Contrary to the movement of price, statistically volatility indeed is predictable. A simple GARCH model will be effective.
lol...... you are not serious... you can predict stat vols??? how big is your hedge fund lol.... re read the question again please....
There are definitely some predictable events that will result in semi-predictable implied volatility movements. Earnings, dividends, acquisitions, crop reports. But, in the absence of predicting the end result of the volatility event, profiting from the predictable IV movement is near impossible.
What about tail risk? Prediction is to within a degree of reliability, not an absolute. Are you saying I can't predict the sun will rise tomorrow, because there's a "tail risk" it won't?
You can't "forecast" implied volatilities from historical data, because IV is derived from current options and underlying prices. Hume's Problem of Induction... no amount of past data provides certainty of the future. As far as predicting future volatility, I think GARCH(1,1) is widely used. Not the most fruitful exercise for the individual trader....
I wouldn't call it forecasting but, simplistically, if for example the near term options rise quickly to within a few percent of their highest IV value over the last 5 years, then you could probably make the prediction the IV will drop to lower levels relatively soon. Making money from that observation is another matter, as is will the IV rise a bit further before dropping? Who knows?
Isnât the problem with GARCH, as used by LTCM, is that it depends on known, previously observed, and therefore quantifiable data? The unknown or unobserved is, obviously, absent. Humeâs philosophy (problem of induction) was an attempt to add intellectual legitimacy to his friend Bishop Berkeleyâs attack on Newtonian physics, specifically the absence of a superior guiding the cosmos. Berkeley was also hostile to calculus. Induction provides a degree of certainty, not absolute. Grant.