Page 4 of 5

General Remarks and so on

Posted: January 13th, 2012, 12:07 pm
by Cuchulainn
qfcl issues

General Remarks and so on

Posted: January 17th, 2012, 5:17 pm
by Polter
outrun, not sure if it's as general as you desire, but there is some benchmark data in QuantLib's test-suite (including the references to relevant journal articles), perhaps that would be of help, e.g.:http://quantlib.sourcearchive.com/docum ... F5611.aspx

General Remarks and so on

Posted: January 17th, 2012, 9:25 pm
by Cuchulainn
QuoteOriginally posted by: outrunI'll have to examine the license though.. Can I cite a verbatim table from Espen Haug?Better still, generate the *same* table of values, which is a great check. Ideally, 4-5 independent sources and compare them all.

General Remarks and so on

Posted: February 6th, 2012, 9:01 am
by Cuchulainn
QuoteOriginally posted by: outrunQuoteOriginally posted by: CuchulainnQuoteOriginally posted by: outrunI'll have to examine the license though.. Can I cite a verbatim table from Espen Haug?Better still, generate the *same* table of values, which is a great check. Ideally, 4-5 independent sources and compare them all.that's what I've just done! I've tested (and passed) against 43 cases from Espen's first book. vanilla option test (javascript version) vanilla option test (C++ version)This is useful for MC and PDE libraries. I now have the gist of the error postprocessor analysis (NPP) done for 2d PDE I would like to do same here as well. The whole testing process can be automated by parameter bumping (and with lots of cores).BTW 2 factor exact solutions in C++ would be very useful. In the style of your 1 factor formulae, lean and meanHere a snippet.

General Remarks and so on

Posted: February 8th, 2012, 6:39 am
by Cuchulainn
Any views on which tensor library is best e.g. foe 3d pde, storing nd data (K, T, s1, s2) as input to post processing?I use boost multi_array which is workable but takes a bot of time to get used to the syntax.

General Remarks and so on

Posted: February 8th, 2012, 1:43 pm
by frenchX
Just a thing which could be nice to include in your code. It's for interest rate derivatives concerning counterparty credit risk and funding costs. Is it possible to include those functionalities ?Credit Valuation Adjustment (CVA)Debt Value Adjustment (DBA)Liquidity Value Adjustment (LVA)Funding Value Adjustment (FVA)The papers of Ancast are the best to start with. Funding CVA DVAPricing of derivatives under CSAand a lastBrigo paperIt's already a VERY HOT topic in the finance community and it will be soon a standart in the pricing of derivatives so I think it has to be included.

General Remarks and so on

Posted: February 8th, 2012, 2:46 pm
by Cuchulainn
QuoteOriginally posted by: frenchXJust a thing which could be nice to include in your code. It's for interest rate derivatives concerning counterparty credit risk and funding costs. Is it possible to include those functionalities ?Credit Valuation Adjustment (CVA)Debt Value Adjustment (DBA)Liquidity Value Adjustment (LVA)Funding Value Adjustment (FVA)The papers of Ancast are the best to start with. Funding CVA DVAPricing of derivatives under CSAand a lastBrigo paperIt's already a VERY HOT topic in the finance community and it will be soon a standart in the pricing of derivatives so I think it has to be included.If you can flesh out the ALGOS for these then I could have a look at implementing it.QF --> Algos -> Numerics -> C++/C#First 2 steps for you, then I'm in. Otherwise, no time in current project NEED1. algos2. test data3. test output. I am just a simple software developers, so please explain as precisely as possible Merci beaucoup.

General Remarks and so on

Posted: February 8th, 2012, 2:58 pm
by quartz
QuoteOriginally posted by: frenchXIs it possible to include those functionalities ?Credit Valuation Adjustment (CVA)Debt Value Adjustment (DBA)Liquidity Value Adjustment (LVA)Funding Value Adjustment (FVA)...It's already a VERY HOT topic in the finance community and it will be soon a standart in the pricing of derivatives so I think it has to be included.For large banks it's a standard already, you're right in that it's a must (in particular with the upcoming Basel regulation).Personally for what I can I'll keep an eye on steering QFCL in that direction. And one of my next project is about it, so sharing some effort might be useful! Are you being involved in the topic?

General Remarks and so on

Posted: February 8th, 2012, 3:14 pm
by Cuchulainn
QuoteOriginally posted by: quartzQuoteOriginally posted by: frenchXIs it possible to include those functionalities ?Credit Valuation Adjustment (CVA)Debt Value Adjustment (DBA)Liquidity Value Adjustment (LVA)Funding Value Adjustment (FVA)...It's already a VERY HOT topic in the finance community and it will be soon a standart in the pricing of derivatives so I think it has to be included.For large banks it's a standard already, you're right in that it's a must (in particular with the upcoming Basel regulation).Personally for what I can I'll keep an eye on steering QFCL in that direction. And one of my next project is about it, so sharing some effort might be useful! Are you being involved in the topic?It means, someone has to be the driver.

General Remarks and so on

Posted: February 8th, 2012, 3:16 pm
by frenchX
I'm far from being an expert on this: Ancast and Piterbarg would be FAR much more efficient than me but let's give it a try. For the quant finance and the algos, I just need to ask one question:Do you use the PDE framework or MC in your current scheme ?Basically in the most complete framework (id est with credit risk and funding cost) the value of a derivative is given byV(t,C,F)=E[Payoff(t,tau^T)+g(t,tau^T,C)+phi(t,tau^T,F)]+E[1(tau<T) *D(t,tau)*theta(C,eps)]Let me explain all the termsE[Payoff(t,tau^T)] is the risk neutral value of the derivative contract (id est the classic one) E[g(t,tau^T,C)] is the expected value of the collateral margining between t and T with C being the collateral accountE[phi(t,tau^T,F)] is the expected value of the funding and investing costs between t and T with F being the cash account needed for tradingTheta(C,eps) is the on default cash flow with "eps" being the amount of losses or costs that the surviving entity would incur upon a default eventThe dynamic of C, F and eps are defined by the ISDA CSA agreement. More on the BCCFVA pricing equation later but that's the basic idea and you would have to incorporate this for Basel III requirement. @Quartz: No I'm not involved in this topic professionaly but more personnaly I can help for the maths and the algos but I'm definitively not a good developper. I'll reread the paper of Brigo carefully and try to sum up simply. His approach (Brigo one) seems to be the most promissing in my opinion more than the two curves discounting one. It's more realistic, more deal dependent and clearer. The problem is that it's harder to implement and more ressource consuming.