I guess this is important since vol is mean reverting. My question is, what data do you use to calculate it? Like say I took the daily HV30 to get the mean. Is that going to tell me where vol is going to revert to in the next 60, 90, or 180 days? No right? Is it as simple as using an HV that is the same time period as what you are predicting for the future? Like if I wanted to speculate where vol would go in the next 60 days, then I would take the mean of HV60 right?
Natenberg advocates using the same time period, eg if the option has 57 days to expiry, use 57 day HV and IV. Both he and McMillan emphasise looking at historical IV as well.
What about the monte carlo probability calculators? I'm referring to the ones that spit out the probability of touch and settling OTM, not the option pricing models that use monte carlo simulations. One of the inputs is volatility so should a person use HV57 if the life of the option contract is 57 days?
It is really about identifying if risk premium built into an option is rich or cheap. There are numerous ways to look at it and each person convinces himself that his is the best. Important factors to remember and take into account are past/upcoming events (e.g. if there was a 20% earnings day in the recent stock history, is it still relevant?), macro environment (e.g. if VIX was 20 but now is 15, lower implied might be justified) etc.
I'm not an expert here, but based on what they both say that would be logical. Hoadley has the Monte Carlo probability calculator built in and it uses the volatility that you associate with the underlying. Thus far I have been deciding that based on 7 day, 21 day and 63 day HV that I track, typically using the 21 or 63 day depending on what is happening in the short term. This for options 30-60 days to expiry. I had thought to make the calculation dynamic so that I can vary the period at will, but I have not got around to tweaking the spreadsheets.
This is a very interesting point, and perhaps a case for excluding volatility spikes from the calculation, unless one expects more of the same in the coming period.
These 2 are not necessarily the same. With 30 days of historical data you get an estimate of the "average" volatility over this period. Now if you want to get an idea of the level of mean reversion (assuming that you believe in such a model) then you need a lot more data depending on how quick or slow it reverts to this mean. Maybe in the order of 4-5 years of daily data? Have a look at this link from the expert himself with a lot of practical data too: http://vlab.stern.nyu.edu/analysis/VOL.SPX:IND-R.GARCH