Topic: many questions

Hello! How to me to make that order would open in an opposite direction, from signals of all trade system. For example when signals of system shows to buy, it has to sells. Not to confuse to reverse. One more, if the history of smaller timeframe isn't loaded, whether strategy will be correctly generated. Example generated strategy tf15, not load 1m and 5. I understand it depends on opening points, it is possible in more detail in what cases it is possible not to use the lowest tf. Possibilities of the program are boundless, but it is difficult to create working system, often there is a falsification under history. OOS not always helps. I will be grateful to councils as to avoid it.  Thanks!

Re: many questions

1. Directional logics in FSB are mirrored. This means if long=rises, then short=falls. If you want to change the direction, then change logics. I think you are trying to make a consistently losing strat into a winning one by changing the opening direction - my top tip to you, don't waste your time on it, it is very hard to achieve because of trading costs!

2. I take it that by "strategy generation" you mean backtest reliability. Well, you can measure it by the number of ambiguous bars, to reduce them you must use the scanner, which loads lower TFs and traces the trades with higher precision resulting in a more accurate backtest with lesser or none ambig. bars.

3. Can you elaborate more on "falsification under history"? I'm not understanding you unfortunately.

Hope this helps a bit!

Re: many questions

1 . Yes, from loss in the winning. Here just it is necessary it isn't necessary to change the signal direction, and it is necessary to change the position direction. Not everywhere it will bring at a victory, but in rare instances very much even can.
2 . With accuracy and the scanner everything is clear. Question in what cases it is possible to do without smaller tf. And whether this strategy will work at practice.
3 . The bringing, indicators are arranged under history. I tried to create strategy for 1m and 5m, used OOS of 30%, 50000 bars. In a tester showed good results, but with an exit of a modern history, merged at once or gradually. As such it is possible to avoid

Re: many questions

2. The answer is when there is no ambiguouty. If there are no ambig. bars, then scanning lower TFs don't add anything to the test.

3. I get your point, but it is something that comes along with this business. One must find a way how to make the strategy building efficient, recognizing the problems and pitfalls and trying to solve or avoid them. You might want to switch to walk-forward testing/optimizing, it can give a better and more clearer outcome.

5 (edited by petr.vhs 2013-03-24 04:53:38)

Re: many questions

2 . Thanks, i understood.
3 . Approximately and i do, only for check. how to apply it to creation. Optimization too will adjust only to a certain piece how to find an optimum set for different pieces. How it is done by you

Re: many questions

The following post explains it pretty well.