Topic: Generator idea..

I like to use as few indicators as possible because the market shifts often thereby requiring maintennace and optimization.

Today I am working with Donchian Channel only...... I blocked all of the other indicators and the stop loss and take profit and break even.

Leaving me with Donchian Channel  only in opening and opening logic and closing logic...... and let the generator do its thing until it finds a solution that I want to use.

If there is no acceptable solution, I may add stop loss and see what happens, or break even or a combination, the idea being to use as little as possible. The generator is really fast and powerful, sometimes it takes quite a while to find a solution and other times it is fast.

I like to use the Heiken Ashi and Round Number, the price action indicators, the oscillators have caused me problems in the past.

I limit the number of logic conditions to 2 at the very most, more than that will be huge problems, I actually prefer to use one only.

Just my thoughts on the generator, someone else may have better ideas.

My 'secret' goal is to push EA Studio until I can net 3000 pips per day....

Re: Generator idea..

Very interesting and motivating ideas.

I have developed hundreds of strategies suggested by the generator (in FSB freeware).  After "moderate" (I hope) analysis and optimization I started trading a handful of them in demo and real. Nevertheless, after a couple of months of trading I am a little worried - nearly all of them perform worse than expected, in comparison with backtests (that included also 20% OOS periods).  I know, you may now say that I have over-optimized or that forex markets have changed recently or that I am too impatient or something similar.
But I mentioned this experience rather as a little irrational feeling and not as an exhausting rational analysis.

In contrast with "generator-made" strategies, I have better results with strategies developed manually, without using generator, even if also here I am still not quite satisfied.

Interesting is that using both methods I have achieved better results with index CFDs in comparison with forex so far. Does it mean than currency pairs behave more "chaotically" ....? Or is this only an abnormality caused by my specific approach?

Most of above mentioned strategies trade only with 1 contract without adding positions. Can you say, if you use adding positions and how much important is it for the success of your strategies? I ask this also for the reason that generator (and optimizer) do not allow to play much with existing features dealing with multiple positions.

Another thing is combining strategies traded in different timeframes of the same currency pair (or CFD). How important is this for the success of your strategies? Unfortunately, here I have not found any help from automatic tools provided by FSB; probably it is too complex. Of course, the professional version has a great Portfolio feature, but this is nothing that can be used together with generator and optimizer I think.

Re: Generator idea..

I would not be one to say you have over optimized...... that term can be mis-used, I use the optimizer to get better solutions.... perhaps optimize to get a better stop loss and then do more generating.

I agree with the idea of constant positions.... I have seen great differences when changing the position mix, my feeling is that changing positions can be similar to martingale.

I suggest that if you have more than one robot in one time frame for the same instrument that you use different types of indicators on each...... eg make one for price action and one for oscillators and one for trend and one for volume.... different approaches might act like insurance.

I like the cfd's, particularly Google, nice moves almost every day, whereas the forex pairs have not a lot of volatility recently.

For a test..... I am running a robot with only Macd cross on  4 hour charts, I just started this.... using only Macd cross, stop loss, take profit and trailing stop and data from Jan 1 2007, demo account... 

If It shows anything worthwhile I will share the results.... if it goes bust. hmmmm    LOL

I am scared of time frames less than 4 hours. mainly because I have burned repeatedly.

I do not like daily because I do not have enough data, I guess I should search for more data.

Do you try 'long only' and 'short only' on the same time frame?

Do you limit the hours for trading specific pairs? 

I have had good luck on one hour charts by limiting days of the week and trading hours, but have decided that 4 hours is better for me.

Just some thoughts.

My 'secret' goal is to push EA Studio until I can net 3000 pips per day....

Re: Generator idea..

You need to do most to understand the concept behind the each strategy.

If I am using the generator then I try to build a strategy towards a specific idea or concept. I try to simulate and backtest the situation with several ways to get deeper understanding about the drawbacks of the strategy.

Market has a continiously changing motion, it's quite a difficult to build exeptionalities to prevent strategy to run all the time. Sometimes it is necessary to manually swiths off or on trader when right market conditions apply.

Re: Generator idea..

Bold

I wonder if you have some ideas as to what software improvements would improve analysis based on your experience.

My 'secret' goal is to push EA Studio until I can net 3000 pips per day....

Re: Generator idea..

Blaiserboy, zenoni,
Thanks for your ideas. I was aware of majority of them, but some are new inspirations to think about. Here my thoughts to some of them:

[Blaiserboy]: I like the cfd's, particularly Google, nice moves almost every day, whereas the forex pairs have not a lot of volatility recently.
[bold]: Glad to hear that I am not alone who has made good experience with CFDs.

[Blaiserboy]: I am scared of time frames less than 4 hours. mainly because I have burned repeatedly.
[bold]: Theoretically I agree. But in practise my best systems (in real) run on 5 min and 15 min! Even with oscillators that are generally supposed to be tricky. Isn't it crazy?

[Blaiserboy]: Do you try 'long only' and 'short only' on the same time frame?
[bold]: In forex not, because I suppose they are symetrical. With CFDs sometimes yes, especially in longer timeframes (stock markets usually fall quickly than they rise).

[Blaiserboy]: Do you limit the hours for trading specific pairs?
I have had good luck on one hour charts by limiting days of the week and trading hours, but have decided that 4 hours is better for me.
[bold]: Sometimes I try to limit hours/days, but mostly not. I have not found enough benefits for real trading yet.

[zenoni]: You need to do most to understand the concept behind the each strategy.
If I am using the generator then I try to build a strategy towards a specific idea or concept. I try to simulate and backtest the situation with several ways to get deeper understanding about the drawbacks of the strategy.
[Blaiserboy]: I suggest that if you have more than one robot in one time frame for the same instrument that you use different types of indicators on each...... eg make one for price action and one for oscillators and one for trend and one for volume.... different approaches might act like insurance.
[bold]: Basically I try to build and diversify my strategies this way as ground ideas in the beginning. Then I start to use generator and optimizer - and, paradoxically, here lies a "dangerous trap" for me, that is not visible during development, but only later, after some real trading period. As I let generator (and optimizer) to "improve" initial ideas and add new instances, borders between initialy diversified ideas tend to fade discreetly. But for me it becomes quite hard to distinquish where the improvment is still healthy and where no more. First  because generated (optimized) equity curves and their parameters might look very optimally, but they tend to perform in the similar way as other, originally diversified, strategies step by step. And second, repeated usage of generator (optimizer) produces more and more "nice" similar strategies; their number growths in a geometric way. Since every single sub-strategy requires some manual effort, it quickly starts to be very time consuming and untransparent. I must always solve, if and what of sub-strategies are already optimal enough and where to stop with futher "improvements". For all these reasons I welcomed Blaiserboy's interesting ideas and experience that also strategies with only few parameters can continually work profitably.

[zenoni]: Market has a continiously changing motion, it's quite a difficult to build exeptionalities to prevent strategy to run all the time. Sometimes it is necessary to manually swiths off or on trader when right market conditions apply.
[bold]: This statement also inspired me to express some improvement, as Blaiserboy asked. I will try to summarize them in the next post.

Re: Generator idea..

Blaiserboy wrote:

Bold

I wonder if you have some ideas as to what software improvements would improve analysis based on your experience.


Today's discussions inspired me to express some ideas on software improvements. Please take them with reserve, I write them without respect, if they have real sense and if they could be really programmed (Sorry, especially to Mr. Popov, if my ideas are too naive and unrealistic. I am not a programmer, but just an excited intermediate trader :-)  ):

1. Placing OOS period before backtest period
Usual backtest approach is first to run generator on defined past period and then to check on an OOS period in the "(un)known" future. However markets always change their structure, sometimes quickly, within months or even weeks. An example: generator runs on a past period from 2010 to 2013. Then I can have a look how this strategy would perform in the future (= 2014, until now). The weakness of this approach is that the strategy is thus optimized to past market conditions 2010 - 2013, but it does not reflect current conditions in 2014 that are in fact more important for successful trading of the developed strategy.
IDEA: Would it be alternatively possible to place the OOS period before the test period? I.e. to run the generator backtest from 2011 to 2014 (now). The robustness of resulting strategies could then be roughly checked in OOS periods 2010 and earlier, depending on available historical data and developer's intention. This past OOS data could also partly indicate, if the backtested period is (not) over-optimized and how much.

2. Equity curve trading
It is an eternal dilemma of every developer/trader, if and when to stop the strategy, as soon as it does not perform well (equity stagnates or falls). And when to start trading again, if the situation improves. There is a known "mechanical" approach for such situation, called equity trading, if I am not wrong.
IDEA: Would it be possible to apply e.g. x-period moving average (or some similar useful indicator) over equity curve? If the equity line would fall under the moving average, trading of the system would be stopped/interrupted until the equity returns above MA again.

3. Multi Markets also for different timeframes
There is a Multi Market feature that allows to compare how the developed strategy behaves in different markets.
IDEA: Would it be possible to make the same feature for different timeframes within one underlyings?
First it would be able to better cross-check the developed strategy, how robust it is. If it performs (relatively) well also in other timeframes.
Second, in some sort of strategies it would be possible to check, if the strategy might diversify trading (= stabilize risk parameters) when using the same strategy concurrently in several timeframes.
Personally, I have developed some strategies that perform well in several neighbouring timeframes with identical or only slightly changed parameters. But it took me much longer, without such feature.

4. Automatically generated Portfolio of Strategies
In FSB professional there already exists a nice feature that shows the equity and statistics of the portfolio of manually selected strategies.
IDEA: Would it be possible to automatize the Portfolio feature? In an ideal case it would help to assemble two or more several strategies that fit best together and create the best portfolio equity line.
Example: There might be a collection of 20 strategies of one underlying in different timeframes. The "Portfolio Generator" would check all their combinations and offer the list of best of them.

Re: Generator idea..

Some excellent ideas.!!

Thanks for this.

My 'secret' goal is to push EA Studio until I can net 3000 pips per day....

Re: Generator idea..

I also also agrees that there are excellent ideas.

Thank you.

The work on my FSB project as a long and continuous project. I may say that I'll continue developing it for the foreseeable feature. I'll use these ideas for sure.

Re: Generator idea..

I wonder these are some of the very fantastic ideas given by any trader. I would really like to thank you for this. Also, I am surly going to follow each of your posts mate.