Machine learning in trading: theory, models, practice and algo-trading - page 1853

 
Aleksey Vyazmikin:

I do not hang on the gluing - in general, we need to deal with it. You cannot check and send me the working version ;)

Alexei, I have been working on the Expert Advisor and IDnikator. I have added the delta record in my Expert Advisor and checked it with mega cool indicator that has a lot of checks and so far all readings are similar on some minutes. I have to modify the indicator. I have an Openvaya broker where the EA is working as I have laid it out before and am laying it out now. I have added my own delta calculation. No thanks :-) I've already added the indicator and I'll pay 37 rubles for the indicator.

Files:
 

Alexey, I'm sorry, but it looks like our deal is off. I have torn down a lot of your code in the indicator, which, as practice has shown, was unnecessary, because now I compile it and it loads the stories as it should be. I hope it will not spoil our relationship. To be honest I am ashamed of these words, but as a result I have made delta and corrected the indicator at least to the state it had, but the fact that it now draws the history as it should. It only remains to organize the first bar. But I think I can handle it.

All in all, thank you for your help, I appreciate it. By the way, do you know the binary classification models with the teacher?

 
Mihail Marchukajtes:

Alexey, I'm sorry, but it looks like our deal is off. I have torn down a lot of your code in the indicator, which, as practice has shown, was unnecessary, because now I compile it and it loads the stories as it should be. I hope it will not spoil our relationship. To be honest I am ashamed of these words, but as a result I have made delta and corrected the indicator at least to the state it had, but the fact that it now draws the history as it should. It only remains to organize the first bar. But I think I can handle it.

All in all, thank you for your help, I appreciate it. By the way, do you know the binary classification models with the teacher?

So, put that "demolished". I will learn to optimize the code:)))

And the logic is interesting, about payment, if you did it yourself, then you do not need to pay - I do not take offense, because I did just for the fun.

Glad to help.

Yes, I also use the binary classification.

 
I'll be sure to post it. I've already done the reading of the missing lines and now I just need to write them into the buffers.
 

I was finally able to make this indicator to work in real time, but at each new minute bar it loads the entire file. I have not succeeded in making it load only the missing data. If the history file is large enough, it will have a noticeable impact on resource consumption. But the indicator buffer is guaranteed to contain all the data that is in the file....

The advantage of this combination is that in addition to the TOI in the file is also written delta, which saves you from requesting a copy with an astronomical number of cycles when requesting a less than long history

Files:
OI_Test.mq5  13 kb
 
Mihail Marchukajtes:

I was able to make this indicator work in real time, but at each new minute bar it is loading the entire file. I have not succeeded in making it load only the missing data. If the history file is large enough, it will have a noticeable impact on resource consumption. But the indicator buffer is guaranteed to contain all the data that is in the file....

The advantage of this combination is that delta is written to the file in addition to the OM, that avoids the need to request a copy with an astronomical number of cycles, when requesting a more or less long history.

Yeah, clearly overdid it - the running time in the tester of your version

2020.07.03 01:13:23.234 Si Splice,M1: total time from login to stop testing 0:00:37.099 (including 0:00:00.976 for history data synchronization)

of mine is .

2020.07.03 01:14:40.128 Si Splice,M1: total time from login to stop testing 0:00:03.368 (including 0:00:00.796 for history data synchronization)

Removed everything that was done to speed up and catch rare bugs.

I still do not understand what works there wrong :( Peeping into the future, I can do as you want of course, but reading from a file of current data - tinny.

The indicator in the tester draws the garbage based on historical data :( I can't test the models.

 
Aleksey Vyazmikin:

Yes, they clearly overdid it - the running time in the tester of your version of

mine

Removed all that was done to speed up and catch rare bugs.

I still do not understand what works there wrong :( I can do a peek into the future, as you want of course, but reading from a file of current data - tinny.

And your indicator in the tester draws the garbage based on historical data :( I can't test the models.

Strangely, when compiling the buffer is drawn according to the information in the file. What it slows down, so it completely reloads the entire file on each bar. And most likely there will be problems with the tester. Too bad :-(
 
Mihail Marchukajtes:
Strangely, when compiling, the buffer is drawn according to the information in the file. The fact that it slows down, so it reloads the entire file on every bar. And most likely there will be problems with the tester. Too bad :-(

In the tester, try to see how it works in terms of rendering. I haven't checked it with the current data.

In the tester it should not slow down, because reading only goes if the file has changed, and it does not change, and therefore it is not loaded.

You can send data through global variables of the terminal - when you write the new data you transfer it to the buffer indicator, but it's for the real world. In general, I would not worry, the bar closing data is correct in the indicator, even if it differs a little from the file.

 
Aleksey Vyazmikin:

In the tester, try to see how it works in terms of rendering. I have not checked on the current data.

In the tester should not slow down, because reading only goes if the file has changed, and it does not change, and therefore is not loaded.

You can send data through global variables of the terminal - when you write the new data you transfer it to the buffer indicator, but it's for the real world. I would not bother at all, the bar closing data is correct in the indicator, even if it differs a little from the file.

I would not see any difference in my version. I would not worry about the tester. When I place the trading Expert Advisor on the chart, it is initialized with arrows, and then these arrows are saved in the file for the tester. Then I use a special Expert Advisor to trade it in the Strategy Tester. The main thing is that in real life it does everything correctly. That it could be left unattended, for some time, of course :-)
 
Mihail Marchukajtes:
Well, in my version there were no differences. I'm not worried about the tester. When I place the trading advisor on the chart, I initialize it with placing arrows and then I save these arrows into a file for the tester. Then I use a special Expert Advisor to trade it in the Strategy Tester. The main thing is that in real life it does everything correctly. That it could be left unattended for some time, of course :-).

If I use your indicator it will not show the history from the file after I loaded it on the chart, for example on TF M15.

Reason: