wu :: forums
« wu :: forums - Yet another pigeonhole »

Welcome, Guest. Please Login or Register.
Nov 25th, 2024, 2:52am

RIDDLES SITE WRITE MATH! Home Home Help Help Search Search Members Members Login Login Register Register
   wu :: forums
   riddles
   medium
(Moderators: Icarus, Eigenray, Grimbal, towr, SMQ, ThudnBlunder, william wu)
   Yet another pigeonhole
« Previous topic | Next topic »
Pages: 1  Reply Reply Notify of replies Notify of replies Send Topic Send Topic Print Print
   Author  Topic: Yet another pigeonhole  (Read 541 times)
ecoist
Senior Riddler
****





   


Gender: male
Posts: 405
Yet another pigeonhole  
« on: May 16th, 2008, 9:07pm »
Quote Quote Modify Modify

Show that every set of 90 numbers selected from {1,...,2008} contains four different numbers such that the sum of two of the four equals the sum of the remaining two.
IP Logged
Grimbal
wu::riddles Moderator
Uberpuzzler
*****






   


Gender: male
Posts: 7527
Re: Yet another pigeonhole  
« Reply #1 on: May 17th, 2008, 1:49am »
Quote Quote Modify Modify

I would use the pigeonhole principle  Roll Eyes.
Create the set of all ordered pairs of 2 numbers chosen among the 90.
That set has 90*89/2 = 4005 elements.
Now compute the sum of each pair.
The minimum total is 3, the maximum is 2008+2007 = 4015, that is 4013 possible values.
Then uh.....
Well, the idea was to find 4013<4005 so that there would have been an a+b and c+d such that a+b=c+d.  
But there are still too many holes and too little pigeons.
Back to the drawing board...
IP Logged
rmsgrey
Uberpuzzler
*****





134688278 134688278   rmsgrey   rmsgrey


Gender: male
Posts: 2873
Re: Yet another pigeonhole  
« Reply #2 on: May 17th, 2008, 7:01am »
Quote Quote Modify Modify

on May 17th, 2008, 1:49am, Grimbal wrote:
I would use the pigeonhole principle  Roll Eyes.

So would I:
 

Again, start with the set of 4005 distinct pairs, but, rather than computing their sums, compute their differences.
 
Take any pair, (a,b) with a-b=x. For any pair, (c,d) with c-d=x, a+d=a+(c-x)=(a-x)+c=b+c, so when (a,b,c,d) are all distinct, any repeated pairwise difference gives you the set of four numbers.
 
There are three ways (a,b,c,d) can have repeated values:
a=c,b=d - which means that (a,b) and (c,d) are the same pair
a=d, b=c-2x
b=c, a=d+2x
The latter two each allow a different pair from (a,b) to share the same difference without giving a set of four by having one element in common - forming an arithmetic progression of three elements, (i,j,k). Extending the progression to a fourth element, (i,j,k,l), gives two non-overlapping pairs with the same difference, so gives a set of four.
 
So, to avoid having a set of four with the required property, each difference can occur at most twice. There are 2007 possible differences (from 1 to 2008-1), but, a repeated difference must be less than or equal to half the maximum possible difference, so only 1003 differences can repeat, accounting for at most 2007+1003=3010 pairs, so any subset of 79 or more numbers (79*78/2=3081 pairs) must contain a set of four with a shared pairwise sum.

 
I suspect ecoist may have had another proof in mind since my result is stronger than required
IP Logged
ecoist
Senior Riddler
****





   


Gender: male
Posts: 405
Re: Yet another pigeonhole  
« Reply #3 on: May 17th, 2008, 10:08am »
Quote Quote Modify Modify

Right, rmsgrey!  My proof was lazily based on the earlier pigeonhole concerning parallelograms on a checkerboard.  I knew my proof was weak because it required 11 numbers from {1,...,30}; yet it seems that 9 is sufficient.  Even smaller than the 10 your argument requires!
IP Logged
Pages: 1  Reply Reply Notify of replies Notify of replies Send Topic Send Topic Print Print

« Previous topic | Next topic »

Powered by YaBB 1 Gold - SP 1.4!
Forum software copyright © 2000-2004 Yet another Bulletin Board