You are not thinking in terms of finite precision mathematics. Consider
x=1/3/10^8. With 15 figure accuracy, your calculation would be

1.00000000000000
.00000000333333333333333
-----------------------------------
1.00000000333333
-1.00000000000000
-----------------------------------
0.00000000333333

The final result only caries 6 figures, not because any operation involved
less than 15 figures, but because most of those 15 figures were wasted on
accurate representation of the 1 in 1+x, with the result that most of the
accuracy for x was lost.

On top of the basic issues of finite mathemetics, you also must consider
that most terminating decimal fractions (such as 1E-1) are non-terminating
binary fractions that can only be approximated (just as 1/3 can only be
approximated as a decimal fraction), and hence you will see some of these
finite precision issues in "simple" calculations where you don't expect them
to be relevant.

I am not aware of any instance where Excel performs basic arithmetic
incorrectly.

Jerry

"engineer" wrote:

I don't know if you count it as an error exactly, but try calculating the
formula y= 1+x-1, which should result in the value x. If x is small, say
1E-8, and you display the result of the calculation to say 12 figures, you
will find the result is not x, but may be more or less than x depending on
the value chosen. Accuracy exists only to 8 sig figs, despite claims that
Excel has 15 figure accuracy.

"Clarkinson" wrote:

Hi,
We had this guy in the office today making out that it was easy to find
faults in Computer software.
He claimed it was easy to show and prove that Excel calculates eroneously.
He didn't specify how or under what cirmstances and didn't have time to do a
demonstration.

Anyone else heard of this?

I check "Excel makes mistakes" on Google and found a few articles dated 2003
dealing with a statistical covariance problem .

Apart from that, nothing much. Is there something I've been missing ?

TIA,
EJC