1 (edited by deadlef 2017-06-09 17:58:47)

Topic: How did you start get strategys?

Hi,

i like to know how you start generating strategys. Did you have bar closing and bar opening locked everytime?what you do in your acceptance criteria? how long did you let the generator running 60mins enough? what your data bars maximum or not? Whats with Out of sample?

Re: How did you start get strategys?

I think this strongly depends on person to person. Everyone has to develop his own way. I am quite skeptical and therefore I take a base that I trust. I have been looking for a strategy based on Ichimoku indicator for several weeks now. Currently FSB Pro is running on a strategy. I put the counter to 10000 with a number of indicators that are blocked and I know that something will certainly come out during this weekend. My objective is to develop a long-term strategy. I occasionally look at the results but I'll let it run the whole weekend. In addition, I have another EA that I developed as well with the Ichimoku indicator as base, but it is only valid for short duration. For that, I had to a couple of days. Generation took a couple of hours and I optimized it within a few days. All based on relatively few data (30000 bars on M15) Hence, since my EA is based on less data this will not continue to run forever. I leave it live for two to three weeks and then it's time to change. If I test this EA on much more data then I will inevitably get a big loss. As far as I experienced with FSB, good results can be achieved. It may not come without some practice and everyone has his own way here. On the forum you will find a lot. As I'm not the best, I can only say this. Let the game run without too many conditions and after half day or even after a couple of hours you'll see nice strategies appear. Also, I never used OOS first. OSS is a great tool. Try it.

Re: How did you start get strategys?

yes but i have tried now a lot of strategys . i only find strategys also simple with good fsb backtests but on real account they not profitable. also test much on demo accounts.but the test are very long. like to test for a month to check if its good. but after 2 weeks i also see that i have no strategy found yet which make profit. i tried all different things with different settings. all of them looks great in fsb but on real not. i have let the generator generate 5 days a week but not better results. i have tried 2 slots on open and close..of this i have generate from the top ten another strategys with one more slots each to get better results from the best..but also this not working on real accounts. dont know which are the best configs. i have tried to do my own idea of strategy but also this not good results.

Re: How did you start get strategys?

I keep the number of indicator slots to a minimum, 1 or 2. The strategies I get seem to work ok. Too many indicators and there will be curve fitting in future.
I also work with daily charts or 4 hour charts.....
Learn how to create strategies on the higher time frames and then move to the lower timer frames

Just my opinion.

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

Re: How did you start get strategys?

I still trade on demo with my strategy portfolio for 8 months, but it is with small profit, so not very bad. It means that I am doing something correctly. Hope chosen best performers will do better alone in new account in the future smile My guidelines would be: to use as low variables(indicators) as possible, use lots of data, at least 50 trades for each variable, always use some chunk of data for unseen confirmation, do not trust generated/mined indicator values - change them to default and slightly re optimize. If you can't get similar results with unseen data probably strategy was curve fitted while data mining and has no prediction power.

Re: How did you start get strategys?

i also let generate strats with e.g eurusd h1 minimum 0,01 - max 0,05 lots. say 2 open and 2 close slots. i tried backtest data with 5 years and 1 year. i get much results with good profit but only in fsb. on real not. also i didnt get results with more than  2 profit factor. after generate i check and filter out with strategys. than i do optimization with 40% or 25% oos. after that monte carlo. all looks great. but when i use in demo account nothing looks great. Also i use Stopp loss. cause i tried without but too much losses.

Re: How did you start get strategys?

How long you trade demo live your strategies? What point is then you decide that strategy does not work? Maybe you turn off your systems too quick.

When I create strategy I set lotsize/risk for 10% max draw dawn. And after system reaches 15% I would turn it off.

Re: How did you start get strategys?

i test them in demo accs for 4 weeks. after that i check it with mt4 tracker and look which of my portfolio are makeing best results. but even of 100 strategies nothing good and profitable but on fsb showing awesome blalance graphs.

Re: How did you start get strategys?

If it looks too good to be true, it often is. Can you post some examples of awesome graphs? There might be some bug involved.

10 (edited by yonkuro 2017-08-02 22:22:43)

Re: How did you start get strategys?

I always lock bar opening and closing slot, I dont use OOS. I only use H1 TF, besause it's the largest TF which I think is independent from timezone difference. I also block some indicators which I think are sensitive to timezone difference. Then I start to generate.

First I use 0 slot on closing logic, so I can get a strategy with good entry point, then I optimize it. Then I lock the opening slots, increase the closing slots, and run the generator again, and then I optimize it again twice, first I optimze the indicators, second I optimize it to find SL, TP and BE.

The most acceptance criteria I pay attention are SQN and stagnation.

do or do not there is no try

Re: How did you start get strategys?

yonkuro wrote:

I always lock bar opening and closing slot, I dont use OOS. I only use H1 TF, besause it's the largest TF which I think is independent from timezone difference. I also block some indicators which I think are sensitive to timezone difference. Then I start to generate.

First I use 0 slot on closing logic, so I can get a strategy with good entry point, then I optimize it. Then I lock the opening slots, increase the closing slots, and run the generator again, and then I optimize it again twice, first I optimze the indicators, second I optimize it to find SL, TP and BE.

The most acceptance criteria I pay attention are SQN and stagnation.


and with that you have profitable strats?

Re: How did you start get strategys?

yonkuro wrote:

I always lock bar opening and closing slot, I dont use OOS. I only use H1 TF, besause it's the largest TF which I think is independent from timezone difference. I also block some indicators which I think are sensitive to timezone difference. Then I start to generate.

First I use 0 slot on closing logic, so I can get a strategy with good entry point, then I optimize it. Then I lock the opening slots, increase the closing slots, and run the generator again, and then I optimize it again twice, first I optimze the indicators, second I optimize it to find SL, TP and BE.

The most acceptance criteria I pay attention are SQN and stagnation.


i just checked my demo acounts and strategys running and see a big mistake. I thought that every strategy i export will have automatically a different magic number. So i export 10 Strategys and they all have the same magic number. So i have never get results cause when one magic number reaches his maximum lotsize the other strategys wouldnt do anything.

I have ever thought that FSB randomly make a magic number.

13 (edited by yonkuro 2017-08-04 16:25:22)

Re: How did you start get strategys?

If you use "Export All Strategies" button, you'll get same magic numbers, if you want unique magic numbers, you should use "Export Expert Advisor to MT4/5", and export the strategies one by one

https://s2.postimg.org/q5gfagx6t/SS_04082017.png

And about my strategies, I got what I expected, however I still have to backtest them on MT4 which have same historical data as FSB Pro, just to make sure that the indicators I use work properly and give similar result to FSB.

I always use bar opening so I can backest them accurately in a short time,

do or do not there is no try

Re: How did you start get strategys?

I was afraid to make FSB exporting magic numbers to the experts because, as Footon said "we will become lazy to check them". We always have to check the Magic Numbers before running an expert.

Trade Safe!

Re: How did you start get strategys?

yonkuro wrote:

I always lock bar opening and closing slot, I dont use OOS. I only use H1 TF...

It was interesting to me to read this from you.  After much trial and error these are also the conditions I've settled on -- at least for now.

What follows below is just my two cents based on my experience using FSB Pro and EA Studio for a few months.  So, don't take it too seriously.  And, of course, I am interested to learn more about other approaches...

I prefer H1 for two reasons:
1. I think the data is sufficiently reliable -- as compared to M5 or M15.
2. I use live micro accounts (rather than demo accounts) for testing since I just don't have the patience to wait weeks or months before moving an EA to a live account.  Using MT4 Tracker it seems that between 40-60% of the EAs created by FSB Pro and EA Studio are immediately profitable.  Even though the micro accounts may contain some poor performers, overall they continue to be profitable.

Regarding OOS -- I've also stopped using that.  In addition, I've found back testing for long periods (e.g. > 2 years) does more harm than good.  When you back test for long periods the strategies you generate have poorer statistics than when you test for shorter periods (e.g. 6 months).  I think that is because you are trying to fit more data points into a profitable formula.  The more data points, the poorer will be the fit.  But if you back test for 6 months you will get many strategies with good statistics.  Since Popov's software makes it easy to generate strategies, I no longer have to become attached to a particular strategy.  I'll use them while they perform well, and over time simply replace the poor performers with new ones.  In my hands, using OOS provides no advantage.  I want to generate strategies using the latest and greatest data in the short-term with the understanding that, over time, they'll eventually get replaced by better performers.

One last point -- I agree that OOS allows you to select for more robust strategies -- i.e. ones that will stand the test of time.  But those come at a cost -- poorer statistics.  With Popov's software we no longer have to be concerned with only using strategies that are profitable for long periods.  I'm now content to use strategies for shorter periods, knowing I can replace poor performers at any time.

Re: How did you start get strategys?

Thanks for your opinion sleytus, and I really like your tracker smile

do or do not there is no try

Re: How did you start get strategys?

sleytus wrote:

Regarding OOS -- I've also stopped using that....

Very interesting, because I came to the conclusion, that all generated strategies are kind of "curve fitted" whatever data I use. So now I always use maximum OOS and check how the EA performs in the range of the (for it before) unknown data.
I am really "shocked" sometimes, how similar good looking strategies suddenly differ extremely in the "test range".
So I only trust OOS checked EA's, but that is only my way.

Re: How did you start get strategys?

bru1 wrote:

...So I only trust OOS checked EA...

What you say is true.  I've had the same experience -- so it is interesting to also hear from others.  And I used to always use OOS, but no more.

Yes -- all strategies include curve fitting.   Since that is the case, then I think it would be more valuable if OOS was reversed -- i.e. for 33% OOS the oldest 33% of the data was OOS and the most recent 67% was "in sample".   Consider this example -- suppose you backtest for 3 years using 33% OOS.  That means the first two years are "in sample" and the last year is OOS.  Since curve fitting is part of the game, why would I want to use a strategy whose settings haven't been optimized for the most recent 1 year worth of data.  Yes -- you would select only those strategies that performed well over the entire 3-year period.  But you would be missing out on those strategies that would have performed well over the most recent year if you had included their most recent data when optimizing.

Another example -- suppose I have two strategies that I back test for 1 year.  One uses 33% OOS and the other just "in sample" data.  And in the end their statistics look very similar -- let's say they look identical.  Which is the better strategy and why?  Would you trust only one in a live account, or both or neither? 

Prior to Popov's software I would be inclined to spend a lot of time creating one EA with the hope of using it for a long time.  So, any test -- like OOS -- that improved "robustness" was valuable.  But using Popov's tools I now have more EAs than I know what to do with.  So I no longer consider a strategy a bad one if it performs well for a few months and then I have to replace it.  I think that is one of the major paradigm shifts that Popov has introduced -- either intentionally or unintentionally.

I know this is getting long, so I'll wrap it up.  I've noticed that settings make a big, big difference.  That is, a strategy is composed of 1 or more indicators.  With the proper settings, it works great.  And if the settings are off, then it looks terrible.  My point is that small changes to settings can make a big difference.  This is why people advocate to "refresh" your strategies from time-to-time.  If you are one who is willing to refresh your strategies, then I think that is also consistent with not using OOS.

Again -- just my two cents.  Since I'm relatively new to forex then I want to better understand what really is behind the different trading strategies that people use and how those now fit in with FSB Pro and EA Studio.