Message-ID: <18638084.1075856641315.JavaMail.evans@thyme>
Date: Wed, 11 Apr 2001 08:48:00 -0700 (PDT)
From: david.port@enron.com
To: vince.kaminski@enron.com
Subject: RE: VaR
Cc: john.lavorato@enron.com, tanya.tamarchenko@enron.com
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Bcc: john.lavorato@enron.com, tanya.tamarchenko@enron.com
X-From: David Port
X-To: Vince J Kaminski
X-cc: John J Lavorato, Tanya Tamarchenko
X-bcc: 
X-Folder: \Vincent_Kaminski_Jun2001_5\Notes Folders\Var
X-Origin: Kaminski-V
X-FileName: vkamins.nsf

Vince

Thanks for the update - especially your point 3 which echoes our own 
discussion yesterday: I want to clarify responsibilities for maintenance and 
administration of VAR. 
We need transparency around the process so that mapping decisions, for 
example, are easily made, supported and reviewed. 

It became clear after last Friday's events, that we are in an awkward 
paradigm with respect to how we handle VAR here, particularly at "pressure 
points".

I am putting together the scope of a "general overhaul" of the VAR process, 
including method and administration, to which I hope we would all buy in, so 
be assured of our co-operation.

DP 


 -----Original Message-----
From:  Kaminski, Vince  
Sent: Wednesday, April 11, 2001 2:49 PM
To: Port, David
Cc: Lavorato, John; Kaminski, Vince; Tamarchenko, Tanya
Subject: VaR

David,

During today's VaR coordination meeting  we had a discussion of issues 
related to mapping of the forward price curves into core locations.

Mapping is a  necessity dictated by the limitations of the computer system:
we have to reduce the dimensionality of the problem to stay within the bounds
of available CPU memory. Also, in some cases the quality of price discovery 
is poor 
and it's difficult to model the price curves independently: we solve the 
problem by mapping
them into more liquid and better behaved core  locations curves.

We have agreed on the following:

1. Winston will investigate the IT side and determine to what extent we can 
increase the number
of forward price curves that are simulated as basic (core) curves. He will 
investigate the impact of a larger
number of the core curves on the time required to complete the VaR run.

2. The curves associated with the biggest 10-20 positions in each commodity 
should be
modeled as core curves (i.e. no mapping into other locations). It makes sense 
to monitor
the biggest risks separately and avoid aggregating them into  less 
transparent aggregates.

3. The results of an automated clustering (mapping) procedures should be 
systematically
monitored by a human and corrected if they misrepresent the risks of the 
trading positions.
This responsibility should be vested with one person (right now the 
responsibility is
dispersed through the organization and this means in practice that nobody
is responsible). Research can allocate one person to this task;
cooperation of trading and RAC will be critical.

Vince