Most software packages has a GARCH (ARCH + ARMA ) forecast function. I am not sure how to use it. If it forecasts future Variance, it doesn't tell the direction of the movement. Variance is defined as : low variance means closer to mean, high variance means far from mean. But there is no negative Variance. How it would be helpful to let's say future value of SPY index ?
This is above my pay grade, but if you are really interested, read this. Maybe it will help: http://www.stern.nyu.edu/rengle/GARCH101.PDF
The GARCH number is just another input into a fancier model of much greater complexity...PhD level stuff. Using Garch by itself....you might as well use a moving average oscillator. It is interesting how the rise in comp sci/quant finance in the recent decade has seemingly debunked technical analysis, when technical analysis was created by engineer/quant types who were also working with the very same underlying math. Anybody of you certified eggheads/academics think this is worthy of PhD thesis? Is it done? Overdone?
It is interesting you mentioned "another model" and "moving average oscillator" Indeed one example I could find on internet is adding GARCH forecast to VMA to predict S&P .
Actually, I thought they were the same except that GARCH uses the statistical measure of variance. I once gleamed over a research paper that I could barely understand and it said EGARCH was best. But, take that with a grain of salt. It is the same as saying EMA is best. Try to read Eaun Sinclair's discussion of GARCH in Volatility Trading. It might uncover something useable. His writing is advanced yet very pragmatic which makes him a favorite options traders.
if you are trading options, maybe having a forecast of future realized vol would be helpful. if trading trends or chops, maybe being able to quantify strength of a move would be helpful. if trading baskets, maybe having a sense of covariances would be helpful. etc.
I use GARCH as an overlay to overall volatility measurements. In normal-speak, using it in concert with something more straight forward like VIX. If you're long (and/or trying to avoid being short) volatility it's more predictive than a single variable. "Rule Based Investing" by Chiente Hsu goes into it quite a bit. Amazing book in general which really challenges the conventional wisdom regarding diversification.