[François Villeneuve]

As part of our process to update the TOPS product, we conduct rigourous testing to make sure nothing has changed. We validate our models and want to ensure that the validated models stay the way they are.

We recently try to update to the 2006-04-10 version of TOPS from the 2005-06-01 version. Many of our trades that use Swaption2(...) showed variations. We looked in the README file and found nothing between those two dates that would indicate that there were any changes made to Swaption2(...). Perhaps some changes in other functions affected Swaption2(...).

We are not adverse to changes, but seek to know what they are and understand them. That way, we can document them and be confident in the conceptual soudness of the models we are using.

So can anyone help by telling me what changed in Swaption2(...) ? Thank you.

François

[Rich]

The changes made on 060220 and 060301 affect all models, and may be the cause of what you're seeing (though they shouldn't be). But we will look at the code for the two dates and see if there are other differences which were slippped through the cracks but were not documented.

Can you post an example of a difference you're seeing to help us find the cause?

Note too that Swaption2 was changed on 060516, the current version, partly due to a bug that was found by one of your colleagues. Are you sure you're not comparing to this 060516 version?

[François Villeneuve]

1) We'll post/send you and example for this. It's the same value in some deals, and different in others.

2) We are all aware that you made more changes to Swaption2(...). We wanted to finish the update process for the 20060410 version first, then "validate" the changes you made in 20060516 and finally, install that one in production. (Everybody here is on the same page, so you should be getting a consistent message from us. That should make your life easier.)

[François Villeneuve]

Did you receive my two spreadsheets? Have you been able to understand why, in once case, there is a difference, and why, in the other case, there are no differences? Thanks.

[Rich]

Yes, we did receive them. And have been poring over a few hundred thousand lines of code to find where the change is coming from, but without luck so far. We have determined the change occurred sometime in june of 2005, though that doesn't help you much. But we're going to keep looking.

[François Villeneuve]

Hi Rich. Any update on this?

[Rich]

I'm readying a test where all the older files are rebuilt and then going into the debugger to see where intermediate values differ from the current version. This is a bit tricky, and I expect to see it set up late this week.

[Rich]

We have found the change that causes the values to differ. In the 050704 version of TOPS a change was made which was not documented in the readme.txt file. An internal TOPS routine used to convert dates that are input in curvers and schedules (in your case, the coupon schedule) which contained dates before the valuation date was returning an error code, even though the inputs were OK. this was causing a problem in the swaption routines calculation of accrued interest. The newer values are the correct values.

We apologize for the error, and the time it took to locate the source of the problem. Thank you for your patience.

A newer version with, among other things, an updated readme.txt file, has been uploaded to the ftp site.