I cannot make sense of the YIELDMAT results, no matter what definition of
YTM that I try. (There seems to be more than one.)

It would help me if someone posted the math formula or algorithm that
YIELDMAT uses.

And it would help me if someone would show how the YIELDMAT example in Excel
2003 help page is computed using that math formula or algorithm.

Some points to note in the help page description:


1. The synopsis at the top says that interest is presumed to be paid at
maturity, not at the coupon frequency.

I think that should impact the YTM (IRR) computation. But in one example
that I tried[*]), YIELDMAT came close (but not that close) to the result of
my reverse-engineered method of computing YTM only when I assume a regular
cash flow of coupons.

On the other hand, I cannot even come close to the results of the help
page result, no matter what frequency of cash flows I assume.
[*] http://www.moneychimp.com/articles/f.../fmbondytm.htm .


2. The YIELDMAT example suggests that the units of the YIELDMAT rate
parameter is "percent semiannual coupon".

I am not sure how that relates to an annual rate.

Moreover, since that requirement is not stated in the synopsis, I wonder
if the example annotation is simply incorrect.


3. The example uses the 30/360 mode ("basis"), not the actual/365 mode which
I understand better.

If it makes the explanation easier, I would not mind switching the help
page example to actual/365 (mode 3). In that case, YIELDMAT result is
6.09636299211303%, which you can round to Percent with 4 decimal places or
more, as the example does.

But I would also like to understand how 30/360 should be computed; that
is, how it affects the math formula or algorithm.


4. With respect to the 30/360 mode, I would like to understand how the
maturity date of 11/3/2008 might have been determined.

Of course, it could be arbitrary. But my curiosity is piqued by the fact
that it is not 11/8/2007 plus 360 days, but that date plus 1.

I wonder if there is some market convention for determining the one-year
maturity date when using the 30/360 mode.

And I wonder if/how that affects the YIELDMAT mathematical or
algorithmic computation when using the 30/360 mode.


5. I do not understand why the YIELDMAT implementation does not tolerate
settlement date equal to issue date.

I understand that they are usually not equal in the real world. But it
would simplify comparisons with reverse-engineered solutions.

Note: The help page does not identify this limitation. I stumbled upon
it empirically.