How to effectively verify the accuracy of a model(may be complicate) for exotic option, if there is no enough data of market price? Is there any related reference?


If your exotic contract specification can degenerate into a lighter exotic structure for which you can observe quotes, make sure you match them.

If you have nothing at all, try to assess how well you would have done, in average, by selling the exotic and hedging it according to your model assumptions for various past realisations of the market (ideally different regimes). If you always lost money in all scenarii, then your model - and the manner in which it captures the true market dynamics - is probably not suited for the sell-side. If you always made money, then your probably charging too high of a price and will not be competitive, hence not suited for the sell-side either.

As mentioned in the comments below, it may also help you to read the answers given to this related question.

| improve this answer | |
  • 1
    $\begingroup$ Adding to @Quantuple's answer: if you can derive put-call parity relations for variants of your exotic option (or degenerate the option into lighter options that admit parity relations), check whether your model verifies these parity relations. The advantage of this test is that you do not necessarily need real market data to perform it: e.g. you can verify the Black-Scholes model enforces put-call parity by simply inputting arbitrary values of $S_0$, $\sigma$, $r$, etc. $\endgroup$ – Daneel Olivaw Sep 29 '17 at 11:11
  • $\begingroup$ I second that @Daneel Olivaw. It actually reminds me of an answer to a related question I posted some time ago, here quant.stackexchange.com/a/27495/19887 $\endgroup$ – Quantuple Sep 29 '17 at 11:32
  • $\begingroup$ @Quantuple i was halfway through writing up an answer to this when you posted that link. I have typed almost exactly the same thing. The methods i use the most are the common sense and incremental validation - in reality it is just about being scientific about it, and understanding that you are fitting a model to a subset of the space of instruments, and then extrapolating outside of that region. $\endgroup$ – will Sep 29 '17 at 11:48
  • $\begingroup$ That's right @will. At the end of the day when pricing exotics all that we're doing is "arbitrage free extrapolation" :) $\endgroup$ – Quantuple Sep 29 '17 at 11:53
  • 1
    $\begingroup$ @Quantuple yeah. So provided you actually understand this, you can apply a lot of the same checks that you might to an extrapolation scheme - do i match at the boundary? Is the boundary smooth? what is the limiting behaviour? $\endgroup$ – will Sep 29 '17 at 11:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.