Re: Info about adaptVanDelta argument in VannaVolgaBarrierEngine
on behalf on my dear former colleague Yue Tian (who wrote the engine): I think the adaptVanDelta flag is inspired by the Vanna-Volga Model in Murex, which also provides such a flag. The idea is to replicate the “true" vanilla price computed on the original smile in the edge case that the barrier option actually collapses to the vanilla. For example the up and out price is corrected by
noTouchProbability x ( trueVanillaPrice - vannaVolgaSmileVanillaPrice )
so if noTouchProbability tends to one, the up and out price tends to the vanilla price computed on the original smile. The latter price has to be provided as an input to the model if adaptVanDelta = true.
I don’t actually know if this method has a footing in the literature or if it is a proprietary approach.