How do I check whether 'Optimisation' or 'Forward Optimisation' is in progress? - page 10
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Question - how do I assemble these 12 runs or what do I need to do to cram these runs into such a program?
If you do it in pure MQL with a regular tester, it's quite time consuming. The user will have to reset dates in the tester manually, because we will need standard forward runs after each backward optimization.
Then we collect frames, filter and summarize in a report.
I am talking about how to analyze the RESULTS of a volking forward. And its results are those single backing-forward runs which occur after back optimization.
Yes, but any results for the same group of parameters are included in the set of all results of the full runN, and any beautifulK results are included in the set of beautiful results of the full run i.e.K⊂ NK. I.e. no advantage is given by the volking forward.
The advantage may be in the filtering speed of these nice results. But there is no advantage either, as optimization is performed on repeating chunks. Yes, there can be an advantage in more efficient screening out results which show slippage on some segments. This advantage can be in taking into account the frequency of trades and equity growth at each sector.
It is possible to filter the beautiful results without failures, taking into account the frequency of trades and equity growth by other means, for example with genetic algorithm with its own criteria (which can analyze the entire history of trades of the run evaluating their quality). Which is faster. By the way, no forward testing on the history in fact does not provide any guarantees in the future. You still choose one of the beautiful results, you have no other.
In addition, the data on trades can be written to files or the database only during optimization on your computer. When optimising in the cloud, we can only get a standard report.
i.e. and testing agents working in the cloud return data
Then we collect the frames, filter them and summarise them in a report.
I have a question - I use auto-optimizer first to optimize the back segment, and then a forward run, which contains both back and forward segments.
This is a single run, its purpose - control. I judge the EA's performance by the results of twelve passes with a shift of backward step.
What should I do to make your program be able to display the balance graphs of all twelve of these runs? That is, make a report of the individual back-end runs.
What should I do so that your software can display the balance graphs of all twelve of these runs? That is, make a report from the individual back-end runs.
My program doesn't (and can't) do that. This is pure MQL variant for the 1st regular optimization run with or without forward. Try to collect the history of balance increments for several runs, in one or more files, and then display it in excel, for example.
I have a question - I use the auto-optimizer to optimise the back site first, and then a forward run, which contains both the back and forward sites.
It is a single run, and its purpose is control. I judge the EA's performance by the results of twelve passes with a shift of backward step.
What should I do to make your program be able to display the balance graphs of all twelve of these runs? I.e. make a report of the individual back-end runs.
I save the settings I like from optimization to the standard .set file 2016-01-01-01.opt, 2016-02-01-01.opt, 2016-03-01.opt etc. I put all the files in the Files directory of the terminal.
I copy the input variables in the Expert Advisor to the usual duplicate variables, which are used in the Expert Advisor instead of the input variables.
Then when the EA runs on the first tick of each day I check if there is a file for the current day. Feb. 1 is found 2016-02-01.opt and it is counted. All values from the file are copied to the working variables. And the Expert Advisor behaves as if on Feb. 1 you have changed the input variables. Thus, input variables required for the current plot are substituted several times during the entire testing period.
As a result, we have a balance and equity chart for all forward sections.
This method also takes into account open positions at the time of a lot change. Positions opened before February 1st, according to the rules of January 1st, will be reproduced by the Expert Advisor according to the new settings (which is actually correct).
I save the settings I like from the optimization to the standard .set file 2016-01-01.opt, 2016-02-01.opt, 2016-03-01.opt etc. I put all the files in the Files directory of the terminal.
I copy the input variables in the Expert Advisor to the usual duplicate variables, which are used in the Expert Advisor instead of the input variables.
Then when the advisor runs on the first tick of each day, I check if there is a file for the current day. 1 feb. is found 2016-02-01.opt and counts. All values from the file are copied to the working variables. And the Expert Advisor behaves as if on Feb. 1 you have changed the input variables. Thus, input variables needed for the current plot are substituted several times during the whole testing period.
As a result, we have a balance and equity chart for all forward sections.
This method also takes into account open positions at the time of a lot change. Positions opened before February 1st, according to the rules of January 1st, will be reproduced by the Expert Advisor according to the new settings (which is actually correct).
Well, yes, it's also an option. The disadvantage is that you have to wait until the end of the whole process to see the overall picture.
Sometimes I terminate the process if I see that something went wrong.
In addition, the probability of unsynchronization of history of different tools increases with increasing history - but it is a rare event.
Otherwise, I do not need to bother with combining charts.
Well, yes, that's also an option. The disadvantage is that you have to wait until the end of the whole process to see the big picture.
Sometimes I interrupt the process if I see that something has gone wrong
In addition, the probability of unsynchronization of history of different tools increases with increasing history - but this event is quite rare.
Otherwise, you don't need to bother combining charts.
Why wait? You have made 2 passes of the forward, saved 2 files - and start the test for 2 months (or other timeframes that you have chosen). And you will get the result for 2 timeframes.
For example, if I see, that there is a strong drawdown, then I do a new forward until the day following the drawdown, for example 2016-03-12.opt. On the assumption that the market has changed and re-optimization is needed.
Why wait? Make 2 forward passes, save 2 files - and run the test for 2 months (or other time periods of your choice). And you will get results for 2 sections.
I can not interrupt - Autotester is running continuously, and if you can not see the intermediate results - is not clear to stop or not.
And, by the way, having a general graph, it will be difficult to understand the situation by segments 0, say, this month is profitable or not.
I can't interrupt - the autotester runs continuously, and if you can't see intermediate results - it's not clear if you should stop or not.
I do it manually, the autotester hasn't figured it out yet. For my method - all ok. For yours, it seems that something needs to be improved.
And by the way, having a general graph, it would be difficult to understand the situation by segment 0, say, this month is profitable or not.
Why is it not visible? You can see it on the chart in the tester - just hover your mouse over the curve and you see the date of the deal.
I do it manually, I haven't invented an autotester yet. For my method, it's OK. For yours, it looks like something needs to be worked out.
Why is it not visible? You can see it on the chart in the tester - put your mouse on the curve and you see the date of the deal.