Tagged: algo trading plan
- October 26, 2019 at 17:48 #24960
Lately, I have been running the Reactor constantly, using Walk Forward optimization, Walk Forward validation, and Monte Carlo, and have been getting between 3 and 10 strategies into the collection each time for several different forex pairs on the M5 and M15 timeframes.
I had stopped demo trading for a while, as the strategies I was generating were failing badly, running a £1000 demo account down to £750.
With the new settings mentioned above, I have steadily built up approximately 50 EAs on the same demo account, and the balance is up to £ 870 after two weeks.
However, being that I am keeping notes of what the parameters of each EA are in a spreadsheet, I am noticing that I now have a lot of strategies that are very similar (Donchian Channel is a very, very popular indicator!), and most of the strategies have no TP, and even fewer have SL.
I read through all eleven pages of this post last night, and in an effort to implement better risk diversification, and at the same time, simplify and strengthen my strategy management method I have come up with the following plan.
It will be a lot of work, so before I embark on it, I would like to know if anyone here with some experience could add to, or advise any changes to it?
I have included some of your ideas here Jacpin, so would be very appreciative to hear your thoughts?!
EA Studio Algo Trading Plan
1) Forex pairs: EURUSD EURGBP USDJPY GBPJPY EURJPY GBPUSD AUDCAD AUDNZD NZDCAD
a) Have an even mix of each currency throughout portfolio
b) No currency appears more than three times
c) low spreads
2) Timeframes: M5, M15 and M30
3) Always 0.01 lot size, to keep it simple. Profits will come as the EA collection grows.
4) For risk diversification, create several types of Reactor settings for each timeframe. For instance:
a) one with no SL or TP
b) one with trailing SL and no TP
c) one with trailing SL and TP
d) one with Always SL and TP
5) Common Acceptance Criteria must always have:
a) minimum profit factor: 1.1
b) minimum count of trades: 300
c) maximum equity drawdown: 15 %
d) minimum 70% win/loss ratio
e) In Sample: minimum profit factor: 1.1
f) Out Of Sample: minimum profit factor: 1.1
6) Strategies that pass each set of Reactor settings to be tested in a separate MT5 demo account until FX Blue analysis decides if strategy is good enough to trade live
7) Using FX Blue, strategies that win minimum of 7 out of 10 trades will be moved to live after a minimum of two weeks testing
8) Using FX Blue, strategies that win less than 7 out of 10 trades will be deleted
9) Live strategies that win less than 7 out of 10 trades will be sent back to demo for monitoringOctober 26, 2019 at 20:56 #24970
I am impressed with the work you are doing, and I like that you have your algo trading plan. This is how things go in the right direction.
I like your acceptance criteria… I would look for a higher number of trades.
Regarding the strategies with our without SL and TP, give it a try, but analyze which one works best for you. If you see that having no SL brings you more significant losses, do not keep it as a method to diversify more the risk. You have many other things to diversify the risk with.
Also, the strategies that do not meet your rule in the Demo account for two weeks might meed them a bit later. Do not hurry to delete them. Think of how you can improve them and what made them lose. Look closer to the statistics there.October 27, 2019 at 0:00 #24984
I’ve been meaning to ask, what minimum count of trades would you set for each timeframe?
I remember in another post you gave a guideline as to how many bars we should run the Reactor with for each timeframe, I suppose the minimum number of trades would be different for each TF too?October 27, 2019 at 0:10 #24985
And when you say there are many other things that we can diversify the risk with, can you give some examples?
Maybe another timeframe?
Different types of strategies for different market conditions? (Not sure how to go about creating these, or when to implement them – maybe they just run constantly on a hedging basis?)
SimonOctober 27, 2019 at 16:42 #25133
Glad to hear back from you.
How normally I proceed with the minimum count of trades:
1. When I generate EAs for the first time for asset and time frame, I start from higher to lower number. For example, I place 1000, and if I see there are no generated strategies, I lower it to 900, and I keep doing this until I see the first strategies showing.
2. To the assets and the time frames that I generate new strategies every month, I add another 20 because there is new Historical data added.
Regarding the risk-diversification you can so many things:
- trading different assets
- different time frames
- different acceptance when you run the reactor
- different robustness tests
- portfolio Experts and regular Experts
- different methods to filter the strategies in the Demo account
- different approach to move the EAs from demo to live and the other way around
and these are just the ones that come to mind at the moment.
Every different approach that you take in the complete process is risk-diversification.October 28, 2019 at 6:30 #25163
Thanks for your reply Petko.
Regarding the second part of your process, would it be right to say that we are looking for around 20 trades per month for an EA?
And are you validating existing EAs on the all of the original data plus the recent month? So the data horizon is always growing? Would there be a point where there are too many bars for a particular timeframe?
SimonOctober 29, 2019 at 7:06 #25398
the number of trades depends really on the EA itself. Usually, the more entry conditions you have, the fewer trades you will see.
I am not sure if I get your next questions, but when I track the performance of the EAs, I look at the last 2 weeks and the last 1 month. So I want my EAs to be still profitable in the shorter and in the longer term.
I would not use more than 300k bars for M1 and more than 150k on M15, and more than 50k on H1.
Kind regards,November 29, 2019 at 14:49 #28486ThapeloParticipant
recently I notice that with EA Studio I succeed to generate strategies with many counts of trades. I reach over 1000 which is just great, but I was wondering why before I Was not able to do it. Is it some update that I missed in EA Studio?
One of the reasons I come up with is that I have more Historical data now…December 15, 2019 at 22:37 #30891RomanParticipant
@Thapelo – It very well could be. I find anywhere between 300 up to 2400 trades depending on the pair. I’ve found that the higher number of trade strats don’t last as long though.
@Simon – How you develop your strategies depends on a few things, but keep one very important thing in mind “Risk Management”. Your parameters may be too strict, and can possibly filter them in the collection after you’ve got many to chose from. If you’ve only got 10 you’re searches are too tight, causing you to find nothing but over-optimized strats on those timeframes. Those timeframes are also very noisy and I’d always have a sl and tp. If you’re algo takes the wrong position its good to just tap out rather than hold through long drawdowns with no SL, you end up losing money in Swap if it holds for too long. Also if it has no TP you could have HUGE profits that end up retracing back to your entry price and never closes or closes in a loss. My experience so far, always have TPs and SLs. Losen your search parameters and then filter them in the collection to what you like to see in your strats. Keep your overall risk in mind when spreading it out through your assets, need to keep “Armageddon” moments in mind.December 21, 2019 at 23:03 #31795
Very well said, Roman.
I also put SL and TP in all of the EAs I use and attach to the courses. Sometimes you might find these great strategies without SL because it is easier for the generator to find parameters for the indicators that will work without the SL, and the TP will be always hit. However, these strategies are more likely to end up on opened trades with considerable losses in opened positions.
And yes, it is better to have more strategies in the collection and filter them from there. Anyway, if you go too strict and have nothing in the collection, it is just nothing at the end of the day.February 7, 2020 at 9:55 #36906MeryjonestexasParticipant
Thanks for explaining that. I wasn’t sure why you always put SL and now I saw that I have 3 open trades in my portfolio that are not closing now for 2 weeks. I figured it out that they did not have a SL and might stay open for months.March 12, 2020 at 21:21 #41432HaliffaParticipant
I started trading with Expert Advisors created by FSB Pro where I select the option to add to the position at least 3 times.
Does anyone have experience in combining EAs in one account that add to the position and such that do not add?
I wonder how the balance will go if my EAs that add open more trades? This way the risk exposure will be more on those that add to the position.
Please, correct me if I am wrong.April 26, 2021 at 22:39 #84320
Bringing to the front that interesting topic…I like the idea of adding on the next same directional signal, but it should be with a limit – I usually use 3-6 times maximum.
As well, I prefer to add if the price goes against me, so I improve my average price.
You must be logged in to reply to this topic. Login here