Forums  > Pricing & Modelling  > Rough Volatility paper?  
     
Page 1 of 1
Display using:  

HockeyPlayer


Total Posts: 125
Joined: Nov 2005
 
Posted: 2017-06-14 20:07

Rough Volatility Barcelona 2017

I'd be interested in comments,

ronin


Total Posts: 684
Joined: May 2006
 
Posted: 2017-06-16 10:48
The model is interesting.

But surely there are enough classes of parametric volatility models already. Who needs another one, and what for?

"There is a SIX am?" -- Arthur

mtsm


Total Posts: 255
Joined: Dec 2010
 
Posted: 2017-06-16 15:17
I could not agree more.

I don't think that the research agenda is dictated by need, usefulness, etc..., but by established people in tenured positions or managing roles, who know how to do that sort of modeling, can get this stuff published, but don't master possibly more timely techniques. It's hard to learn something new admittedly.

Quant finance 1.0 has been slowly decaying since the early to mid 2000s....

silverside


Total Posts: 1446
Joined: Jun 2004
 
Posted: 2021-08-26 14:57
I find the "rough volatility" idea extremely interesting : it seems to be possible to fit the market volatility surface using a conceptually sound process for volatility.

I see that more recently, a calibration approach using neural networks has enabled near-real-time fits.

Is anyone using this in practice ? Or is it still an intellectual curiosity ?

I never worked much with Equity Derivatives and wouldnt be up to speed with e.g. Stochastic Local Vol, but this does seem like a really cool idea!

nikol


Total Posts: 1377
Joined: Jun 2005
 
Posted: 2021-08-29 07:20
Regarding ANN driven pricings - how can I hedge?
Thinking logically, I have to train ANN for each derivative type and then apply hedge ANN, no?

... What is a man
If his chief good and market of his time
Be but to sleep and feed? (c)

silverside


Total Posts: 1446
Joined: Jun 2004
 
Posted: 2021-08-29 10:45
I skimmed a recent paper...

The idea was that the neural network (which took a long time to train) would recognise the "type" of volatility surface to give a rough calibration, and then move to a quicker, semi analytic calibration taking mere seconds.

I don't think it mentioned hedging but perhaps the old Jacobian trick could be used.

I think the calibration in that paper was to vanillas. I didn't keep the link but it may have been mentioned on risk.net

nikol


Total Posts: 1377
Joined: Jun 2005
 
Posted: 2021-08-29 13:18
Ah, sorry. As it happens - I read your message, but then "moved sideways" following associative path and replied to those pictures generated by my imagination :))

For ANN related path I am thinking into this direction:
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3591734

Still, apology for the thread hijacking.

However, let me reformulate the same for "Rough volatility". My initial reply might help to corner the idea.

If I use a new process with a "very good fitting" capability, then the usefulness of that depends on the usage:
- market making? - very likely, but still I have to capture relationships with all other products across strikes and maturities to keep risks under control
- pricing and hedging of exotic structures? - requires effort to develop pricing machinery for the new process and make translation of all greeks especially vega related.

... What is a man
If his chief good and market of his time
Be but to sleep and feed? (c)
Previous Thread :: Next Thread 
Page 1 of 1