Ridiculous Excel Result

[Gandalf]

I was trying to confirm that a sequence converged, using Excel.

Specific details here, if you care (http://www.actuary.ca/phpBB/viewtopic.php?p=214895#214895)

Out of the blue, Excel evaluated one term very different than its true value. A tiny difference in the value of x (depending on whether I determined the x's recursively or calculated each x from scratch) made a big difference in the value of f(x).

That's very disconcerting. Does anyone have ideas about knowing when your spreadsheet will evaluate expressions wrong?

[urysohn]

Using Excel 2000, I cannot duplicate the error...b18 shows as 4.54E-05. You can check the Microsoft site for a list of known bugs. This sounds like the 2^15 bug (or the 2-to-some-power bug. I think it's 15 but not sure), where 2^15 would always equal 1, whether it was entered or the result of a calculation. Either the same bug or one to the same effect. If you can locate the bug, they should have a patch for it as well. For the bug I'm thinking, the patch was lumped in with the "linked sheets not updating properly" bug.

(I know that with all this precise terminology I'm throwing around, you (a) will have no trouble locating the patches, and (b) are amazed I was eliminated from Survivor Jeopardy so quickly :P)