Hi there:Let's say I have an exotic product on multiunderlyings;Say n yrs. tenure (maturity) and early redemption features on a set of observation dates.(T1, T2, ... Tn)Now, i have a model using MC numerical scheme on the correlated Brownian Motion.1) Well, I doubt if the model gives reliable result, thus want to keep track on the daily P&L.Assuming that the trader executes DELTA HEDGING ON DAILY BASIS. Is there any way for me to break down the daily P&L such that I can know how much of it is attributed by WRONGCORRELATION ESTIMATE and WRONG VOLATILITY ESTIMATE .I saw on a book pulished by DB that they do have a formula derived using Ito's lemma. But thereare only terms like theta (time decay), cross-gamma and delta....How can I make use of this formula ???2) IN pricing the structure, I am naively using the correlation no. onBloomberg CORR function. As well known, it is HIGHLY UNSTABLE...What is the cutting-edge stuff on correlation modeling ??Do they do something like Exponetial Moving Average like their counterpartson Volatility Modeling (ARCH, GARCH... etc.)3) Does it mean that having a CORRELATION FIGURE WITH HIGH CONFIDENCE INTERVAL + good implied volatilty => the tradercan make money using the model ????Any paper, material, books... will definitely help me a lot.Appreciate your help & input...Mr. Money

Not too sure what you mean...can u elabroate further ????

OK, got what you mean.But the trader says that the vol. is correct.They are concerned that my correlation is wrong.Basically, I use the Corr. figure from Bloomberg.As u see, there is the T-Stat. on those values...Not sure which one to use ?? ( e.g. 1 yr data may contain too much stale info., regime shift 3 mth data may not be enough to provide good confidence interval/t-stat values)I am thinking if I can use some kind of Exponetial Moving Average on the correlations values ???OK, assumed that the Vol. is correct, then any difference is PURELY DUE TO DIFFERENCT CORRELATION VALUE.But, if the Vol. is also incorrect, then how much can i attribute the diff. to each factor ??? that's the challenge.Is there any closed form formula on such issue ???Many thanksMr. Money

QuoteOriginally posted by: MoneyHi there:I saw on a book pulished by DB that they do have a formula derived using Ito's lemma. But thereare only terms like theta (time decay), cross-gamma and delta....How can I make use of this formula ???2) IN pricing the structure, I am naively using the correlation no. onBloomberg CORR function. As well known, it is HIGHLY UNSTABLE...What is the cutting-edge stuff on correlation modeling ??Do they do something like Exponetial Moving Average like their counterpartson Volatility Modeling (ARCH, GARCH... etc.)Any paper, material, books... will definitely help me a lot.Appreciate your help & input...Mr. MoneyP&L Attribution is not easily done, paul has drawn a nice roadmap.There is correlation GARCHI presume it will be easy to do but you may have to adapt your exisiting routines. I haven't done it can't tell.

P&L attribution is frequently a regulatory requirement, so it is pretty universal, so much so that it if found in almost all front-end trading and risk management systems. Whether it is worth something is a different story.

Attribution is quite important in my shop. We sell illiquid exotics and then hedge out the risk we don't want. Management wants to know if we're laying off the risk we intended and, if not, where the ball was dropped. Was it in the trade execution, a persistent problem with assumptions, a recent market agitation, etc.As Mr. Money is finding out, this level of reporting is potentially much more difficult than putting the valuation/trading system together in the first place.Good Luck!

Paul, it is done at some places by risk mgmt too. As DavidJN mentions the regulators love it. The problem imho has to do with "data" being sparse in some cases. (And the other issue is that of relevance of that data)Making it impossible to do this sometimes. And then we have jumps to mess things up further.

What does "delta" mean when you have multiple underlyings? A vector of deltas for each underlying? Or a single delta for some sort of "Market" value? Or what?Likewise "implied vol".I tried to follow this thread, but find I cannot without knowing the answers to these questions.Does anyone have any references where these quantities are explicitly defined for an "exotic product on multiunderlyings"?

C'mon...My boss is giving me pressure everyday..So any idea on how to split the P&L into these two parts ???THere don't seem to have any closed-form formula ???Mr. Money

QuoteOriginally posted by: FermionWhat does "delta" mean when you have multiple underlyings? A vector of deltas for each underlying? Or a single delta for some sort of "Market" value? Or what?Likewise "implied vol".I tried to follow this thread, but find I cannot without knowing the answers to these questions.Does anyone have any references where these quantities are explicitly defined for an "exotic product on multiunderlyings"?No answers from anyone? Am I being obtuse?

so do I, wanna to know the answer as well...