Number of Strategies Generated in 10 hours?

[Tanya Jay]

Hello algo traders!

I signed up for EA studio and gave it a try for the first time last night.

I tried to create the strategies last night for 10 hours. It generated only 4063 strategies in 10 hours. Isn’t it too little? 

Below are my testing criteria:

– GBPJPY; H1 – 9221 bars
– From 2021-01-05 07:00 to 2022-06-27 20:00
** I only have this length of data provided from IC markets as I just downloaded it yesterday **

Strategy properties
– I keep it as is, with no change. (Long and short, TP/SL 10-100)

Generator settings
– Out of sample: in sample
– Working minutes: 600
– Everything else stays the same

Acceptance Criteria
– 300 trades
– PF 1.2

Can someone tell me what I’m missing?

TIA 🙂

[Samuel Jackson]

Hi Tia,

There are a few issues with your settings (I’ll expand on this later as on the train on way to work now). But your low generation number isn’t to do with your settings.

I believe that you have made your window inactive or shut your screen off on laptop maybe? EA studio needs to run on an active browser.

Can I ask a few questions to help you get the most generated EAs for your setup.

1- What browser are you using? (Chrome is by far the best)

2-Are you running on a desktop or laptop

3-What computer processor does your computer have?

We can tackle the low generation number first and then once sorted will discuss some of the other parts of your settings 😉

[Tanya Jay]

Hi Samuel,

Thanks for your response.

To answer your question,

1- What browser are you using? (Chrome is by far the best)
– Chrome

2-Are you running on a desktop or laptop
– I’m running on the VPS using chrome. Same VPS I use to run EAs on demo accounts. I just opened and ran them on 4 different chrome windows (not tabs). I’m traveling with my laptop all the time so I need to run them on VPS.

3-What computer processor does your computer have?
– Interserver VPS. 2 Cores. 60GB SSD. 4TB transfer.

What did I do wrong here?

[Samuel Jackson]

Hi Tia,

That all seems fine, 4 windows sounds about right for your vps. Although make sure you don’t push your vps too hard as multiple terminals as wel as 4 browsers could be a bit much on a 2 core vps.

Have you checked your processor utilization?

You didn’t minimize the chrome windows did you? You should make sure that you resize them and that all 4 are active on the screen.

As a quick test set then running on the vps, take note of the number of strategies generated and then log out of your vps (but make sure the windows are active). Then log back in after 15 minutes and make sure the strategies generated number has increased significantly (depends on settings but approx +10,000 say)

Then you will know that all is well. If that still doesn’t work then let me know.

It could also be that you are running too many browsers for your vps which can cause a drastic slowdown of Ea studio but if you confirm the results from the above test first it would be good

[Tanya Jay]

Hi,

Thanks for the thoughts. I upgraded my VPS to 6 cores (12 GB) and will test them tonight.

Can you recommend good settings as you mentioned earlier that there is something wrong with my settings above?

[Samuel Jackson]

Hey Tanya,

2 main issues is the amount of data is not enough, especially to generate 300 trades on h1.

also some OOS would be a good idea and also Monte Carlo for robustness.

what courses have you been through?

also what symbols are you planning to trade etc?

Did you resolve the issue of low generations first with a single EA browser as it seemed to be a window becoming inactive issue?

Your 6 core vps should be able to handle about 4-6 browsers comfortably depending on how many mt4 terminals you are also running but make sure to check your processor utilization also.

[Tanya Jay]

Hi Samuel,

Did you resolve the issue of low generations first with a single EA browser as it seemed to be a window becoming inactive issue?
– Yes, I did. Upgrading RAM helps.

What courses have you been through? also what symbols are you planning to trade etc?
– I bought 8-10 different courses on Udemy (Top5, 100 EAs on EURUSD, Crypto monthly, US30, etc). I bought them and tested them out on my demo right now. I mainly bought different ones because I want to diversify my portfolio and learn as much as I can. However, 90% of all courses are repetitive, except for this EA uploaded parts.

2 main issues is the amount of data is not enough, especially to generate 300 trades on h1.
– I see. That’s the only available data I got from downloading from my broker (IC Markets). What would you recommend me to do in this scenario? Also, what’s the ideal OOS %? 20% or 30%?

[Samuel Jackson]

Hey Tanya,

I would recommend using the premium data (or historical data tool – both are the same really) to generate your strategies using more bars and more trades.

Then once you have a collection of EAs that have say 300 trades over 5 years on this data. I would download the collection and run it through the validator with your IC data selected to check that it performs well over that data also. Obviously when running through the validator you would relax the number of trades made. So for example if you had min trades set to 300 over 5 years using premium data then if only have a years data with your broker you would reduce the number of trades to 300/5 = 60.

I’d recommend 20% OOS personally but only use OOS for the strategy generation and not when running them through the validator for your broker data. Likewise with your monte carlo. Validator on broker data should purely be just checking that the statistics are correct.

 

 

[Tanya Jay]

Understood. I will try your recommendation.

I have 2 more questions:

1. Based on your experiences, do you recommend running walk-forward optimization AND Monte Carlo validation? Does it necessarily mean it will generate better results/strategies, compared to 20% OOS AND Monte Carlo validation?

2. For M15, how far back of the data horizon should I choose to create strategies? 6 months? 1 year? 3 years?

Thanks!
Tanya

[Samuel Jackson]

Hi Tanya,

I would stick with just using the Monte Carlo. If you want to make it more stringent you can try increasing the number of tests and passing percentage or make the validation acceptance criteria better. I’ve found 20 and 80 works well though.

The walk forward is a more advanced tool(but that doesn’t mean better!). I have used it in the past but I no longer do.

For robust strategies I would recommend the combination of a good amount of data with plenty trades, OOS and Monte Carlo.

I only trade M15, H1 and H4 and I typically like the use 5 years data for M15 and H1 and about 10 years for H4. Can use more but I wouldn’t use less.

[Tanya Jay]

Thanks for the recommendations.

I have no intention of trading anything less than M15 as well.

I will use 300 trades, 20% OOS, and 20/80 Monte Carlo as a start then.

[Tanya Jay]

Another question I have is, what would be the best settings for TP and SL?

May use?
Fixed or trailing?
10 – 100? 10 – 300?

[Samuel Jackson]

Good plan Tanya,

I personally keep things simple and always use a take profit and stop loss. I just use a fixed stop loss also.

For M15 I think a good range is 20-120, but I would increase the upper range a bit for higher timeframes (say 150 for H1 and 200 for H4).

If you are getting plenty strategies coming through using those ranges over approximately 5 years of data then all good 🙂

If I run my reactor and get about 50 EAs overnight on a pair and timeframe then I am pretty happy. If a pair has too much more than this then I’ll make a note that can change the settings a bit for that pair and if too few then likewise I will need to relax the settings.

[Tanya Jay]

Thanks for the tips!

– May I ask the reasons why you use fixed SL instead of keeping it open? (can be fixed or trailing or don’t use it at all)

– When running Reactor, how many strategies should I aim for per 1 pair per TF to build a solid pool? What’s your ideal number?

– After getting collections, do you recommend adding the strategies into a portfolio? For example, EURUSD M15 portfolio and EURUSD H1 portfolio. Or it’s better to run it as an individual EA?

If there are other tips about using EA studio expert and/or building strategies, pleaseeeee share them away! 🙂

[Samuel Jackson]

Hey Tanya,

I think a stop loss should always be used so it’s really only trailing or fixed to choose between.

It’s kind of just a habit that I choose fixed to be honest so definitely feel free to try trailing also and see what gives you the best strategies. You could set two identical reactors running with only difference being one uses fixed and one uses trailing and compare that way to see what’s generally giving you the best looking strategies. Or just set fixed or trailing in the generator so it can choose between the two, both options are all good.

For the portfolios I would definitely recommend working with portfolios rather than individual EAs but I would also recommend not having overly large portfolios unless they are very uncorrelated (which is not easy for same symbol and timeframe).

I would stick to portfolios of about 30 personally, just a preference to limit too many positions being opened at the same time but this really isn’t a big deal unless doing scalping that requires super quick executions. Opening too many positions at once can cause several seconds delay on execution which you will be able to see in MT4.

It’s fine having lots of portfolios on different timeframes and symbols (say 10 portfolios of 30 EAs) as it’s unlikely that for different symbols and timeframes trades will open at the exact same time. But again it’s really not big deal just what I do.

 

[Samuel Jackson]

Oh also for number of strategies, it’s much better to have more in the collection than less as you can use the performance filter to trim things.

It sucks when nothing goes in after 10hrs but I also don’t like it when too many are going in too fast. If I get 50-150 overnight then I am happy and I’ll trim those collections down to a portfolio of 10-30.

[Tanya Jay]

Hey Samuel,

Thanks for the insights. I will use 10 portfolios with 30 EAs in each portfolio as a ballpark and a starting pool then. And after that, I will pick about 12 winning EAs to run in live/mock demo.

[Samuel Jackson]

Hey Tanya,

Sounds like a good plan. I wouldn’t be too rigid about the 12 EAs. I’d say aim for 10-20 EAs and stick with 0.1 lots size on your 10k account and you will get a feel for how much your balance will change over the course of a month.

It of course depends on the EAs you select as one EA may typically trade more often than another for example.

More important of course is to be moving your balance in the right direction but you will want to also get a feel for the risk you are taking on your account every month.

For example, if you have a trading plan for a month then you should be able to feel pretty sure that even if you have a really bad trading month that your balance is very unlikely to drop below a certain threshold.

As an example of you had 10 EAs trading then for each one you could get the average number of trades a month, win loss ratio and typical size of losing and winning trade. Then you could assume that the win loss ratio was halved to simulate a bad month and calculate what the loss would be for the month and scale your lot size accordingly.

I’m not saying that this is a good way as I just came up with it on the spot as a quick example but it would give you something tangible to estimate as a start and you can improve your calculation as you gain experience.

In a nutshell the point I am trying to make is just that you should have a reason for the lot size you are trading that is as much about the risk you are willing to take as it is about the profit target you are aiming for. But always aim to keep things as simple as they can be 🙂

[Author]

Posts