MG VaR

Ted,

We have the first running version of VaR model for the MG positions. Many
thanks to Tanya and Anjam for taking lead
in this effort. We had also great help from Cantekin and Kirstee. We want to
validate some inputs
before releasing the number. Specifically, I want to make sure that both
London and Houston work
with the same position numbers every day to avoid confusion.

A few comments.

1. We plan to set up the models to run VaR concurrently in Houston
(Cantekin) and London (Kirstee) with the same inputs.
The reason to do it this way is that we may have to respond quickly to
multiple inquiries in both locations. (Cantekin and Kirstee,
please stay in touch and synchronize the inputs ).

2. It is critical that we run the model daily and receive the updated
positions and prices on a regular basis.

3. We shall continue work in Houston to improve the model. Two critical areas:

a. VaR model, by its nature, does not capture one big risk MG has in its
portfolios. This is the liquidity risk
resulting from different cash consequences of the hedges and the
underlying positions.
b. We should go through due diligence to validate the process generating
the position numbers (i.e. look
into the details of specific transactions). From what
I know, MG has a fairly rudimentary risk
management system and I am not quite sure how they generate option
prices (and deltas). They get some
option prices by calling brokers.

4. I think the current VaR is within 80-85% of the true number (assuming the
inputs are right). But, as I have said,
VaR does not capture all the risks that we have to worry about.

Vince