Forwarded for Erich Friedman, whose messages are being rejected by the list server: On Sun, Jul 28, 2013 at 9:56 AM, Erich Friedman <efriedma@stetson.edu>wrote:

...

More natural questions: What's the first feasible (H,L) with L > H? How about with L >= H + k for k=1, 2, 3, ...? Must some (H,L) work for any k?

i'm in the weird position of receiving math-fun posts, but am unable to post to the list.

but (3,4) is feasible.

randomly weigh AB against CD until AB<CD and D is removed. then weigh A against B. if they don't balance, C is heavy. then randomly weigh CE against FG until they don't balance. the remaining coin from the heavy side is guaranteed to be heavy.

erich