said Jean-Luc Godard , the French-Swiss film director, once. I just found this quote by chance and it took me to add some words to my Bebop Finance post.
We, quant finance software makers, take from books ... models. To model instruments from the simple to the most sophisticated, we apply interest rate models from Hull-White, Black Karasinski, one- and more-factor, to LMM with stochastic interest rate volatility and jumps ... equity models from Black Scholes to Exponential Levy, BS extensions to Dupire, Heston, Bates (again with stochastic volatility and jumps). Correspondingly with derived models for FX, inflation and so on - we take models from ...
Then we solve PDEs or PIDEs, or do (Quasi-)Montecarlo-Simulaten with least square techniques and we need to know how to do the parameter identification right, for calibration and recalibration. Easy? Not so easy, if you combine the right deal types - models for underlying - solvers and calibrators, correctly
Interestingly enough, there is not so much literature going into that depth. Anderson, Piterbarg, Interest Rate Modeling do it to a certain level (not to write the, say, C++ code) in their great 3 volume book.
At the other hand we get the insight that more recalibration will be required and core competences of high-end numerics need to be extended to inverse problem knowledge and, intrinsic, global optimization. Large portfolio across scenario volumes force us put best efforts into the concrete implementation (in our case hybrid programming with grid and CUDA computing - but after we have optimizes software by, say, principal component analysis, techniques from asymptotic math and transforms) - we take models to ...
We strongly believe, made-to really matters ....
