How to Do but not Overdo it

Yesterday, our first compact seminar on Model&Methods Risk in Quantitative Finance. In our home town Linz. We invited core users from who we could expect unmistakable responses and assessments.

Andreas Binder (picture) and Michael Aichínger guided through the most beautiful but also dangerous fields of mathematical finance and pointed out where one can still make fundamental errors in concrete derivatives analytics examples.

Tree based methods, finite differences, finite elements? Finite elements! But what happens, if you, when solving a reaction-convection-diffusion-type short rate PDE don't apply streamline diffusion? How do I treat Bermudan rights in Montecarlo? Or: if one applies the Longstaff Schwartz algorithm in the form originally published, it works well for Bermudan Calls and alike, but not for, say, a Callable Reverse Floater. Andreas and Michael presented full explanations and the required improvements. I also found the examples, why more factors do not necessarily explain more and why parameter identification in calibration is so difficult, insightful. And why enormous speed-up is required doing result optimization in Montecarlo and that GPU co-processing became state of the art at the right time.

Without getting a red face, I summarise, positive response is an understatement. We  are fully motivated to go with the seminar where we find interest. Vienna, 17-Jun-09 and Zurich, 15-Jul-09 fixed.