Simon,
The answer is, yes and no. That tells you a lot, doesn't it? :-)
I personally use the program nearly every day. If you are just thinking of playing around with it, then it becomes a 70 dollar toy, something that you just throw some data at to see what it says. That's all it will be.
If you are planning to seriously use the program for trading, you should get the commercial version with at least 40 neurons. In the commercial version, you get to see the last 10 days of volatility bars {something that is very important for trading} and the Max Profit trading strategy among some other features. You asked if you will get better signals from the more expensive commercial version. The answer is no. Identical nets with identical settings will produce nearly the same results whether they are trained by the shareware version or commercial version. The commercial version just lets you see what you are doing better during the time period that really counts {the last few days}, and allows you to optimize the net for maximum profitability.
In my opinion, which comes from working with this program for the last three years, if you want to seriously use this program and get decent results, the first thing you have to do is ignore most of what's in the help files. Secondly, you must have some type of program that will produce price and indicator files that Neuro can read. Thirdly, do not pollute the nets using data files of outside influences as inputs. And last, but not least, set up a verify period when you start a new net and ignore what the training set looks like, paying attention to what the verify period looks like, and only what the verify period looks like. The training set is unimportant. What you are really looking for are accurate signals in data that the net has never seen. Only then will you know whether or not the net will predict accurately next week or next month.
I am very particular with the nets that I produce. They must meet my criteria for accuracy or they are discarded. Accuracy can be only determined two ways. Using a verify period of at least 6 months or longer, or by withholding data and then reinserting it after the net is fully trained and viewing the trading signals in that data. For absolute certainty, I like the second way better. Reason being, because then I know for certain that the net has never had any access, in any way, to that data until I made it available.
If you need any more info, let me know.
Jay
|
