Predicting HV - Thoughts on a Trading Model

[atticus]

Quote from nonprophet:

have i lost my sanity or is this the looney board?

Too much MDMA, perhaps? To the thread-starter: stat vol prediction isn't terribly relevant if implied vols diverge.

The implied vol-prediction is [loosely] governed by path-dependency -- markets go up, vols go down... to predict index vol is to predict direction. It's why banks quote risk-reversal vols. I realize the thread-starter is attempting to predict RV, but it's pointless, IMO. Here's a prediction; index vols will rise during the next session in which we see a decline of 30 on SPX.

 

[nonprophet]

Quote from atticus:

The implied vol-prediction is [loosely] governed by path-dependency -- markets go up, vols go down... to predict index vol is to predict direction. It's why banks quote risk-reversal vols. I realize the thread-starter is attempting to predict RV, but it's pointless, IMO. Here's a prediction; index vols will rise during the next session in which we see a decline of 30 on SPX.

You are using index vol and implied vol (of index options) interchangeably and apparently you realize that. If there is indeed a (empirical, not theoretical) relation between index vol and direction then I admit that’s news to me.

 

[atticus]

Quote from nonprophet:

You are using index vol and implied vol (of index options) interchangeably and apparently you realize that. If there is indeed a (empirical, not theoretical) relation between index vol and direction then I admit that’s news to me.

Mea culpa for not differentiating. The implied-vol path is dictated by mag. and direction. Market up = < implied vol[strips]/deltas/dvegas. Market down = > strips/deltas/dvegas.

In practice, ask yourself why upside calls in a rally are often such a poor-play after a significant decline?

 

[neuralmarket]

Hi all,

Thanks for commenting, this discussion has been helpful and I appreciate everyone's feedback.

RE: the comment about Binomial Validation. I understand what you're getting at and it would make sense to do that for the purposes you list. However I think in this case its not useful since the model tries to recognize volatility patterns by building on new data each week. As we know, volatility is "sticky" so rising vol begets rising vol.

The 5 day forward prediction then becomes dependent on the previous instances of volatility, so IMHO, the binomial validation won't work since it requires each instance to be independent.


RE: Other comments, the model is purely directional (which as I said is right 60-70% of the time) and it does spit out at number for HV as part of the process. that number is notorious for being wrong. The model has a tendency to get the direction correct but I wouldn't use it to predict an exact HV.

Please feel free to comment further. I will investigate the variance swaps.

Thanks.

 

[sellvol]

Quote from atticus:


To the OP -- why not trade VIX futures and/or options?

I second that.

OP - sell/buy the implied distro in vix derivs into a declining/rising HV signal. When trading pure vol, I'd take corr risk(implied over stat) over strike dependence(delta hedging) any day of the week.

 

[nitro]

Quote from atticus:

...To the thread-starter: stat vol prediction isn't terribly relevant if implied vols diverge.

The implied vol-prediction is [loosely] governed by path-dependency -- markets go up, vols go down... to predict index vol is to predict direction. It's why banks quote risk-reversal vols. I realize the thread-starter is attempting to predict RV, but it's pointless, IMO. Here's a prediction; index vols will rise during the next session in which we see a decline of 30 on SPX.

While that is an astute observation imo, it is not the case that this is always true, or even true most of the time. For example, the markets went up today, and vol went up today in the SPX options. It is also not the case that if the markets go down, that vol in SPX options automatically go up either. An extreme case of this was in the '87 crash when short calls actually lost money because their iv exploded!!!!

All that said, "vol" as measured by VIX does appear to reflect what you are saying, so that would be a good trading vehicle for this sort of strategy. Single SPX options or spreads, I would be very careful.

nitro

 

[atticus]

Quote from nitro:

While that is an astute observation imo, it is not the case that this is always true, or even true most of the time. For example, the markets went up today, and vol went up today in the SPX options. It is also not the case that if the markets go down, that vol in SPX options automatically go up either. An extreme case of this was in the '87 crash when short calls actually lost money because their iv exploded!!!!

All that said, "vol" as measured by VIX does appear to reflect what you are saying, so that would be a good trading vehicle for this sort of strategy. Single SPX options or spreads, I would be very careful.

nitro

I stated "loosely" as a qualifier. Within one deviation anything can happen. Referring to a 3-handle move in the SPX certainly isn't indicative. It's noise. I'll add another qualifier; if spot moves >one sigma my argument holds more often than not.

 

[nitro]

Quote from atticus:

I stated "loosely" as a qualifier. Within one deviation anything can happen. Referring to a 3-handle move in the SPX certainly isn't indicative. It's noise. I'll add another qualifier; if spot moves >one sigma my argument holds more often than not.

It would make an interesting study...

nitro

 

[nonprophet]

Quote from nitro:

It is also not the case that if the markets go down, that vol in SPX options automatically go up either. An extreme case of this was in the '87 crash when short calls actually lost money because their iv exploded!!!!

This is not contradictory at all, is it? I mean the '87 crash *confirms* the statement....?

That said, let's not forget the man predicts HV (RV, stat vol), and not IV.

 

[segv]

Quote from nonprophet:

I meant replicating a log contract by himself. Why can't that be done?

Because it is extremely expensive.