Predicting HV - Thoughts on a Trading Model

I did an unscientific quickie on correlation between HV and price (daily close).

Attached QQQ with 60-day HV and 1-year correlation. Pos correlation during 2001, but just as neg during 2003. Tried diff lookbacks etc. but bottom line: corr all over the place (i.e. long-term irrelevant).

“A contract on the implied volatilities of traded options, along the lines being proposed by the Chicago Board Options Exchange, might go some way toward meeting the need for a volatility hedge. But changes in actual volatility and changes in the implied volatility in option prices are quite distinct. An implied volatility contract does not provide a good hedge against actual volatility.”

This is quoted from the Neuberger paper, publ. 1994.

I just wonder why exchanges 10 years since created these complex IV (VIX) contracts, but never introduced a “simple” log contract to hedge/trade actual vol. There must be a reason.....

 

non , you should of use historical IV and not HV. You are missing impact of otm puts/itm calls skew on inverse correl

 

To OP:

When you use your neural net to predict HV, what is the mse and worst error? An accuracy of 60 or 70% means nothing without knowing the potential error (i.e. the risk).

It is better to predict IV instead of HV. IMO, better to predict a distribution of IV instead of one exact value.

I am interested to hear any ideas on using neural net for trading options. Keep posting.

 

Sorry IV, I completely concur on IV/price inv correl, but thread starter clearly states he predicts HV using a neural net, presumably from just price series.

Even if one could predict with 100% accuracy, say actual vol during the next 30 days, there is no trading vehicle for this. Buying/selling under/overpriced options based on current IV is very imperfect because IV's have a life of their own, as you know.

With a Log contract if HV increases +1 you make a buck, if -1 you lose one, nothing “implied” here. I see a big gaping hole called Volatility Exchange. Any mavericks out there with more money and less time than me, contact me and let’s get this started.

 

u are right , non , this thread is about HV. My bad

 

bump

 

Because you need an infinite number of strikes and hedging those. Otherwise, you are back to delta hedging the residual risk. I would love to have access to variance swaps. Alas, I am stock in retail hell too. Even when I recently had access to serious money to trade (not retail), I still did not have access to VS. It seems to be a bank-to-bank or bank-to-hedge-fund trade.

To the original poster. Being right just about five day ahead HV is no guarantee of success since hedged profits are (underlier) path dependent. However, the closer you get to expiration, probably the better your chances are. I haven't thought it through completely even in this case, as gamma explodes nearing expiration probably exactly for this reason, i.e., to make trading the convergence of HV and IV that much harder/costly.

 

Yep, in fact if the IV of calls did not also go up, there would be an arb based on P/C parity.

I would say this is the number one thing that new traders don't understand about options, and even experienced option traders don't think through. Same strike CALLS ARE PUTS and vice versa from the standpoint of vola through synthetics. Every options market maker knows this.

 

I agree with this. The threshold of a professional-level understanding of options is when you know beyond needing to think about it that a put is a call and a call is a put because at their core they're both premium - gammas, vegas and thetas.

If you're trading options and this is not clear as day, then acquiring that understanding should be your goal.

 

As well as acquiring the understanding of the few rare cases when put - call parity doesn't hold. But there's been many a discussion of these topics here before.