Topic: how to build strategies with a high number of trades

Hi,
As others have pointed, using EA Studio with the Portfolio EA feature in combination with MT4 Tracker is a good way to find strategies that have good probabilities to continue working well in the future.

I'm trading in demo accounts hundreds of strategies at the same time using the Portfolio EA feature of EA Studio and then, after some time (some months or number of trades), selecting only the strategies that performed better in demo to move them to a live account.

The issue is now when to move the strategies from demo to live. I've read some post from some of you askingg your demo strateies at least 100 trades in demo before moving them to real. Or at least 3 months in demo or similar.

So far, my strateggies have been in demo for 3 months but my problem is most of them do not have more tan 20 to 30 trades. Some have 40 trades, but no more.  I'm usually using H4 timeframe to build them, and although it seems I'm getting some nice strategies I will have to wait at least  more months before getting a bunch of strategies with a good number of trades.

I wanted to ask what settings you think it is better to use (or what settings you are using) to get strategies with a lot (more) trades. I'm using H4 timeframe and I usually use a selection criteria of SQN=2 to filter them. My broker gives me data from 2010 to now (8 years) when I use H4 timeframe, but it gives me only 2 years of data in the H1 timeframe. I've also used the close ath the end of the session feature (nice feature) and this way I try to force my strategies to open and close their operations fast (hoping the strats to have more operations). I'm asking my H4 strategies at least 300 hundred trades (and SQN=2) to pass my selection criteria, and I usually get nice strategies. If I ask more than that it seems the strategies are harder to get and they are not as nice as before.

What settings / timeframes are you using with EA studio to build portfolios? What number of trades do you think it is statistically significant to move a strategy from demo to live?

Thanks all in advance
best regards

Re: how to build strategies with a high number of trades

rjectweb wrote:

Hi,
As others have pointed, using EA Studio with the Portfolio EA feature in combination with MT4 Tracker is a good way to find strategies that have good probabilities to continue working well in the future.

I'm trading in demo accounts hundreds of strategies at the same time using the Portfolio EA feature of EA Studio and then, after some time (some months or number of trades), selecting only the strategies that performed better in demo to move them to a live account.

The issue is now when to move the strategies from demo to live. I've read some post from some of you askingg your demo strateies at least 100 trades in demo before moving them to real. Or at least 3 months in demo or similar.

So far, my strateggies have been in demo for 3 months but my problem is most of them do not have more tan 20 to 30 trades. Some have 40 trades, but no more.  I'm usually using H4 timeframe to build them, and although it seems I'm getting some nice strategies I will have to wait at least  more months before getting a bunch of strategies with a good number of trades.

I wanted to ask what settings you think it is better to use (or what settings you are using) to get strategies with a lot (more) trades. I'm using H4 timeframe and I usually use a selection criteria of SQN=2 to filter them. My broker gives me data from 2010 to now (8 years) when I use H4 timeframe, but it gives me only 2 years of data in the H1 timeframe. I've also used the close ath the end of the session feature (nice feature) and this way I try to force my strategies to open and close their operations fast (hoping the strats to have more operations). I'm asking my H4 strategies at least 300 hundred trades (and SQN=2) to pass my selection criteria, and I usually get nice strategies. If I ask more than that it seems the strategies are harder to get and they are not as nice as before.

What settings / timeframes are you using with EA studio to build portfolios? What number of trades do you think it is statistically significant to move a strategy from demo to live?

Thanks all in advance
best regards

Hi, i make this with 100 trades on Demo...

but with M15 and H1, H4 it takes much more time for that. And it is not my Timeline, i feel me better, when the trade after any minutes / hour closed.

But i have after Demo in Live ~ 1 month good results and after this, the most goes bad.

I dont know why, i must test it more time, i hope it will make better over time.

3 (edited by rantampla 2017-10-14 13:00:33)

Re: how to build strategies with a high number of trades

Look the Picture
a very good Strategie for 3 months Demo and then 3 months live and much trades.....

1 week and all loss sad

The Picture is only the live Account

Big Thanks to Sleytus, with MT4 Tracker i can that see, very nice your Software smile


https://s1.postimg.org/4l0js97ewr/bad.png

Re: how to build strategies with a high number of trades

The number of trades alone does not give information about the significance of a trading system. The advantage of FSB is just to "skip" the pass of writing and "skip" the demo account test; i do not understand why losing months to test in demo. Do you really think that using the generator contributes to making you make money? Popov has created a great tool but without your constant quest you will not gain. The number of trades offers great help in statistical terms, if we look at the "distribution", to understand where we are. An investment fund that earns 5 consecutive years may look very good, but if we look at it with all of the funds in the world, we realize it might be just lucky (that's what happens by creating strategies as we do). If it was so simple to earn hedge funds like Two Sigma, Reinassence and Bridgwater would be formed by 3 people, instead they work over 100 ph.d.

Good trading

Re: how to build strategies with a high number of trades

Hi rantampla,
thanks a lot for answering. Yes, it is sad to see a good strategy going bad. What are you planning to do with that? Are you going to give it more time to recover again or remove it from your live account?. Do you have an expiration criteria or a method to remove strategies from live trading?
Thanks in advance.

Your live account seems to have very good strategies, with many trades. I'm not having so many trades because I'm using H4 timeframe. I send you a capture of some of mine. One of my bests so far has 47 trades. Planning to move it to a live account soon.

Best regards

https://s1.postimg.org/8qxeju4ea3/H4_mt4tracker.png

Re: how to build strategies with a high number of trades

D5QM54S7 wrote:

The number of trades alone does not give information about the significance of a trading system. The advantage of FSB is just to "skip" the pass of writing and "skip" the demo account test; i do not understand why losing months to test in demo. Do you really think that using the generator contributes to making you make money? Popov has created a great tool but without your constant quest you will not gain. The number of trades offers great help in statistical terms, if we look at the "distribution", to understand where we are. An investment fund that earns 5 consecutive years may look very good, but if we look at it with all of the funds in the world, we realize it might be just lucky (that's what happens by creating strategies as we do). If it was so simple to earn hedge funds like Two Sigma, Reinassence and Bridgwater would be formed by 3 people, instead they work over 100 ph.d.

Good trading

No i dont can say that that is right.

I have long time test it, 75 percent with very good results in FSB / EAStudio....after create and put in live, the most make loss trades...

Thanks Sleytus i have use MT4-Tracker and wait 3 months / min 100 trades and good results and then i put on live.

When i also use it without Demo, then i loss with 100 Strategies ~ 500-1000 Dollar in fast time....

With the Demo / MT4 Tracker i can this big loss make smaller....but after few months now, it is not very good, much strategies loss after few months in live....

Re: how to build strategies with a high number of trades

rjectweb wrote:

Hi rantampla,
thanks a lot for answering. Yes, it is sad to see a good strategy going bad. What are you planning to do with that? Are you going to give it more time to recover again or remove it from your live account?. Do you have an expiration criteria or a method to remove strategies from live trading?
Thanks in advance.

Your live account seems to have very good strategies, with many trades. I'm not having so many trades because I'm using H4 timeframe. I send you a capture of some of mine. One of my bests so far has 47 trades. Planning to move it to a live account soon.

Best regards

https://s1.postimg.org/8qxeju4ea3/H4_mt4tracker.png

I let it run, but when it generate 50 dollar loss with 0.01 then i kill it...

i have now a new projekt....5 weeks run the reactor with m15 and hard settings. today, after 5 weeks i have 100 Strategies finish. I hope it makes better results, but i am not sure, i put it on Demo next week and i hope it makes good.

With H4 i have 2 months back a portfolio put on demo, but the same, ~ 30 trades, not sure for live i think.

M15 and H1 i have the best results.

Re: how to build strategies with a high number of trades

Hi D5QM54S7,

thanks a lot for answering and contributing to this topic.

I'm not sure if I understand your point of view well. It would be great if you could elaborate a little more and tell us about your experience. As far as I understand your post, you are not using a demo account so you must be selecting very well the strategies you are trading in your live account. Could you please share with us what criteria you use to pick the good strategies from the bad ones. Thanks a lot in advance. I'm hear to learn, as you said, if I knew it all already I would be rich by now smile

In my humble opinion (I might be wrong) I think that using the portfolio EA feature with hundreds of strategies in a demo account might work for me to find some strats that are not (much) curve fitted and that are more likely to work in the future. I'm using H4 timeframe and building strategies that have worked for 8-9 years accross very different market scenarios. If they have survived for 9 years, I think they can be traded in demo for a few months before I move them to a live account.

What do you think about this? What is your expereience trading without a demo account? Could you share something about your method? Thanks a lot in advance and thanks again for contributing to the topic.

Best regards

Re: how to build strategies with a high number of trades

I let it run, but when it generate 50 dollar loss with 0.01 then i kill it...

i have now a new projekt....5 weeks run the reactor with m15 and hard settings. today, after 5 weeks i have 100 Strategies finish. I hope it makes better results, but i am not sure, i put it on Demo next week and i hope it makes good.

With H4 i have 2 months back a portfolio put on demo, but the same, ~ 30 trades, not sure for live i think.

M15 and H1 i have the best results.

Hi rantampla, thanks a lot for answering.
I think your approach is very similar to mine and I have big expectations on this combination EA Studio Portfolio + Mt4 tracker (thanks a lot mr Popov and Sleytus). Yes, I have an account of M15 and H1 strategies and they look good. I think they will expire sooner but so ffar they look good.

Thanks again rantambla. Please keep us informed about your new project and your new account.
Warmest regards

Re: how to build strategies with a high number of trades

rjectweb wrote:

I let it run, but when it generate 50 dollar loss with 0.01 then i kill it...

i have now a new projekt....5 weeks run the reactor with m15 and hard settings. today, after 5 weeks i have 100 Strategies finish. I hope it makes better results, but i am not sure, i put it on Demo next week and i hope it makes good.

With H4 i have 2 months back a portfolio put on demo, but the same, ~ 30 trades, not sure for live i think.

M15 and H1 i have the best results.

Hi rantampla, thanks a lot for answering.
I think your approach is very similar to mine and I have big expectations on this combination EA Studio Portfolio + Mt4 tracker (thanks a lot mr Popov and Sleytus). Yes, I have an account of M15 and H1 strategies and they look good. I think they will expire sooner but so ffar they look good.

Thanks again rantambla. Please keep us informed about your new project and your new account.
Warmest regards

Yes, let us write, what we see smile

Re: how to build strategies with a high number of trades

Hello -- I'd like to add my 2 cents...

rantampla wrote:

...1 week and all loss...

From the chart it looks like there was one bad day when the account lost a lot.  I've also seen that recently -- where my strategies are making consistent, small profits and, then, one day all my profits disappear.  When I then go back and look at the MT4 EURUSD chart then I can see why -- usually there is a sudden, big change in the market.  Probably caused by some world event, I don't know. 

If you were to remove that one bad trade around #250 then your portfolio seems to be performing in the neutral -- i.e. not gaining and not losing.  That is, one strategy lost a lot and it made the overall portfolio look bad.

I think "curve fitting" has been given a bad name -- actually, I think that "optimizing"  or "training" is a better description.  If a strategy begins to lose it means it doesn't know what to do with the new incoming data.  And the only way I know how to improve this is by optimizing with different types of data sets -- so the strategy has experience with more types of patterns.  I don't think using 8 years of old data necessarily helps because those past patterns may never appear again in the future.


rjectweb wrote:

... One of my bests so far has 47 trades. Planning to move it to a live account soon...

It looks like you have a number of strategies with good potential.  I would pay attention to the 'Return / DD' ratio.

Re: how to build strategies with a high number of trades

sleytus wrote:

Hello -- I'd like to add my 2 cents...

rantampla wrote:

...1 week and all loss...

From the chart it looks like there was one bad day when the account lost a lot.  I've also seen that recently -- where my strategies are making consistent, small profits and, then, one day all my profits disappear.  When I then go back and look at the MT4 EURUSD chart then I can see why -- usually there is a sudden, big change in the market.  Probably caused by some world event, I don't know. 

If you were to remove that one bad trade around #250 then your portfolio seems to be performing in the neutral -- i.e. not gaining and not losing.  That is, one strategy lost a lot and it made the overall portfolio look bad.

I think "curve fitting" has been given a bad name -- actually, I think that "optimizing"  or "training" is a better description.  If a strategy begins to lose it means it doesn't know what to do with the new incoming data.  And the only way I know how to improve this is by optimizing with different types of data sets -- so the strategy has experience with more types of patterns.  I don't think using 8 years of old data necessarily helps because those past patterns may never appear again in the future.


rjectweb wrote:

... One of my bests so far has 47 trades. Planning to move it to a live account soon...

It looks like you have a number of strategies with good potential.  I would pay attention to the 'Return / DD' ratio.

Thanks Sleytus, you are right,

but it was normal news situation, it was not a big change, i think it was a normal change...

Ok, maybee small big change smile

I let it run and see, i write here, next time, if good or not smile

13 (edited by D5QM54S7 2017-10-15 17:51:19)

Re: how to build strategies with a high number of trades

Rantampla demonstrate why the use of the mt4 tracker would be better than the FSB wallet function :-) if you do not lie to yourself (low spreads, overfit the system, apply a strategy only to a special currency pair etc.) the in-sample and out-sample functions are perfect and perfectly replace your 3 months demo. The demo test may serve to verify broker execution or VPS validity. Still, why do 100 trades? a distribution curve will have little confidence either with 20,50,100 trades => a high confidence you could start having with 1000,2000 trades analyzed. For me the number of trades is not relevant precisely for this reason (I trade daily, 4h and 1h timeframe). MT4 traker for me is a useful tool to track a yearly account (or every 6 months) a real account on which EAs were launched earlier.

Rjectweb I'm not rich, I'm young, I'm not trading alive (I've almost finished studying) but I burned lots of money in trading (I think is the fastest way to learn) and others have earned it. Each of us has different strategies, what works with others probably does not work with us, so in the previous comment I told you that research is all. We know that what has not worked in the past will not work in the future, but what has worked in the past may work in the future => here it has to work! :-). Fortunately, the market, unlike coins, has a memory and what needs to be done is to exploit this. We know that the best in the world are quantitative traders (statistics, maths, computers) and use all very complex models, but we do not handle so much money so we claim less. Many professors have theorized the possibility of killing an EA when it does not work and inserting another EA that works => bad news, this does not seem to work (the best results were obtained while remaining loyal to the plan even if something does not work). Finally, we are sure that killing and continuously inserting EA into our portfolio is not overfit? To sum up, be honest with yourself, do not just look for profits, optimize it well but do not exaggerate (if the foundations of a home are not strong then everything collapses), trading is a job that does not study and does not search has already failed departure. If you have questions ask when I can answer :-)

Sorry both for my English

Re: how to build strategies with a high number of trades

D5QM54S7 wrote:

Rantampla demonstrate why the use of the mt4 tracker would be better than the FSB wallet function :-) if you do not lie to yourself (low spreads, overfit the system, apply a strategy only to a special currency pair etc.) the in-sample and out-sample functions are perfect and perfectly replace your 3 months demo. The demo test may serve to verify broker execution or VPS validity. Still, why do 100 trades? a distribution curve will have little confidence either with 20,50,100 trades => a high confidence you could start having with 1000,2000 trades analyzed. For me the number of trades is not relevant precisely for this reason (I trade daily, 4h and 1h timeframe). MT4 traker for me is a useful tool to track a yearly account (or every 6 months) a real account on which EAs were launched earlier.

Rjectweb I'm not rich, I'm young, I'm not trading alive (I've almost finished studying) but I burned lots of money in trading (I think is the fastest way to learn) and others have earned it. Each of us has different strategies, what works with others probably does not work with us, so in the previous comment I told you that research is all. We know that what has not worked in the past will not work in the future, but what has worked in the past may work in the future => here it has to work! :-). Fortunately, the market, unlike coins, has a memory and what needs to be done is to exploit this. We know that the best in the world are quantitative traders (statistics, maths, computers) and use all very complex models, but we do not handle so much money so we claim less. Many professors have theorized the possibility of killing an EA when it does not work and inserting another EA that works => bad news, this does not seem to work (the best results were obtained while remaining loyal to the plan even if something does not work). Finally, we are sure that killing and continuously inserting EA into our portfolio is not overfit? To sum up, be honest with yourself, do not just look for profits, optimize it well but do not exaggerate (if the foundations of a home are not strong then everything collapses), trading is a job that does not study and does not search has already failed departure. If you have questions ask when I can answer :-)

Sorry both for my English

Hello.

I use FSB and EAStudio now much months, i use all setting, in sample, put of sample, monte carlo, multi market......

I search 7.200 and more Minutes for good Strategies.....

I use my own Broker Data...

But every time, the most Strategies make a loss or make after small time a loss....

MT4-Tracker is the only Tool, which i can all Strategies good analyse ....

And when you think, it is a stagnation Phase, let it run, it comes back.....I have all Strategies, which in live bad or not used, run many months now in Demo, and i can see, it was not stagnation, it was most loss trades....

When you have other Experience, good, i wish you the best with it, maybe you can give a hint, what you make in settings, when you use EAStudio or FSB.

smile

15 (edited by sleytus 2017-10-15 20:03:06)

Re: how to build strategies with a high number of trades

I think many of us have had (and continue to have) experiences like rantampla.  That is, we can create strategies that look good during generation, optimization and testing -- but then perform poorly in a live account.

I think it is fair to consider a strategy similar to a math formula -- e.g. do you remember from Algebra 101 the formula for a straight line:  y = mx + b?  When you optimize a strategy using back testing data then you are solving for the best 'm' and 'b'.

When you later use your new formula in a live account and it performs poorly then that means your formula is not using the best 'm' and 'b' for the new, live data set.  And why is that?  Its because the data set you used to compute the best 'm' and 'b' does not match well or include patterns that your equation later encounters in a live account.

Popov's software does an amazing, amazing job of creating high quality strategies and portfolios and providing a framework for testing them very rapidly.  But the software can't predict future data sets.  I think that is where we need to be more creative.

One last point -- learned from a more experienced trader on this forum -- go for longer time frames.  The data seems less prone to noise and false signals.  Actually -- I said that wrong.  With data sets of longer time frames I think strategies are less prone to be "tricked" into announcing a wrong signal.

Re: how to build strategies with a high number of trades

sleytus wrote:

I think many of us have had (and continue to have) experiences like rantampla.  That is, we can create strategies that look good during generation, optimization and testing -- but then perform poorly in a live account.

I think it is fair to consider a strategy similar to a math formula -- e.g. do you remember from Algebra 101 the formula for a straight line:  y = mx + b?  When you optimize a strategy using back testing data then you are solving for the best 'm' and 'b'.

When you later use your new formula in a live account and it performs poorly then that means your formula is not using the best 'm' and 'b' for the new, live data set.  And why is that?  Its because the data set you used to compute the best 'm' and 'b' does not match well or include patterns that your equation later encounters in a live account.

Popov's software does an amazing, amazing job of creating high quality strategies and portfolios and providing a framework for testing them very rapidly.  But the software can't predict future data sets.  I think that is where we need to be more creative.

One last point -- learned from a more experienced trader on this forum -- go for longer time frames.  The data seems less prone to noise and false signals.  Actually -- I said that wrong.  With data sets of longer time frames I think strategies are less prone to be "tricked" into announcing a wrong signal.

Yes, ok, Longer Time Frame i have not used till now, but i will it use smile

Re: how to build strategies with a high number of trades

My input to this discussion:
Find attached the chart of my real-EA-Testaccount. I am running currently 10 EA's (which I had selected out of many on my Demo-Account) and had good results in the beginning but now it turned worse.
One EA which turned to making more losses I have stoped, other EA which had good performance I have increased the "weight"
The only reason why my performance deceased I assume is that the presently strong long time "Long-trend" is an exeption and was not in the data I used for creating.
But I accept that we cannot really determin the exact reason, because if that was possible the real professionals who have much more manpower and recources for development than me would all be very rich.
I see myselfe only as a small ligth in this brutal game, but makes fun and keeps me curious.
So I just try to get a number of EA's running which in the average make more plus than minus.
Is'nt it strange (but for me understandable) that in this forum too we all seem to do the same, we might all do the same tests with the same EA's created from the same generator, but we have not one thread where we can find strategies that are proven successful.
Best regards

Post's attachments

FK_171028.jpg 73.23 kb, 1 downloads since 2017-11-03 

You don't have the permssions to download the attachments of this post.