UM_manager, I think the concept of price tracking is viable, but prefer to do that on a finer interval using PDE's instead of NN's. Like this:
I work with a similar method. Regions, definitely, statistics, definitely. So far I have never traded it but if it is profitable I will not be selling the signals.
can you give examples of algo's you use? Show us something more than pretty charts and results and statistically speaking, shouldn't the error increase as the prediction moves further into the future? Your upper and lower limits look very parallel
It is your choice, maxpi. But other guys give recommendations absolutely free of charge. It's their choice.. Hi, nbates! I tried to use PDEs and correlations, but NNs have proved to be much better. On one of stages of my algorithm the value calculated by means of NN is used as a parameter. SwingOutOn5! Of course, "error increases as the prediction moves further into the future", pics show it!
"If price movements were predictable to everybody then markets would not exist" "If price movements were predictable to a small group of traders then markets would still exist" Spike500, 2005