I am currently building a workflow for successful forex strategies. Join me!

I’ve been using StrategyQuant for more than a decade, but believe it or not, I haven’t even used its full capabilities until now.

My process is to create thousands of strategies in the builder and then run them through a very selective robustness check. The (few) surviving strategies are then activated on an MT4 demo account, where they must execute at least 25 trades before I consider them for use on a live account.

So far, I am very happy with this process and would like to automate more of the production of successful systems. That’s why I’ve been diving into the custom project features that can be used to create custom workflows.

While I am building some sample workflows for successful forex strategies, I would be happy if anyone on this forum is interested in the same and willing to share their experiences.

Especially, I would be interested to know if anyone has ever started such a project on their own?

Hi Gerhard,

When I watched SQ’s video tutorials, I was very happy to see that you can automate this.
Since then, I only create/test strategies via automation. It saves a lot of time and effort! 🙂

You are welcome to message me in German if you like ;). Here in the forum english is better.

I have also spent quite a lot of time trying to get my workflows running well. I’m getting close, but I’m not there yet. I’ve built dozens of workflows and have gotten good at doing that. But, my workflows tends to get stricter over time, and then, one change later, I get no strategies being generated any more. So, I still need help, but I’m willing to share what I’ve done and the lessons I’ve learnt.

IMO, the biggest problem with workflows is the maintenance of them. How do you quickly change dates, symbols/instruments, etc. Ideally, you don’t want to run a single workflow and then sit idle for 10 hours while you sleep, waiting for you to configure the next symbol. You want your workflows to run 24/7 and produce results.

Contact me here on the forum via PM, or on Discord as @SkipZA

Hi Conmariin,

thanks for your reply. So from what you describe I think we are very likeminded and want to make the best out of StrategyQuants capabilities to automate the process of trading systems creation.

I appreciate you giving me your details to contact you privately (or in German), but I would rather keep this discussion open in the forum and invite others as well to join the thread.

I´ll share the workflow I have developed for EURUSD 1H systems in a short while, and we could use this approach to jointly discuss improvements.

Best regards

Gerhard

Hi FirestarZA,

thanks for your feedback. I completely resonate with you: one of the difficulties in workflows is that once you change one parameter, you may end up with no results at all. It took me quite some time to get a workflow up and running, which consistently produces systems on a daily basis. If they are good enough to withstand a live test still has to be proven (I´m running them on demo accounts already to test this).

Maintenance is also quite hard to do, I certainly agree. And because of this, I think it would be very beneficial to discuss different approaches to “program” those workflows to make them easy to maintain as well.

I´ll share the workflow I have developed for EURUSD 1H systems in a short while, and we could use this approach to jointly discuss improvements.

Best regards

Gerhard

https://algotrading.de/strategyquant

FirestarZA wrote:

IMO, the biggest problem with workflows is the maintenance of them. How do you quickly change dates, symbols/instruments, etc. Ideally, you don’t want to run a single workflow and then sit idle for 10 hours while you sleep, waiting for you to configure the next symbol. You want your workflows to run 24/7 and produce results.

Hi,

I don’t know if I understood you correctly, but you can produce multiple workflows under “Custom Projects”. I created one for one symbol and than copied it multiple times for other pairs and timeframes. So I can run and produce several pairs simultaneously.

Correct me if I understood wrong.

That is correct, yes. I also do the same. However, I try to have one workflow per type of strategy. So, one for trend following, one for mean reversal, one for scalping, etc. I then modify my symbols based on what currency pair I want to build for.

There are pros and cons to each approach. Your approach (which I’ve tried in the past) has a ton of workflows. While this by itself isn’t a huge problem, updating them becomes a huge pain. If you find a mistake/error in a workflow, you have to update 28 workflows (one for each of the 28 currency pairs) to fix it. And, since that’s a manual process, you can easily make another mistake and mess it up a second time. Same goes for updating dates to include later date ranges. You have to do that many times.

The approach I follow means I only have to use the “modify symbol” option when I want to change symbols. Also, updating dates only has to happen once.

This is what I mean with maintenance of the workflows.

Aaah okay. I understand. Your “Custom Priject-focus lies on dividing in strategies and mine focus in dividing in pairs and timeframes. Okay, with focus on strategies you have less workflows. Correct.

On which pairs do you run your workflows? I take only Eurusd, Gbpusd, Usdjpy, Eurjpy, Gbpjpy and Xauusd. I made with theses pairs faster and better results.

I currently have a live portfolio that runs strategies on the following currency pairs:

AUDCAD,AUDNZD,EURJPY,EURCHF,EURUSD,GBPJPY,USDCAD

I then have, in addition to the above, other strategies that are currently not in use, running on:

GBPUSD and XAUUSD

My end goal is to have as many uncorrelated strategies as I can get (aiming for 100 or close to), running on all the Major and Minor currency pairs, as well as a few of the metals (not all), and each of these also running on the following time frames (M15/M30/H1/D1).

So that’s 28 currency pairs, x 4 TF each = 112 strategies, at least, without the metals. And once I’m done with that, I’ll start working on other markets.

Ambitious goals, and I may not get there (it may be unrealistic), but this is what I’m working towards.

That´s great to hear that the three of us follow more or less the same idea, but with different approaches. I´m currently using workflows the way Conmariin does, meaning copying an existing workflow and changing it only for a new symbol. Building different strategies would be desireable, too (like FirestarZA does).

The reason why I didn´t build a lot of different workflows up until now is because I´m not sure what a good workflow is. A good workflow – according to my definition – would be one that produces similar results in live trading. So what I do is the following:

I have just one workflow running which runs for 24 hours. After that it has created approx. 1400 strategies, which are then handed over to heavy retesting. In the end between 1 to 5 strategies survive. The whole process starts over so that I get 1 to 5 strategies per day.

These strategies are then activated on a demo account of my broker, running 24/7 on a VPS. I use an “EA” which uploads completed trades automatically to a mySQL database and I use an automated script which creates stats on the live performance every 24 hours.

This way I can compare live performance and simulated performance. But I have only just started.

I would think of duplicating workflows only once I know what robustness tests give me enough confidence that the represent live performance to a certain degree.

I run my completed strategies on a live account immediately, but with 0.01 lots (ideally a cent account would be better, as then you can set risk and let it run, only moving from a cent account to a real account without changing settings). Once it runs on there for about 25 trades or two months, I export the results and look at the report in Quant Analyzer. I then build a portfolio every couple of new EAs I get, in QA. I run a few what-if tests, MC tests and so on, on the results, and make my decision about what my live portfolio will contain. This then gets run on higher risk (2% account balance per trade). So far, my results are ok, but I’m not yet making money (or losing really, just breaking even). This may be a result of not yet having enough strategies, or having a particularly bad trading time (or heaven forbid, a particularly good trading time and I’ll lose once I get out of it). Either way, building more strategies is the MO right now. I’m only adding one or two strategies per week at the moment, which is way too slow.

Have you ever thought of swapping systems based on their performance? I´m a firm believer that there are no systems that perform well over an extended period of time. That´s why I keep an eye on the ranking of my strategies and turn them off (or replace them by others) when their KPIs begin to drop.

I think this is one of the reasons why also StrategyQuant says: you´ve got to produce new systems all the time. It is not a one time effort and then you just relax and earn money.

For this reason I do not require my systems to be profitable over an extended period. I just wait until they have performed 25 trades on a demo account, then they are qualified to be used for live trading. But only if they rank in the top x range as listed on the table I have attached.

I do that, yes. I keep 2 lists of strategies. One list is my competitors list, and then I have a smaller list, which is my finalists. The competitors goes into the bucket when I analyse using QA, and I then only take the best of the list, and only if they’re uncorrelated (and pass some other basic tests. I imagine (though I’ve not really done this yet), that I’d do a WFM test at some point as well, to keep my strategies around a bit longer.

Could you tell us more about the step of your workflow ?

Hi Laurent,

I´m actually glad that you asked. Because both FirestarZA and Conmariin (who replied to my post up until now) seem to be very advanced already, having good performing workflows in operation.

As far as I am concerned: I´m still learning and eager to share, but also get input from other members on areas to improve or simply some tips and tricks to do thiongs better.

My primary goal is to develop a workflow for the EURUSD on the 1H timeframe, which shows a close match of backtesting results with real live performance on a demo account.

I´m quite happy with what I have accieved so far and I´m happy to share that with you:

My workflow consists of 7 tasks :

1. Builder: build strategies for EURUSD 1H within a time window ranging from 1/1/2017 until today with an Out of Sample period right in the middle of the range (11/2018 until 03/2020). Do this for two days. This results in approx. 2000 strategies that meet my initial criteria.

2. Retest the generated strategies on the 1 min timeframe and delete any strategy that doesn´t comply with the initial criteria.

3. Delete all strategies with a profit factor < 1.3

4. Test remaining strategies on another Out of Sample Window ranging from 1/1/2015 until 31/12/2016 (the two years before the initial period). Delete all strategies which have a profit factor < 1.1

5. Retest on Dukascopy data. I forgot to mention that all the steps above where using MT4 data from my broker (Admiral Markets). Now I want to see how resilient my strategies are if the data source is (slightly) different. I require the profit factor to still be at least 95% of what is was with broker data. This removes the dependency of strategies on particular brokers´ datafeeds.

6. Retest on other symbols. I test the strategies against GBPUSD, USDCHF and USDJPY. They should at least be slightly profitable (profit factor > 1).

7. Monte Carlo Retest. I do basically two tests: trades manipulation, where trades are skipped with a 10% probability and Monte Carlo retest, where I randomly change the parameters of the strategy. All these Monte Carlo tests are done 200 times and I require the Return/DD value to reduce only by 50% with 95% confidence level.

This results in 1 to 5 strategies that survived these stringent tests and I put them on a demo account to monitor their live performance.

This is where I stand at the moment. What I´m interested to learn is:

– which robustness tests are other people using?

– which robustness tests help to produce strategies that work in a live environment

– What about optimization of strategies that have successfully passed: is it ok / required to optimize them afterwards or should we leave them untouched?

I would be happy if someone could answer these questions or share their own workflow.

I´m running a startup business to let people rent successful strategies rather than buying them on https://algotrading.de. All of these strategies were built with StrategyQuant.

Hi AlgotradingDE,

I know there are people who use this type of method to determine when to include/exclude strategies on their live account and they report success using it, however I must confess that I’m not convinced it’s the best way. I’ve found that in a strategy’s performance profile a stagnation period of less than 180 is quite good, and it could be a very well performing strategy outside of that max. stagnation. Drawdowns are a part of life for a strategy that otherwise could be a very well performing strategy. However if, in the first 25 – 30 trades it coincides with a stagnation period or a drawdown period, it obviously wouldn’t obtain the best KPI’s on that period, and would be discarded.

I tend to prefer to evaluate whether the strategy’s performance is in accordance with it’s performance profile as backtested etc. in SQX, set limits with regard to DD based on a MC confidence level. If the strategy performs totally different to what is expected compared to SQX’s findings in anyway, it would be eliminated, but until then, for me it would be a valid strategy to be included (subject to correlation tests on the portfolio etc.).

Thanks for bringing this subject up, by the way, as I too am looking for ways to automate the whole generation – testing – evaluation – deployment process as much as possible as well!

Cheers!