Which amount of data (observation) is recommended, in order to receive best results, 4 or 10 years ,other ?
Lots of research on the subject and you can google it. No single answer. For standard GARCH, it's probably good to have arnd 1000 samples or more... Less might be needed for GJR-GARCH or other flavors.
Can you explain how to turn your GARCH data into a profitable trading strategy? I've tried all types of volatility measures to get an edge, but it seems its more important in how you implement the strategy of taking the Garch measures and generate the various trading signals from it.. and this seems to have nothing to do with Garch or Egarch, or any other type of volatility measures themselves.
The problems I have found with statements like this come from my own experience. Sometimes a tool is dismissed as not useful, when in fact it is some other input that is causing the statistic to become independent, and hence your conclusions become invalid. I rate this the number one enemy of the quantitative trader once he has a little enough knowledge to be dangerous, keeping variables dependent so that statistical inference can make sense. It is not hard, but just like any other technique, it requires knowledge and experience.