This a thread intended to help my self as well as others understand the nature of volatility modeling. Disclaimers: 1. I am not an expert; my knowledge of options trading is intermediate at best. However, I do have a very strong mathematical finance background and would like to hear about some practical applications for a "good" volatility model. 2. I trade futures and stocks profitably (a rare directional option trade here and there) and I want to become better educated in options trading. So, in hopes of combining some practical experience from the option traders here with some decent mathematical models, lets have a discussion. I ask a few general (possibly stupid, so bear with me) questions. 1. What do market makers currently used as an "industry standard" to forecast volatility? Is it a GARCH - like form? Is it silly to assume an option market maker uses something other than a history of implied vola? 2. Other than mean-reversion and clustering, the two dominant ideas in many of the vola forecasting models I've seen, what other features does one include in analysing vola? That's it for now. Later I would like to discuss some of the vola models I have been working on and a possible trading application for a good vola model (i.e. vola arb). Thanks, Mike

Your questions may be better answered by others with mathfin backgrounds and pricing model experience: www.nuclearphynance.com www.wilmott.com www.global-derivatives.com Though, there are a few on ET that do have the requisite knowledge also. I am not one of them.

I am sure I am not up to your mathematical background, but I will give you some thoughts. They are only opinions so feel free to ignore them. PHD mathematicians have access to tremendous funds and computing power at the major firms and none of them have ever consistently been able to predict volatility. How do you value an option if a stock can gap up 7 SDs on a takeover, or gap down on a sudden accounting scandal? Did you notice the last SET value in the SPX? The most widely followed index in the world gapped up 15 points at expiration, the April 1480 Calls went out at .50 and settled the next morning at 5.57. Were those correctly priced by the mathematics of the pricing model? Did statistical analysis of historical or implied vol predict it? My only point is warning you of analysis overload. You can use all the math and statistics you want, but you are still trying to predict future unexpected occurrences based on historical information. How do you predict something that has never happed? The analysis is useful, but trading on it profitably is another story. The best option market makers I worked with were great traders; they did not accomplish it with "decent mathematical models."

It's not neccesarily about predicting volatility. There is a lot that can be learned from modelling the nature of volatility. By that I mean, you don't have to be able to predict when a takeover or scandal occurs to be able to model those kinds of events in order to have a better understanding of your risk today. As for SET, this is not a gap as it does not represent an index print. The procedure for the calculation of SET is well understood. 15 points or thereabouts is not-uncommon in my experience.

Moreagr, SET is the settlement value for SPX options calculated on expiration morning. TraderMojo, I would love for you to give me the dates that the SET has gapped more than 15 when the VIX has been in its current range. How long have you been trading options professionally, because a +15 has not happened in more than the last 7 years? The SPX gapped about 12.46, so you're right it did not gap 15 but close enough, and I was talking about valuing options. Either way, that was not my point. I was merely trying to explain that forecasting volatility, while a worthwhile and useful endeavor, does not prepare you to make a living trading options.

GARCH is unreliable as it cannot take into account future events like.... earnings releases / trading statment / key economic data releases, etc, etc. The best forecast of future volatility is... the option implied volatility, IMHO.

Why? I made no such assertion. I'd rather not get into some kind of semantics debate. That is the bane of many forum threads here. Suffice it to say, I don't have exact figures at hand and I don't really follow it that closely but my experience tells me that double-digit differences from Thursday close to Friday SET come along as much as twice a year anecdotally. Therefore, I thought use of SET in your original post was not a good illustration of your point and I don't think gap terminology can be applied to SET due to it's nature. If I'm wrong, so be it.

You would think so but according to one paper by the Federal Reserve Bank of St. Louis: "Research has consistently found that implied volatility is a conditionally biased predictor of realized volatility across asset markets. No solution consideredâincluding a model of priced volatility riskâexplains the conditional bias found in implied volatility. Further, while implied volatility fails to subsume econometric forecasts in encompassing regressions, these forecasts do not significantly improve delta-hedging performance."