[Samuel Jackson]

Hey Tanya,

Great to see you are moving forward, and glad you appreciate my help and your kind words ;-)

So my first point would be that it seems you switch the EAs every 2 weeks? This is fine but 2 weeks and 4 weeks together as selection is pretty similar. I would suggest as one of your filter tests you try just a simple filter of moving the EAs that have been profitable over the whole life on you demo account with over 5 trades (say 1.1 pf) and then as the life of your incubator account extends you could add an additional filter for the most recent month also.

Let’s say now you have a collection of 100 EAs in your pool, in three months you validate and only 50 of them have traded as profitable (or even just broken even) over the last 3 months. Move those 50 to a second tier pool for example that only has strategies that have performed suitably over three months on demo in real-time.

Over time you will build a pool of EAs that is higher quality, you could even repeat the exercise on your tier 2 pool after 6 months say and move all the ones passing that to a tier 3 pool.

Hope that makes sense? It’s just a suggestion of how you could build a higher quality pool of robots over time, you of course need a combination of good robots and sensible filter criteria to make profits.

While you are practicing and learning if you do this as a side then over a year you could make a really nice quality pool of strategies for yourself.

[Samuel Jackson]

Of course nothing to stop you trying another pair combination in tandem Tanya. If pairs are working better for you then try combining them as it is likely that your pool of EAs for these pairs are higher quality (which can be partly lucky sometimes as it is very possible to run the reactor with exact same settings and pick top 10 the exact same what and one with far more overoptimised poorer quality EAs than the other.

This is one of the main reasons it’s important to test them appropriately out of sample, starting with using demo accounts (and patience) and progressing to using EA studio without the demo as your experience grows.

[Samuel Jackson]

Hey Tanya and Petko,

I can add a few points:

Tanya uses IC markets for both demo and mock live account and I know that she validates the strategies using the IC markets broker data.

Although I didn’t select these pairs I advised Tanya to do what I do which is to select 4 pairs to trade together that are as uncorrelated as possible. They look good to me? Also the mix of timeframes M15, H1 and H4 also works well for minimizing correlation on the same symbol.

I’ve got two of those pairs in my own portfolio at the moment and it’s been trading really well :-)

Can you confirm the above is correct Tanya :-)

As to the poor results Tanya:

You have at least a month of trading on the demo and at least a few trades when deciding how to move EAs right?

How many EAs are you putting on your mock live at a time?

Also you could try a few different mock live accounts at the same time and trial a few different selection methods?

But to answer the original question Tanya (at last :-):

My advice would be to maintain a pool of collections and then check them every 3 months or so using the validator for the most recent three months of data, then remove the ones that don’t perform well over the last three months and keep doing this ongoingly.

Over time you will build a higher quality pool.

The higher quality your pool of strategies the less need to continually create more.

Also keep things simple when moving the EAs, just a good profit factor and more than 5 trades for example over the last month.

Keep the demo running as long as possible and keep replacing the strategies that are performing poorly. This is why I actually prefer to work with single EAs rather than portfolio EAs as it makes it cleaner to just remove an EA that’s performing poorly and then replace it with a new one.

Keep practicing as you are Tanya, things will improve with practice, it sounds simple enough but it does take a while to get the knack of things :-)

 

[Samuel Jackson]

Hi Alan,

I assume you mean download to another folder?

This is controlled by your browser. Just go into your chrome settings and change the default download folder in there.

 

[Samuel Jackson]

Regarding the slight out of balance in terms of USD is not a problem at all in my opinion with all those suggested pairs.

I think stick with what has been working, and experiment with small changes at a time rather than a large overhaul.

Yes 10 different strategies on EURUSD can be a balanced portfolio. I would prefer they were spread over different periods though. 2-3 for M15, M30,H1 and H4 would be better.

Yeah I’d think once you have enough on EURUSD move to different pairs. 1000 is more than enough in your pool for any single pair and timeframe though.

[Samuel Jackson]

Hi Tanya,

As you had sent me you collection I was able to check this quickly for you. First of all I changed the dates to 11/05/2017 and 11/05/2022 (since 05/11/2022 is a few months into the future so I assumed a typo?).

I have run it using IC Broker data and Premium data including Monte Carlo and there are 121 strategies passing for premium data validation and 176 for IC broker data.

This is totally fine, it is likely that many only marginally passed the acceptance criteria for IC and only marginally failed for IC data. Also plenty could have been overfit to the data used to generate (IC?) and therefor aren’t performing as well on different broker data which is a good robustness test and best to lose those strategies anyway.

You can also simply load the full 204 collection into EA studio and open up each strategy individually and see how the chart looks when you change between the data sources and remove the ones that differ too much, should give similar results. Probably about an hours work for 204 strategies but a good exercise.

All is well, these results seem very reasonable to me. The broker data wont be hugely different but it can be different enough to get these results.

[Samuel Jackson]

It’s not really a problem having so many strategies, it will help you find strategies faster as you are able to skip the generator having to find strategies from scratch. Just make sure to keep them tidy I guess.

What about other pairs and timeframes? You should focus on building these up if you are approaching too many for EURUSD M15.

Also once you have a big collection this will allow you to run the reactor less often and just use the validator going forward to ensure your pools keep passing the validator settings or also if you decide to change your strategy requirements by modifying the acceptance criteria you will have a large group of strategy’s to begin with.

I would pick the pairs and symbols that you want a pool for, build each up and then going forward use the validator to maintain and keep current and the Reactor to top up as needed going forward.

[Samuel Jackson]

Hi Alan, having too many strategies really isn’t much of a problem. But lets say the you are aiming for roughly 1000 strategies in your pool per symbol timeframe pair.

If at the end of the month you have 2000 and want to trim a bit you could load your existing collections into the validator in chunks to get the best 250-300 and then just keep the final 3-4 collections that come out of the validator. Also as you will be moving the validator end date forward a month some may just naturally fail the acceptance criteria and if you are getting too many passing you can simply increase the acceptance criteria a little and filter the top 300 however you want when you run them through the validator.

How many strategies are you expecting to have??

[Samuel Jackson]

Regarding Petkos 6 month and 1 month, I believe Petko has determined through experience that it works well and has some logical sense but I’m sure he will confirm his reasons when you reads this.

My feel for these date ranges is that the longer filter is serving as more of a general robustness test of the EA and then the shorter filter serving as a selection based on what is performing best on a shorter more recent timeframe. Wether 12 months and 2 months or 4 months and 2 weeks may perform better I don’t know.

Also It’s not to say that the Ea won’t perform well after a month but just that the process should be repeated on a reasonably frequent basis to adapt to changing market conditions.

I agree with the logic of the approach and certainly used to get good results on demo when I started out using it. I remember my first months results were 50% profit that steadily grew over the month (using 10k account and 0.1 lot size) but I personally would be at least halfing my lot size in order to reduce my risk.

I would practice it for a few months and see how you get on (using demo account of course). Fairly easy to run other tests alongside it as it only takes a couple of hours at most once a month.

[Samuel Jackson]

Also I think that for finding EAs that will work long term I would be putting say 100 or so EAs that you have generated onto a demo account and just leave them trading for a few months and then view the results, but a quicker way would be to run the reactor with an end date set the 3 months in the past and then use the validator to check how they performed for the 3 month walk forward.

It all depends on how proficient and confident that you can match EA studio with demo trading effectively.

In my opinion people new to EA studio should start out managing with demo for a while before transitioning to purely using Ea studio to manage a pool of EAs.

[Samuel Jackson]

No worries Steve, glad it makes sense and is helpful :-)

Thats fair enough in wanting to leave them on indefinitely however I believe there should be some ongoing tuning of the EAs wether that be to keep the same EAs and tune them ongoingly as Petko does in the top 12 course you are going through.Or switch out new robots monthly like is done in the professional course.

Or you can manage your robots by moving too performing EAs from a demo account to live or mock live account (which is just an alternative way of achieveing the same thing as is performed in the professional course without demo trading really)

i personally prefer just managing my robots by having lots on demo and then using FXBlue to monitor and select the EAs but it’s the same as professional course method so just a preference.

[Samuel Jackson]

Also it’s by design that Ea studio portfolio only does a single asset portfolio. It’s purpose is to create a portfolio expert advisor that can only be attached to a single chart so all the strategies in the portfolio must be for the same symbol and timeframe.

I initially thought it would be good to have a feature to do a combined asset portfolio but its not something I would use anymore even if it existed actually.

[Samuel Jackson]

Yes I agree with Petko on large stop loss and small take profit being able to work well.

Your logic is reasonable, however remember it is the average winning trade / average losing trade x the probability of winning that is important.

For example what would you prefer:

Strategy A: 1:1 stop loss or take profit ratio with a 60% win/loss ratio

Strategy B: 2:1 stop loss to take profit ratio but with a 80% win/loss ratio

Let’s use 100 dollars risk as an example;

Strategy A should make 10 trades and win 600 and lose 400 which will mean 200 profit with a profit ratio of 1.5

Strategy B should make 10 trades and win 400 and lose 200 which is also 200 profit but has a profit ratio of 2

That example is only expecting a much smaller relative increase in probability of winning against a halving of the tp/sl ratio also and still outperforms it.

The probability of the trades winning combined with the risk to reward ratio of each trade is what’s important.

[Samuel Jackson]

Hi Steve, yes you are right in EA studio you can only create a single asset portfolio. But can check the combined results manually easily enough.

Also I would say that expecting 7 months results to be good for a strategy only developed using 2 years of data on M5 is also pretty unrealistic.

2-3 months would be a better wf amount in my opinion.

Also remember you are not managing robots in anyway with that system but simply putting them on and leaving them for 7 months (which is ages).

[Samuel Jackson]

Also you seem to be very restrictive in the strategies you are targeting with EA studio, care to explain your logic further?

Often principles of discretionary trading don’t cross over properly to algo trading

[Author]

Posts