Hi there,

I'd use something like this to count trades:

Hello Amine,


after running enough backtests I'm about to add all my entry signals and the trailer to a multi-entry strategy as I intended. But before running it in real time one more problem needs to be fixed. It just occured when I was adding entry long and entry short (I built and tested them separately): The strategy ignores my restriction, that there mustn't be more than just one open position at the same time. The strategy always opens two positions - one in short direction and one in long direction. What can be the reason of that happening? I didn't change the logic of the trailer - did I overlook some detail again?

Therefore I'm asking you once again for checking the strategy!


Thank you in advance for you help and for your patience!
Yours sincerely
Lunas

Hi

I see;
As far as I see the multiple positions are ,done almost instantly in both sides short and long. this is strange to me as the entry conditions for long should be different for the short side. I'll let you verify that... But beside this, I see that the strategy waits till the acknowledgment of the opened position is received before switching the "Go" variable to off. So there is 2 solutions at this level:
If you're sure about your entry conditions in both long and short sides, you can change the link that connects the assign block after 'Open @ market " block from "Fill OK" to "One flow" this will executed the assign block straight away.
If you find a wrong logic for your entry conditions, I think you can keep the assign as it is, but this is subject to further tests before confirming definitely.

Cheers

Hello Amine,


thanks for your quick response!

Admittetly: The entry conditions seem to be the same for both sides. But there are slight differences (I added some notes in the strategy). When checking the strategy once again, I found a well hidden mistake in the use of the BBands - I made the necessary correction (hoping, that this could be the reason for the trouble), but it didn't have any influence on the fact, that the strategy opens long and short at the same time. After that I changed the link of the "assign block" as you suggested. Now the strategy doesn't open long and short at the same time any more. But another problem occured: now the strategy trades "long only" - even after a short signal....strange, isn't it? You can test it for example on 2017-04-28, 10:39GMT and 11:26GMT.

Obviously something more needs to be changed....
Therefore I ask you again to check the entry conditions of the strategy - you surely have further ideas of what can be the reason.


Thank you in advance
Yours sincerely
Lunas

Hi Lunas !

The reason of not getting nay short trade is actually quite simple: As conditions are pretty much the same the first flow out of your"Multiple action" block is linked to the Long Entry, the second one goes to the short conditions. So as the flow is distributed in sequence( within these milliseconds) the first conditions to read and executed are the Long one. You can test this easily by switching the order out of Multiple action block.
As for the 28Apr and the non placed trades:



There is already a long open trade and Go is switched to false --> The short side is blocked : Normal behavior as per the strategy conditions.

Cheers

Hello Amine,


thank you for the next instant answer and sorry for my late answer.


Alright, it is a relief, that the reason for the problem is simple. But now the need of a further adjustment arises: By definition, the strategy processes the data twice: first long and afterwards short. I thought the entry conditions were clear, so that the strategy was able to distinguish between four consecutive green and four consecutive red candles. Let's suppose a short entry occured: The strategy should open short even if it processes the long part first. After processing the long part the strategy should know: we have four consecutive green candles -> long entry condition not met. Proceed with the short entry part.

To get past the problem of long positions after short signals I thought of the following - what do you think of it?
Can it be possible, that there is an issue of overlapping? Can there be interferences between long and short? A candle can be part of both a long signal and of a short signal (please see attached screenshot). Let's suppose we have already three consecutive green candles. Now the likelihood of a short entry is bigger (we just need one more green candle and a doji) than the likelihood of a long entry (we need four red candles and a doji). Can we block the long part so that only the short part will be processed at this point (with an additional "assign" block?)? The blocking will be lifted either after a successful short entry or when no doji occured (i.e. the entry signal was not completed by definition). Supposed the expected short entry was not completed and the blocking was lifted afterwards: A consequence might be that the long part starts running again when a new signal is forming already. But I assume: If the long part (or the strategy as a whole) begins to run after the first of the four consecutive candles already occured, the strategy can catch it nevertheless (with "get historical candle").

Intricate, but a challenge as well! If anything remains unclear, if some of my ideas are unintelligible, please do not hesitate to ask!


Thank you in advance for your patience and your answer!

Yours sincerely
Lunas

Hello Amine,


here's a better version of my screenshot - my last one was a little bit overloaded. Hope, this one helps.


Yours sincerely
Lunas

Hi there,

Here is your strategy "restructured"... the major change is the creation of 2 Boolean triggers "Go Long" and "Go short" as previously the short side was always blocked by "Go" being equal to False.
No the strategy trades on both sides but a small issue arises: There is some attempts to modify a position status while this one is already closed. I'll check this later ...

Hello Amine,


thank you for reading my long descriptions and for taking the time to improve / debug the strategy once again.....seems like it was a lot of work. By the way: You arranged the VJF-blocks very clearly - that is quite helpful to keep track and there are a lot of blocks indeed. The colour of the lines is helpful as well.

I consider this to be a progress: now it is just a small issue and several issues have been occured before and all have been solved. I look forward to hearing from you regarding that "small issue".


Thank you in advance.

Yours sincerely
Lunas

Hello Amine,


thanks again for your reconstruction. At first sight the strategy seems to work as it should. Then I detected a position which was opened one 10sec candle too late and one position which was opened without a signal. The latter one was openend straight after the previous position was closed - is it that what you meant by the attempt to modify the position status? Did you have any success with your investigation concerning this issue?

Besides this little problem: In my backtests with the previous version of my strategy another problem occured which was a little bit annoying: Sometimes no position was opened after a valid signal. What is the reason? As long as a position is open while a new entry signal is forming itsself no position will be opened even if the previous position was closed before the new signal is finished (please see attached screenshot). I was afraid this would need a further post after the current issue would have been solved. But the backtest with the latest version of the strategy turned out a result without that annoying problem. I'll have to do some more backtests but at the moment it looks good. With the reconstruction of my strategy you obviously solved not only the problem with opening a long position after a short signal but also the problem with a blocked valid entry signal! Very impressive!


Thank you in advance for your help and for your answer.

Yours sincerely
Lunas

Hi there,

Here is my findings:

- Regarding your trailing stop flow:
You're using 1mn candle Low price to determine your next price step therefore what's the point of using ONTick start point ?? --> That's the reason why I removed OnTick method as you can see
There was an error related to attempts to change the SL price with the same price when I tested in JForex Historical tester: This is due to the above reason as the trailing step part is checking the the previous candle High and Low on tick basis meaning the execution frequency is based on every tick while the previous candle (1mn) H&L are the same till a new candle is formed --> inconsistent;
- I've implemented a position viewer to get rid of the errors related to "Order state immutable". Position Viewer is much more reliable when it comes to checking previous positions status.

Now regarding your findings with positions not triggered. Can you please point out the date and time of such phenomena or paste a screenshot with full chart picture ?

Thanks!

Hello Amine,


alright, here's the requested data:
running "strategy_with_blocked_entry" on 2017-05-18, 16:22 GMT turns out a blocked entry (please see attached screenshot).
running "strategy_almost_ready_ver3a" on 2017-05-23, 12:16 GMT turns out, that the entry is not blocked (please see attached screenshot).

In contrast to the information in my previous post the "strategy_almost_ready_ver3a" sometimes also blocks entry signals (possibly 2017-05-18, 16:22 GMT, I'm sorry I forgot to jot down the exact time). Because of the fact, that the said strategy is working regularly on 2017-05-23, I'm sure it can work regularly all the time.

Sometimes there are positions without entry signals (2017-05-23).

Strangely the trailer in "strategy_almost_ready_ver3a" doesn't move properly. But we can ignore this issue for the moment because I'd like to use a different version of the trailer (you can see this trailer in the attached strategy). If I made some modifications (such as using "candle open" instead of "last tick"): Is the logic of this trailer also usable with "onCandle"?

You switched the logic of the strategy completely to "onCandle". Is it correct, that this is an advantage for my computer?


As always: Thank you in advance for your continuous and patient help!
Yours sincerely
Lunas

Hi there,

Well, it looks like your requirement and expections are different from what is effectively developed as I can see the following facts that meet the strategy specifications:

1/ at 13:12




2/ at 16:20




The above tests were made on the version 4 I sent you earlier, please use this version for your tests otherwise we'll fall in redundant issues that I already fixed in ver4.




I switched the logic because the OnTick method is useless with regards to your trailing logic which is based on a parameter that belongs to 1mn candle.

Hello Amine,


sorry for the late answer.


Thanks for your friendly advice . You're right: No further excursion into the world of outdated strategies.

Working with your latest version it turned out:
no positions without entry signal
no blocked entry signal

As you noticed earlier than me, there was a discrepancy between the entry specifications in the strategy and my expectation of what exactly is the entry signal. There was no blocked entry at all - just the entry conditions were not met. That's why I remodeled the entry slightly and this new entry signal is working properly.
After finishing the changes in the entry, I tried to change the trailer a little bit as well. I'd like to use a trailer similar to the one in my earlier post: Moving in some bigger steps instead of candle by candle. I was almost successful. In EURUSD the trailer is working properly, in USDJPY it isn't working at all....Is it because of the two decimals in USDJPY instead of four decimals in EURUSD? I tried to find a solution using "PipsSize" according to the "User Guide". But it didn't work.
So after fixing all bugs the following question remains: How can I use a trailer in these instruments together? Please see attached strategies - they are identical apart from the instrument. Trailer in EURUSD is working for example: 2017-05-19, 8:20 GMT
Trailer in USDJPY is never working but a good opportunity to test is: 2017-05-26, 2:11 GMT


Thank you again for your continuous help!

Yours sincerely
Lunas

Hi there !

Here is the right way to use pip size so that your calculations remains dynamic regardless of the traded instrument.

Hello Amine,


thanks for your helpful screenshot. Now the trailer works indeed in all the currency pairs the strategy is meant to trade.


Yours sincerely
Lunas

Hello Amine,


in the meantime I assembled all parts of the strategy, did some more backtests and first real time tests (Demo) as well. And it turned out that there is one problem regarding the trailer, but just occasionally: Sometimes the trailer is working properly and sometimes - without any change in the logic - it isn't working at all. This is happening in backtests and real time as well. What can be the reason for that? If there was a mistake in the logic, the trailer couldn't work at least sometimes. That's why I ask you again to check the trailer. You can backtest the strategy at that time:
2017-06-21 09:54 GMT (trailer is working)
2017-07-03 03:05 GMT (trailer is not working)

For reasons of clarity I removed some of the entry signals. If you need them for your investigation nevertheless, I'll upload the extensive version of the strategy.

Thank you in advance for your answer!


Yours sincerely
Lunas

Hello there !

I've looked at your idea and instead of troubleshooting the error I suggest to look at this alternative solution and test it.
your questions are welcome.

Hello Amine,


great that you thought up a less complex trailer . I did some backtests so far but unfortunately no real time test yet. But it seems like it needs a little adjustment: actually the trailer moves into the wrong direction.......
You can test it yourself: USDJPY: 2017-06-21, 10:11 GMT and 2017-07-03, 3:05 GMT
Do we eventually need a boolean variable (like "trailing_1_Long" -> "true/false") on every trailing step? Further question regarding the use of variables: Isn't it better or even necessary to use unique variables on every entry signal? At the moment I'm using "OpenPositions" and "PositionSize" in every entry signal. Can the global use of this variable be the reason for the trailer not working from time to time? Never thought about it before! But if I change "OpenPositions" into something like "OpenPositions_EURUSD": What are the consequences for the trailer? Finally it also uses the variable "OpenPositions". Can the trailer cope with several variables like "OpenPositions_EURUSD", "OpenPositions_GBPUSD", "OpnePositions_USDJPY"?

While gaining experience on real time test I'm changing one of my entry signals a little (please see attached screenshot). Obviously something like a counter is needed now but can a counter work backward? And can the use of "Get Historical Candles" simplify the entry signal compared to the extensive use of "Get Historical Candle"? I'll try to finish the changed entry signal on my own if you could give me some advice (highly appreciated).

Thank you in advance for your help!


Yours sincerely
Lunas

Hello Amine,


regarding your suggested trailer:
After some more tests using breakpoint it seems that the SL moves into the wrong direction as soon as the data processing reaches the block "Set Stop Loss". Despite the fact, that the SL moves into the wrong direction, I consider your trailer to be a big progress - you know: keep the logic as simple as possible and your suggestion is less complex than the version I'm using currently. I'm sure you already know a smart solution for this problem.

regarding the multi-trailer:
Currently the strategy seems to work properly with two trailers (one for every entry signal - I'm planning to use your above mentioned trailer for my multi-trailer as soon as it works properly). But I do get an error message (please see attached file): "no value...." This is no new phenomenon to me - I'll try to solve this soon. More important in respect of the muli-trailer: Is the general approach (please see attached file) correct?

Unfortunately I did not make any progress regarding the changes in one of the entry signals. As shown in my last post, it is planned to count the green and red candles in a given period of time. I'd like to use "Get Historical Candles" but I haven't a clue how to use its data afterwards. And I also don't know whether a counter can work backward. I would be glad if you could give me some advice how to count the number of past green / red candles.


Thank you in advance for your help!

Yours sincerely
Lunas