since GARCH modeling is based on the past, I don't see any real difference between GARCH modeling and technical analysis >>Icecloud,it seems to me the issue is not specifically GARCH then, it’s the Calibration: Should we use historic data or not? Paul suggested the following in the “GARCH and Vol Swap” thread: Calibration to market data. By this I just mean make theta a function of t so that we can fit market prices of various contracts in the usual way. >>

Icecloud/Reza/Wilmott: It is clear that volatility is persistent, and therefore it can be predicted more easyly than market direction....although some famous traders, like Bernie Schaeffer, think exactly the opposite.What I have seen is that most traders use technical analysis. I know this is a page for and by quants, please dont hate me, but I am doing my simple and methodless research and I guess models dont work so well. I have been reading old reports provided by the analysis departament of the place where I work and simple compare them to real volatilities and made a ranking of it. Well, As the typical post-hoc economist I know now why we dont make so much money. No method, garch, egarch, agarch, wighted and non weighted historicals ranked persistently amongst the best or worst ones. I have a certain simpathy for arma on volatility,....just because when I think about them I guess they have it all: they consider clustering, at least for the short term, because they take into account past observations, and they also consider residuals. And in fact when i try this models on a simple excel sheet I have built, they do work and fit the series well. But out of the sample they look horrible...just like the rest of the models.I guess one great paper to be written, or many of them, is about the protitability of these models, and even technical analysis on volatility, and wow if we compare one another. I mean a paper like the one by lebrock on moving averages. I guess you people, who are the smart and wise ones, could do it, dedicate it to me, and send me one copy. Thanks!

You know what we need on the Forum? We need a dedicated postdoc to finish off all the brilliant ideas that we keep coming up with!!! Anyone here from a bank and senior enough to fund some labour?P

That would be the Wilmott Chair in Quantitative Finance. If he does nothing productive after a 6 month stint, we could rename it the Hamilton Chair in Quantitative Finance and give him one of 5 rotating icons.Seriously, a dedicated patrons programme funding a Phd post doc student would be a great solicitation!

Wilmott....you are the senior, rich and famous here.....ergo....

It is clear that volatility is persistent, and therefore it can be predicted more easyly than market direction....although some famous traders, like Bernie Schaeffer, think exactly the opposite. >>Isn't Bernie Schaeffer the man who writes a report with trading recommendations? What's the track record of his recommendations?

Omar,I know he was ranked very bad by fortune, anyway he is famous...maybe for something he did in another business or a previous life. But everyone doing directional trading with options dont trust vol. forecasts...and there are many.

A couple of points on GARCH, volatility forecasting, and technical analysis:First, on the topic of volatility forecasting, everyone should read the above provided paper by Andersen & Bollerslev, the fundamental point is applicable to all volatility forecasts: make sure you are measuring the forecasts against the right measures of observed volatility. I am not saying that their measure of observed vol is necessarily perfect for all occasions, but it clearly dominates the more traditional squared returns methodology. As for comparing GARCH, 2-factor models, et al., vs. technical analysis for forecasting performance, the real limiting factor is rigorous definition of the technical measures used. Before grad school, I was a derivatives broker in Europe. I have used, and to a certain extent, still use technical analysis. But most technical analysis is subjective, which makes it difficult to rigorously analyze. If someone has a specific technical forecasting tool which can be objectified that they would like to compare to more traditional forecasing methods, drop me a line. With a bit of luck, I think I could find a bit of funding to study it.Matt