[Samuel Jackson]

Good Job Alan, this is exactly right.

You can see that for 1 reactor you are getting 6000 generations a minute but between 5-10 reactors your combined generations is 10000. So I would suggest that 5 is pretty optimal for you. Although 3 is getting 9000 generations per minute total which is also pretty good.

You must have good hardware, I find if I keep increasing the reactors then at some point I get a pretty drastic slowdown.

This is good info because it lets you save some resources for other things and also dont need to work the computer hard unnecessarily

[Samuel Jackson]

Hi Gareth/Alan,

Definitely running too many reactors can slow things down but it depends on your hardware. My rule of thumb is a browser for each computer core works well so I run 8 on my 8 core computers.

However in this case I believe the issue was not this but that the browser was inactive and simply stopped churning.

You can easily test the slow down effect by doing the following:

Set a single reactor running and then time how long it takes to generate 1000 strategies (generate not collect), then open a second reactor and set running and time again. Continue doing this until you see a large drop in generation speed and you will know the optimal number of reactors for your computer.

But it’s important is making sure the browser is active so it keeps running.

[Samuel Jackson]

Hi Alan,

Ill have to check what I get but this seems low number of generations to me so I expect you are working your computer very hard. Of course it depends on number of bars but I am pretty sure I get about 1000 generation every couple of minutes.

Maybe easiest to attach your settings file Gareth, there could be something with the account size or leverage preventing trades being taken. I see you are generating using 0.01 lots usually this is 0.1 so I am seeing a few flags of odd settings.

[Samuel Jackson]

Hi Omkar,

Great, yes in this case of using generator I would recommend using Ea studio over FSB pro. Ea studio has a better generator, fsbpro is better if you have your own strategy ideas that you want to create. There’s more too it than that but conceptually if you are using the generator I would recommend sticking with Ea studio for now.

Also make sure you backtest using the open prices model.

Please do the following.

Import your broker data to ea studio

For an ea see backtest results in ea studio

then download the ea and backtest on mt4 using open price model and compare the journal files

The software is robust and matches well, many users including myself have these kind of issues/concerns early on. Let’s see how you go with the above and we can take it from there :-)

 

[Samuel Jackson]

Regarding your struggles with matching fsbpro with mt4/5. I have several suggestions to help with that.

First though can you let me know why you are using fsbpro instead of Ea studio. I am assuming you are manually building strategies that cannot be created in EA studio??

The exercise of matching the backtests is most straight forward using Ea studio and mt4 initially I would say. But if you want to go straight to FSB pro and mt4/5 then all good too. Just needs a couple of additional considerations.

And info about your experience in fsbpro and ea studio will also allow me to help you better. Cheers

[Samuel Jackson]

Hey Omkar,

For point 1, the mt4 limit can be easily increased by changing the chart options. Your broker data will only download so much but you can import data to mt4 much greater than 32k bars.

additionally yes the limit in FSB pro of 200k bars can be easily increased by modifying one of the settings text files, if you can’t find how to do this on google let me know and I’ll help you out.

[Samuel Jackson]

Hi Garett, I agree that this message in itself isn’t a problem really. However with the hugely wide settings you have shown I would expect you to be getting some strategies.

I believe I see the issue though ;-)

you have shown the generated image with only 6000 generations in 54 minutes. This is barely anything so what I expect is happening is the browser is becoming inactive.

Watch this generated number for a minute or so to see how many generations you expect and then keep an eye on if it is moving up properly.

The browsers need to be active so if you are running several, minimize them so they are all visible and they should keep running.

Any further issues just post and will help out.

[Samuel Jackson]

Hi Omkar,

I have a couple of questions also.

1. Have you tried doing this comparison with Mt4 and fsbpro instead of mt5 and fsbpro?

2. Why are you using mt5 rather than mt4?

Also have you compared a control point backtest (OHLC M1) in mt5 with the live trade results?

I have personally found mt5 backtester to be a real pain and pretty buggy. Something does look odd here, but I would recommend doing some tests using mt4 first.

[Samuel Jackson]

Hi Robins,

I cant see any problem with the input file. It seems there could be an issue with your node.js installation though. I would try reinstalling this and if this doesn’t work I would get in tough with Popov himself. Cheers

[Samuel Jackson]

hi Robins,

right click on the “.cmd” file and paste the text here please. I expect you have only asked it to download these pairs and timeframes but I could be wrong.

in the text file you will be able to edit the list of timeframes and symbols that you want to download.

[Samuel Jackson]

Hi Jaylon,

I’d suggest for Petko to help you fully you provide three things:

1- the EA

2-the data imported

3-the EA studio settings file

Ill be able to take a quick look also if you provide this.

To me the most likely thing looks like this particular Strategy has been generated on different data and seems to perform extremely different on this data for some reason. Looks odd that it makes so few trades.

[Samuel Jackson]

Hi Robins,

First point is that EA studio is undoubtedly the best software for this approach given its speed of generation and validation.

I know a trader who does exactly this with EA studio.

I would suggest that to be even more efficient that you save a pool of collections and validate first also, depending on the look back period I would expect most of the strategies you generated the week before to pass again this week (unless you are using an extremely short look back)

If looking for 5 trades in three days then I would expect that m5 and m15 timeframes will work better for this. Usually I expect my h1 EAs to make 1-2 trades a week.

Additionally can you share your trading background so I can help further if you have more questions. I would strongly recommend at this point that your live account is either very small and trading the smallest lot size available or a mock live account until you are getting stable results.

[Samuel Jackson]

Hi Jaylon,

You correct in changing your EA studio settings to avoid this high spread. I remember having a similar issue a long time ago.

If you are no longer collecting strategies then this is to do with your settings.

I would suggest initially running just the generator with no validation and relax your acceptance criteria gradually until you are collecting say 20-50 strategies over 24hrs. Then Filter these in the collection tab and save the collection and run through validator with robustness settings.

This way you will get a feel for what settings are most appropriate.

If you post your generator settings and acceptance criteria I can review and advise better.

Cheers

[Samuel Jackson]

Hi MrPrice,

I see a few main issue with your comparison.

1 – You are using the inherent JFD data in EA studio to compare with your broker terminal data. You should first use the export data script to export your data from mt4 into EA studio for a fair comparison. Additionally this will automatically match all swap values and allow you to set commission. But note that MT4 doesnt include commision and so for a pure comparison its better make this zero and then change it afterwards.

2 – You are running the MT4 backtest using ticks instead of open prices, it is best to use open price modelling in the MT4 backtester.

As to how reliable is EA Studio, very. Experiences like this are always user error.

I would strongly recommend reading the thread below in which I helped someone with similar issues.

portfolio calculations vs reality

[Samuel Jackson]

Hi Maria,

Perfect, that was my next suggestion if the above didn’t work. You can use as many bars as you want that way :-)

Shame we cant lift the bar limitations of EA Studio the same way!

Cheers,

Sam

[Author]

Posts