John Randall wrote:
> Basic idea: Over time, the numbers become closer.  Since their sum is
> constant and positive, they end up nonnegative.  It suffices to show that
> the sum of squares eventually decreases.  It does on one step if x+y+z>0,
> but I cannot yet do the rest.

I was on the right track, but did not quite have the right quantity. 
Spoiler follows.





















Suppose the values on the pentagon are (in order) p,q,r,s,t.  Let

S=(p-r)^2+(q-s)^2+(r-t)^2+(s-p)^2+(t-q)^2

(the sum of squares of differences 2 apart).  Now replace

p=p+r
r=-r
s=s+r

to get a new sum S'.  A calculation shows

S'-S=2r(p+q+r+s+t).

The parenthesized sum is positive by hypothesis, and we only do the
indicated move if r is negative.  Thus S is nonnegative, and decreases on
each move.  The process terminates.

Best wishes,

John


----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to