When Good ENUF is Great

8-Oct-09, the Risky Horror Show was exciting for us, and, I believe, for the participants.
Paul Wilmott traveled around the world and gave many interviews for the most reputable media giving insight in why the risky horror occurs. Paul is really one of the most influencing quants.

Andreas Binder and myself met him one hour before the event and it was great to talk to him: why people think more is always more and suppress danger of complexity?
Example: model calibration and horror oscillations. In the above picture you see a simple Hull White curve fit, nasty enough with annual curve points, but horrible with monthly curve points. You are completely lost in random parameter spaces, if not applying regularisation techniques.
Over 40 experts attended the live seminar and record-breaking many have followed the live web cast.
Andreas gave full explanation on the many traps from
  • models with too many parameters
  • poor numerical methods
  • unsmooth fits of market data and oscillating calibration
  • object functions with many local minma
and how to avoid them.
I used the title, inspired by a "Wired", Sep-09, article "the good enuf rvlutn" (about solutions where "simple" beats "perfect"). This is about managing massive data and not necessarily a metaphor for managing financial risk, but however, as fruit of challenging research enriched with real life examples from our customers, we recommend: TRY TO KEEP IT SIMPLE.

No comments:

Post a Comment