No, "they" did not tell you "only". Instead, "they" told you that such a model will also tell you how to efficiently cross-hedge VIX options with each other or with SPX options. This makes market models very useful for market making or for risk-managing a diverse book. But it's not very useful for proprietary trading - it will fit the existing prices very well and will not tell you that something is out of wack.
What good is a perfectly fitting stochastic volatility model for VIX such as the Rough Heston ??
- Thread starter stochastix
- Start date
why not? if you have a calibrated model , the continuing likelihood will be high if its fits and as the regime shifts the likelihood will plummet indicating something is out of wack. there can be patterns in the calibrated parameters over time that are invisible even dealing with only the observations . when I worked on HFT I found patterns due to Nash equilibrium that were not present in underlying series
I don’t think or don’t know whether you can use BS with VIX, as I haven’t even bothered. I just treat each VIX options expiration date as a different underlying, since that’s what they are - separate underlying futures for each month, although correlated and converging onto single VIX index over time.
I also don’t think VIX can be predicted, whether on 15-sec or longer time frame, but I see that once in a while the spreads between the VIX futures are so extreme, especially during backwardation periods, that I’d get into trading those spreads using options. Though I guess I may be trading the deviation from backwardation, when certain months and related option prices are most out of sync between just couple futures/expires. Though those opportunities are rare and therefore difficult to test since they won’t show up every day. During last 6 months I only spotted the election months (October and this one/November) and more recently December 4 expiry as offering large enough opportunity to trade.
I do see your point about why not trade on 90% chance of winning since it won’t get better than that. The only issue is that everything is so correlated that when the 10% chance of loss happens then you may lose on all trades at once.
So maybe a solution would be to look for less correlation, either across different types of instruments (like commodities), ETFs, etc.; as well as trade VIX with lesser correlation to the market; as well as look for trades that may profit in either direction (strangles) or at least setups that limit your losses, or setups that may be hedged DOTM to prevent catastrophic losses. At least I’m looking at all such solutions, and even started to like certain strangle-like setups (partially delta hedged though still with certain direction so my delta rarely is zero).
I also don’t think VIX can be predicted, whether on 15-sec or longer time frame, but I see that once in a while the spreads between the VIX futures are so extreme, especially during backwardation periods, that I’d get into trading those spreads using options. Though I guess I may be trading the deviation from backwardation, when certain months and related option prices are most out of sync between just couple futures/expires. Though those opportunities are rare and therefore difficult to test since they won’t show up every day. During last 6 months I only spotted the election months (October and this one/November) and more recently December 4 expiry as offering large enough opportunity to trade.
I do see your point about why not trade on 90% chance of winning since it won’t get better than that. The only issue is that everything is so correlated that when the 10% chance of loss happens then you may lose on all trades at once.
So maybe a solution would be to look for less correlation, either across different types of instruments (like commodities), ETFs, etc.; as well as trade VIX with lesser correlation to the market; as well as look for trades that may profit in either direction (strangles) or at least setups that limit your losses, or setups that may be hedged DOTM to prevent catastrophic losses. At least I’m looking at all such solutions, and even started to like certain strangle-like setups (partially delta hedged though still with certain direction so my delta rarely is zero).
Last edited:
afaik, you first need to have a rough model of shocks that would create vix spikes, the size of the spike and duration, and then you can use that to "price" a shock in real time against the vix using your model. The usefulness in a model is greater accuracy (or estimate of error) in your scenario projections.
Well, you can use the BS on VIX, but you shouldn't, because it won't fit worth a damn.
The models I linked to a far beyond Black-Scholes, and are specifically designed for the VIX for the most part.
There are futures pricing models, as well as options pricing models, so yes they are different securities that do converge yes.
Actually, it can, but its quite detailed and technical, at least sign of the short end of the VIX futures term structure can be predicted. and the VIX itself does have some level of inherent predictability but since its not directly tradeable its harder to capture.
I don't usually have many trades open at once, usually a handful of short strangles with DOTM hedges to protect against large moves. The last big move in VIX had me temporarily down 10% but it quickly came back because i left plenty of excess liquidity so I didn't get a margin call. I should have held off and sold the combo a few days later and pocked that extra 10k premium. So now my next task is to work on software that will watch for a good oppurtunity and enter the position for me, if I tell it i want to sell a combo sometime in the next few days..
Good ideas that I'm already doing. I'm looking for series that have a good mean reversion tendency to them and plenty of liquidity so I can use the stochastic models
Cool. I'd just follow up on the first point, because I think it differs with VIX in my mind. First, by BS I mean that I hold option data in Black-type vol/variance surface matrix, while later I may calibrate using some models, but I don't see how any of this could work logically with VIX. At least ES futures or even oil/CL futures have an underlying that you can buy and hold, so different expirys may have certain types of correlations. But VIX, to me, can have uncorrelated prices across different months, creating unusual shapes and spikes in the term structure. So taming VIX would require a model that would work with random order of expiry months (or allow rearranging them), or even be applied to AMZN & GOOG at the same time because they have just as much semi-random correlation, so you'd model both their option chains at the same time. That's what doesn't make sense to me.
A side note regarding predicting VIX, yeah, maybe, I had 90% accuracy on predicting/trading TVIX/UVXY long, sometimes picking peaks intraday, which also made me money by buying UVXY calls. Later I realized that results cannot be guaranteed while losses/risks (depending on trading style) may outweigh the benefits. Now I still trade UVXY options, but quite differently and less directionally.
Last edited:
I get ya, mean/variance will not work in this instance
Yes the lack of a true underlying does make the arbitrage relations different. The VIX has been tamed pretty much, there are now several 'perfectly' fitting joint SPX/VIX models. The joint calibration problem is known as the 'holy grail' of volatility modeling and was previously known as the most difficult problem til this guy solved it about 2 years ago. See
The Joint S&P 500/VIX Smile Calibration Puzzle Solved
Nothing can really be guaranteed and I do understand the feeling of skating on thin ice you refer to .
If you were going to trade options intraday and need to stream many quotes, who would you go with? this isn't HFT so maybe I can get by with IB. i think any other solution is going to be very expensive
lmao. yeah. i know. how does that matter though? I've been trading very well recently despite all the chaos. all the world is , is one giant stochastic process
Here is a thoughtful note on the lack of the cost-of-carry of VIX
A Note on the Premiums and Discounts Embedded in VIX Futures Prices
"This article illustrates the volatile nature of the premiums and discounts embedded in the prices of VIX (Chicago Board Options Exchange Market Volatility Index) futures contracts. The fact that the underlying VIX index cannot be traded leads VIX futures to be priced more on expectations of market participants than on a typical cost-of-carry relationship. As they near expiration, VIX futures, in the aggregate, tend to trade at an increased premium, when trading in contango, and at an increased discount, when trading in backwardation. In addition, the premium in these contracts tends to peak as the VIX index nears a low, and the discount in the contracts tends to bottom as the index nears a high."
A Note on the Premiums and Discounts Embedded in VIX Futures Prices
"This article illustrates the volatile nature of the premiums and discounts embedded in the prices of VIX (Chicago Board Options Exchange Market Volatility Index) futures contracts. The fact that the underlying VIX index cannot be traded leads VIX futures to be priced more on expectations of market participants than on a typical cost-of-carry relationship. As they near expiration, VIX futures, in the aggregate, tend to trade at an increased premium, when trading in contango, and at an increased discount, when trading in backwardation. In addition, the premium in these contracts tends to peak as the VIX index nears a low, and the discount in the contracts tends to bottom as the index nears a high."