Saturday, July 29, 2006

Correlation smile...

Looks like the good old smile (or smirk) problem comes back in the credit derivatives market. We spent some time in the Derivatives class trying to understand why the smile exists in equity derivatives, which implies that the BS assumptions dont apply to the way the market prices these instruments.

Similarly, the one-factor Gaussian Copula model, while widely used, doesnt reflect market beliefs, thus producing an implied correaltion smile for the different tranches of say a CDX index. While an open ended question, some believe that its because the fat tails are not captured adequately by the gaussian copula model.

Wednesday, July 26, 2006

Week 19 - one more to go before T2 ends!!

Derivatives Software Posted by Picasa

As part of the project on Derivatives: Economic Concepts course, we had to develop a working software with input/output forms to value options with extensible features, using implied binomial trees. The project was setup as a case study for advising a "client" about their various strategies to insure their fortunes. My group used MATLAB to develop this sotware..which turned out to be pretty cool at the end.

P.S.: As for the color selection of the GUI, lets just say that our "client" demanded it that way!!

Sunday, July 23, 2006

week 18...the toughest one!!

Week 18 was one of the toughest. We had Computational finance project report and presentation due. The project involved barrier option pricing using MC methods. The project involved developing pricing engines, which my team did by protyping in MATLAB and then developing the engine in C++. We were glad we did the prototyping - it really came through for us to understand the option by parametric studies.

As soon as we got done with this, we had to jump on for another big project due on Tuesday. Derivatives pricing using implied binomial trees....

Thursday, July 13, 2006

Week 17...

So there is a method to the madness. The idea of Monte Carlo simulation for pricing derivatives is well known, but even more fascinating are the mathematical constructs to speed up the process. The basic normal random number generators cluster the points in space, but quasi-random sequences (QRS) and stratification techniques help make the process more efficient. An excellent online java applet to visualize this is given here.

Week 17 - credit final .. and paper was due.

Monte Carlo using Stratification Posted by Picasa

Saturday, July 08, 2006

Where Credit Is Due --

Where Credit Is Due - CFO Magazine - November Issue 2003 -

This old article caught my eye as I took a break looking over some papers for the upcoming credit risk final. Immediately, I see the practicality of the complex models we had been discussing in the class. I am also busy researching the cause of downgrade for a company as part of the paper due for this course. This course is very practical - it helps when the Prof. worked as an MD at a top credit rating agency.

This class has been extremely interesting - the first of a two-series course on credit risk. Next term, we take a more quantitative approach on building models.

Last week was busy with some heavy C++ coding (I am still learning how to code effectively) for the Quant methods course - we are discussing innovative ways to speed up Monte Carlo simulations using methods that would make a theoretical physicist blush - Brownian bridges, Stratification, Quasi-random sequences, Sobol (is that a new language???), ...

This has got to be the craziest quarter yet!!