The Truth About Computational Lies
In when three rights make a wrong I wrote about the risky horror of complicated models: a more complicated model may carry greater risk than a cruder one if they are abused or the users are not qualified to understand or solve them. In short, knowingly or not.
When valuations answer questions wrong
Let us assume that the models are naively used. The traps are manyfold. They may come from uninformative data, the wrong mix of core and use models and the ill-posedness of problems (like inversion) and methods.
The risky horror may have model and/or method reasons
Model validation and verification is not easy.
Validation is about determining the correctness of the model related to the real world problem and Verification is about determining the accuracy and robustness of its implementation.
It is important to distinguish between the problem, models and their (algorithmic) representations.
This is why we organize instruments, models (mathematical models) and methods (computational models) orthogonally. This makes cross-model and cross-method testing easy.
I now and then read about metrics helping the assessment, but I have my doubts ..
What is important in quant finance: knowledge and quantitative skills grounding in numerical methods.
This is why the UnRisk Academy has developed the Workout in Computational Finance as a reference class.
We really enjoyed working with quants at the London event - 30-Jan-14.
This is the first announcement of the next Workout: 26-Jun-14 in Frankfurt.
I'll keep you informed.
Picture from sehfelder