There is certainly no lack of methods for predicting time series. Traditional econometric models (e.g. ARMA) as well as various GARCH formulations have been extensively applied. More recently models based on Pareto-Levy (nonlinear) distributions incorporating embeddings in phase-space reconstructions have had some success.
Neural networks, genetic algorithms, and their A.I. spinoffs are also used with varying results. Often directional traders (myself included) reach a frustration threshold, regarding lackluster performance, beyond which they capitulate, and become a statistic. My guess is that a certain subset of these discouraged traders migrate to the options arena, on the premise of it’s non-directionality. Curiously, though, volatility trading at it’s core, comes full circle back to directional forecasting, just not of an “underlying” but of future volatility.
The inclusion of timing and/or cyclical analysis, may provide an integral complement, to both standard and non-standard time series analysis. My research indicates enhanced system performance characteristics when timing is used. Certain types of “reversal” systems follow naturally from an awareness of time targets. Saturday, June 4,2011 I will give a Webinar presentation where I will present certain time projection procedures.
You can register for our Free Webinars using the link below. Space is very limited on this one, so please do it now.
