Re: [Quantlib-users] FD Greeks computation when the calculated NPV value is too close to the spot value

[Prasanna K. <pk...@ku...>]

Klaus,
Thanks for the tip.  You are correct, the tests pass even after removing the condition.  How the engine-greeks and FD-greeks are compared, plays a role too.

In my copied code, I was purely testing the difference between the engine-greeks and the FD-greeks.  The pure difference between the engine-greeks and the FD-greeks do exceed tolerances.  The tests fail.

However, the ((difference between engine and FD greek) / spot) is always less than the tolerances and the tests pass fine.  Would you happen to know why the division by spot is necessary?  At one level it seems intuitive, but I can’t seem to have a mathematical explanation for this.

Thanks.


> On Jul 15, 2020, at 4:33 PM, Klaus Spanderen <kl...@sp...> wrote:
> 
> Hi
> 
> may be a left-over from the early days. IMO you can remove the condition and the tests will still pass.  May be in the past the discrete delta/gamma calculation in the test case  wasn't accurate enough to meet the tolerances for deep OTM options.
> 
> regards
> Klaus
> 
> 
> 
> 
> _______________________________________________
> QuantLib-users mailing list
> Qua...@li...
> https://lists.sourceforge.net/lists/listinfo/quantlib-users