, giving ownership
of it to p, in ontradi tion with her modesty.
We shall need a ouple of more de nitions to get a few orollaries regarding
the types of modest ouples in a safe-sex orgy:

De nition 3.4 Let (p1 ; p2 ) be a modest ouple. Call the parti ipant that is
rst to be a tive an a tive member, and all the other a passive member.
De ne r(p) for modest pair members, as follows:
If p is a passive member, r(p) is the number of the rst round in whi h p is
a tive.
If p is an a tive member, and her passive partner is q , r(p) = r(q ).
Corollary 3.1 Let (p1 ; p2 ) be a modest ouple, with p1 the a tive member,
and let r = r(p1 ) = r(p2 ).
If (p1 ; p2 ) is either an f- ouple or an m- ouple, then p1 is never passive before
r, and never a tive after it (and vi e versa for p2 ).
If m; f > 1 then this is also true when (p1 ; p2 ) is an h- ouple.
Proof: First note that by de nition p2 is never a tive before r, so p1 is never
passive during this time. But as p1 is the a tive member, she is a tive at
some point before r.
If (p1 ; p2 ) is either an f- ouple or an m- ouple, then the pair members are
never a tive on the same round. Thus, at r, p2 is a tive and p1 is passive.
By the lemma, p1 an not be a tive again, and thus from round r onward,
p1 is never a tive, and p2 is never passive.
If (p1 ; p2 ) are an h- ouple then it must be the ase that both are a tive during
round r. Otherwise, p1 is passive, and by the lemma may not be a tive again.
But if p1 is never a tive after r, and p2 is never a tive before r, when will

they have sex with ea h other?
Furthermore, if m; f > 1, p1 may not be a tive again. Consider the next
round that one of them is a tive. If p2 is a tive in it, then p1 is passive, and,
indeed, by the lemma, may not be a tive again. Otherwise the onverse is
5

true, i.e. p2 is never a tive again. But if p2 is neither a tive after r, nor
before it, when will she have sex with the other parti ipants?

Corollary 3.2 There may not be both a modest m- ouple, and a modest
f- ouple.
Proof: Assume for ontradi tion that there is a modest m- ouple (M ; M ),
and a modest f- ouple (F ; F ), with M and F being the a tive members.
Consider the round t when M and F have sex. By Corollary 3.1, r(F ) 
t < r(M ). Similarly, by onsidering the round that F and M have sex,
we get r(M ) < r(F ). But as r(F ) = r(F ), and r(M ) = r(M ), we get a
a

a

p

a

a

a

p

p

a

a

p

ontradi tion.

a

p

p

a

p

p

a

Corollary 3.3 There may not be more than 2 modest h- ouples.
Proof: Assume for ontradi tion that there are, then in parti ular we have

that there are two ouples where the a tive members are of the same gender.
W.l.o.g. assume that they are women. Denote the ouples by (F1 ; M1 ) and
(F2 ; M2 ). Sin e our assumption implies that m; f > 1, by onsidering the
round when F1 and M2 have sex, and the round when F2 and M1 have sex,
we get a ontradi tion in the same manner as in Corollary 3.2.
We are now ready to prove the lower bound:

Theorem 3.1
7
6

Con(n; n)  d ne 1
Proof: By Corollary 3.3 we know that there are at most 2 modest h- ouple,
so let us forget about them. Now, by Corollary 3.2 we may assume w.l.o.g.
that the only modest ouples are f- ouples. In other words, modest men must
have immodest partners.
Let ea h of the parti ipants hoose one of the ondom sides they own, and
all it their hosen side. It is enough to show that there are at least 13 n sides
whi h are un hosen, and we shall do just that.
Partition the men into gangs in the following manner. Ea h gang leader is an
immodest man, and the other gang members are his modest men partners (if
6

there are any). A gang leader that owns k un hosen sides, has at most k + 2
members in his gang. At best, the gang in ludes himself, his partner for the
hosen side, and his k partners for the un hosen sides. Thus, the number of
un hosen sides a gang leader owns is at least one third the size of his gang.
Summing this up, we get that the number of un hosen sides is at least one
third the number of men, or 31 n.
Re all now the 2 h- ouples. We have a tually shown that the number of
un hosen sides is at least 3 2 , thus the number of ondoms needed is at least
d 21 (2n + 3 2 )e = d 67 n 31 e  d 76 ne 1
n

n

Corollary 3.4
1
2

2
3

2
3

1
2

Con(m; f )  minfd me + d f e; d f e + d meg 1
Proof: The same proof as for Theorem 3.1 works, with the ex eption that
we do loose generality by assuming that the only modest ouples are f- ouples
(ex ept for maybe two h- ouples). This assumption is indeed to our disadvantage if m  f , but in the omplementary ase we may only assume that
the only modest ouples are m- ouples, giving the bound in the orollary. It
is an easy exer ise to verify that the integral values are indeed as stated.

4 Upper Bound
Let us now try to design an orgy that a tually a hieves this lower bound.
Indeed, let us state it as a theorem, and then try to prove it:

Theorem 4.1
1
2

2
3

2
3

1
2

Con(m; f )  minfd me + d f e; d f e + mg + 1
Proof: The proof of Theorem 3.1 suggests how to design the parsimonious
orgy we are looking for. We shall show how to use only d 32 me + d 21 f e + 1
ondoms. By ex hanging the roles of men and women one gets d 23 f e + d 21 me +

1, giving the stated bound.
We saw that the worst ase in the proof above was when all the women were
7

Figure 1: An a tive, modest woman
modest, and two thirds of the men were modest. The immodest males were
ea h leaders of a 3-men gang, that is, ea h had exa tly two partners, and
this was the sole sour e for un hosen ondom sides.
Let us look at the women rst. Sin e they are all modest, they are partitioned
into pairs. For ea h pair, all the a tive member an a tive woman (see Figure
1), and the passive a passive woman. Denote by F the set of a tive women,
and by F the set of passive women. Let us now hazard a guess that there
is some round r su h that no passive woman is a tive before that round, and
no a tive woman is a tive from that round on.
Now let us look at the men. They have to be a tive both before round r,
and after it. By lemma 3.1 it must be the ase that the modest men are
(perhaps) passive at rst, then a tive and then passive again. This implies
that the immodest men are a tive, then passive, then a tive again (or they
might start as passive, and then follow this pattern). In other words, the
proof of theorem 3.1 suggests that the immodest men are rst a tive with
the a tive women, then give away their ondoms to the modest men. These
have sex with the a tive women, and then (after round r) with the passive
women. Finally, the immodest men have sex with the passive women.
To see that this indeed works, denote by M the set of immodest men, and
partition the modest men into M1 and M2 in su h a way that no two men
in the same set have a ommon partner (note that all three sets are of size
3 ). We shall also need one extra ondom. Give a ondom to ea h woman
in F , and to ea h man in M and in M1 . At rst these have sex, and after
A

P

I

n

A

I

8

that ea h has a ondom with one lean side, and one un lean side. Now have
the men in M invert their ondoms, and give them to the men in M2 . Now
let the men in M2 have sex with the women in F , using the extra ondom
in su h a way that the ondoms owned by F , maintain one lean side (the
ondom side previously used by M now be omes unusable, but not to worry,
they have extra sides).
This brings us to round r. Now ea h woman inverts her ondom, whi h still
has one lean side, and gives it to her partner in F . These have sex with
the men in M2 , whi h retain their ondoms from the previous rounds. Now
the men in M1 have sex with F , using the ondoms they previously used,
and the extra ondom to keep the other side of their ondom lean. Finally
M1 invert their ondoms, give them to M , whi h have sex with F , and the
orgy is done (so is the proof).
I

A

A

I

P

P

I

P

5 Orgy Graphs
In this problem we explored a straight safe-sex orgy where ea h possible
straight ouple have sex. Thinking of the parti ipants as vertexes, and the
sex a ts as edges, we an say that a straight safe-sex orgy realizes the graph
K . In general, a safe-sex orgy realizes a graph G, if the parti ipants
orrespond to the vertexes of the graph, ea h orgy round is safe (the ondom
omplex is non-empty, and the tou h set of the sides at its end ontain, are
either empty, or the singletons orresponding to the parti ipants on either
end), and for ea h edge (u; v ) in the graph there is an orgy round where
the parti ipants are u and v . Denote by Con(G) the minimal number of
ondoms required to realize G as a safe-sex orgy. This paper shows that
Con(K )  minfd 32 me + d 21 f e; d 23 f e + d 21 meg.
Similar arguments to those used in the previous se tions show that d 23 en 1 
Con(K )  d 23 en + 1. For the lower bound, note that the proof of Corollary
3.2 implies that in a realization of K there an be only one modest ouple.
Partitioning the parti ipant into gangs, as in the proof of Theorem 3.1, gives
Con(K )  d 23 en 1.
For the upper bound, divide the parti ipants into three groups of size 3 , say
M ; M1 ; M2 . Give a ondom to those in M and in M1 . These an now realize
the orgy among themselves, leaving one side lean. Next give the ondoms
used by M to M2 . M2 an now realize the orgy among themselves, and with
m;f

m;f

n

n

n

n

I

I

I

9

the use of an extra ondom, have sex with those in M1 , keeping on side of
the M1 ondoms lean. Finally, the ondoms from M1 are given to M , and
these have sex with the members of M2 .
I

We leave open the value of Con(G) for other graphs (bipartite or otherwise),
and the s ope of appli ability of the results herein.

10