>Are these results based on closing or intraday prices? How does
>one factor into the results poor execution or slippage?
The results are based on closing prices, with no slippage or commisions entered. NS Trader does allow you to incorporate these, but I chose not to in order to keep it simple. Besides E-Mini commisions through a discount broker are about 1 tick or 25 dollars. The range in recent days has been 5 to 10 points at least, so it's easy to cover the commisions.
>At first blush the results look fairly good. Can you critique
>why you are less than satisfied?
If you look at the trading results, the first two months are poor because of a trending market. This net seems to work only in a choppy market and if you look at the results in chart format, it always comes in a day or two late. I may be overly critical, but when you design a neural net you are never really finished. I also haven't worked on the trading strategy to incorporate trailing stops or exits.
>Can you describe your input variables without giving away too
>much of your system?
Acceleration, Simple Moving Average Offset, Standard Normal, and an Acceleration of the NYSE Advance Decline. It required *a lot* of trial and error work, along with correlation analysis in Excel. But I am definately getting closer to a viable trading system. At $50.00 per SPX point, the system will definately pay off. In fact the current net made close to $15,000 on a $5,000 investment in 6 months. Now that includes the Oct 27 dip which is abnormal, but I think that it still preformed well even without the dip.
And I do know that trading a system in real time is very much different than analyzing previous history. But I think the net would still make a tidy profit...
Optim |
