Forums  > Risk Management  > Vol surface changes as underlying moves  
     
Page 1 of 1
Display using:  

HockeyPlayer


Total Posts: 116
Joined: Nov 2005
 
Posted: 2016-02-24 21:12
We market make in highly liquid, near term options markets. I want to build a risk report that tells us how our portfolio's greeks will change as the underlying moves. This is for risk management in the face of unusually large moves, such as 2, 5 and 10% moves in the underlying. We are only trading vanilla options.

My starting vol curve takes IVs from the current market at every strike and lightly smooths them. I want to 'shock' the underlying & the vol curve in several different directions. Obviously, it is easy to shock vol up by 30% of current vol, but how do I predict how the curve moves when I shock the underlying up 5%?

Given that I am modeling large moves, I'm leaning towards using the "sticky delta" approach where the options with a given delta keep their volatility.

I didn't find any appropriate threads, links to previous discussions appreciated.

Baltazar


Total Posts: 1756
Joined: Jul 2004
 
Posted: 2016-02-24 23:13
The first thing you need to look at is how your ATM IV is going to move with the underlying move.
You can do that with historical data but that's already rather noisy and will depend on the horizon you look at.
The rest of the curve is more a secondary effect. It is easier if your smile is parametric and you study the historical evolution of such parameters with underlying/ATM vol changes.
That's the main drawback of using 'connect the dot' methods for fitting smiles opposed to more parametric approaches.

Qui fait le malin tombe dans le ravin

Patrik
Founding Member

Total Posts: 1325
Joined: Mar 2004
 
Posted: 2016-02-29 18:25
Very market dependent. A good trick here as Baltazar says is to have a "better model", i.e. something where your parameters do not move so wildly just because the underlying moved 5%. Depending on market it may or may not be easy to find such a thing, and as usual it's likely to be regime dependent and change over time.

EDIT: I should add as it's for risk management and not quoting you don't have to use the same model - you could prefer slightly worse fit to market if the risk it churns out for a bigger move is more realistic. Right tool for right job. The downside is that quoting-model greeks and risk-model greeks may not line up perfectly for same inputs as they represent different ways to view the world (with interpolation being mostly the no view of the world case).

Capital Structure Demolition LLC Radiation

EspressoLover


Total Posts: 202
Joined: Jan 2015
 
Posted: 2016-02-29 19:33
Don't know what horizon you're looking at, but one stylized fact to consider is that time of day effects are in play. For example in US equities, IVs are less sensitive to late afternoon price moves than at other times of day.

HockeyPlayer


Total Posts: 116
Joined: Nov 2005
 
Posted: 2016-02-29 19:53
We are looking at this for overnight risk modeling and to tell us how bad it would be if the underlying went limit up/down for 2 straight days.

EspressoLover


Total Posts: 202
Joined: Jan 2015
 
Posted: 2016-03-03 01:52
You may want to consider fitting the time-series of IV against underlying returns. There does tend to be short-term localized persistence, with longer-term reversion to historical relationships. For example YTD-2016, S&P IV has tended to undershoot the underlying.

To be honest I'm not sure if the vol surface snapped at a single point in time reflects this or not. My experience here is with stat-arb on vol futs, so I don't know how much marginal information these time-series signals add to the surface. However, even with large moves the time-series fitted IV-underlying relationship does tend to be very stable.

granchio


Total Posts: 1530
Joined: Apr 2004
 
Posted: 2016-06-28 20:12
@HockeyPlayer did you finalise this or are you still looking for ideas?

"Deserve got nothing to do with it" - Clint

HockeyPlayer


Total Posts: 116
Joined: Nov 2005
 
Posted: 2016-06-28 21:08

We will be trying out one approach in the next couple weeks, but I'd be very happy to hear more ideas.

mtsm


Total Posts: 179
Joined: Dec 2010
 
Posted: 2016-06-28 21:35
What asset class are you interested in?

I can only comment about rates and in my experience, in the sell-side front office, this is the holy grail of options market making - no less than that. Outside of the fee-based business, it is what you live and die by since definite knowledge of the volatility backbone is what lets you take directional vol exposure, hedge your vega exposure with delta, and other stuff a vol desk needs to do... I don't think standard option pricing models (including stoch vol) work at all for this. Still only talking about rates...

In a risk group, the requirements are less stringent presumably and some statistically significant model results may be sufficient.

granchio


Total Posts: 1530
Joined: Apr 2004
 
Posted: 2016-06-30 13:25
@HockeyPlayer if you're in London, happy to have a chat over coffee. As mtsm just hinted just wrote, this is not trivial, and it's faster to talk than to write... Anyway my email should be in my profile if you want to PM me

"Deserve got nothing to do with it" - Clint
Previous Thread :: Next Thread 
Page 1 of 1