Methods of carrying out a rolling forward - page 4
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
Hm, ok. But if there is optimization there, then we still end up with a set on the history.
A set is an intermediate characteristic of an EA operating within a certain period of history. As soon as I get the results of forward testing, I load an intermediate set for a new segment and it is overwritten by the next one. The only thing that matters is the starting set, and even then, the starting conditions have to be equal.
Some kind of incomprehensible testing. Do you end up with a set of parameters or not?
Here is an extract from Wikipedia
And I don't argue that this method of elimination is good for full optimization without genetics. As an alternative. But the final results still find parameters to find beautiful balance curves at some final interval of history. Which, if available, will also be found by genetics.Some kind of incomprehensible testing. Do you end up with a set of parameters or not?
Here is an excerpt from Wikipedia
Walk forward testing allows us to develop a trading system while maintaining a reasonable 'degree of freedom'.
From the same place.
Walk forward testing allows us to develop a trading system while maintaining a reasonable 'degree of freedom'.
From the same place.
Agreed. If it's used for development and monitoring the influence of certain modifications - very good.
Yes, an important point mentioned in the article - it allows you to find the parameters that can be optimized based on the history. You may work with your trading robot in the same way: optimize these parameters after a period of time, set a new set for the bot.
I.e., it helps to write Expert Advisors that can and should be reoptimised.
Agreed. If for development and tracking the impact of certain modifications - very good.
Yes, an important point that is highlighted in the article - it allows you to find the parameters that make sense to optimise by history. You may work with your trading robot in the same way: optimize these parameters after a period of time, set a new set for the bot.
I.e., it helps to write Expert Advisors that can and should be reoptimised.
Here's the best one.
Demand one of these from Metacwot with your hands anyway. In the long run they will, of course, but it may take years, but in the meantime, make an autotester.
Unlike Uncle Fyodor (from Prostokvashino, and not from MQL5.com), you make proper sandwiches :)
Here's the application to the SD, look at the date, the application is open, they say we'll do it.
Here is the application to the SR, look at the date, the application is open, they say we will do it.
I had a conversation with the MC in one of the threads about this. There is a waiting list. In order to jump the queue you have to
1. or some extreme interest of the trading public
2 .or their own conviction or social responsibility of the developers.
The trading public is used to searching under a torch, and those who understand that this makes no sense usually make their own little torch.
Developers themselves are not traders, they are also influenced by myths and habits - hence the hierarchy of priorities oriented towards the majority.
But that doesn't mean we shouldn't discuss it, which is what we do here.)
For example in the next thread we found a good solution to the problem of visual representation of forwards for a volleying forward - it's just necessary to adjust the timeline, likeIgor Volodin did in his toolzine. You just need to make forward alignment.
Some kind of incomprehensible testing. Do you end up with a set of parameters or not?
Here's an excerpt from wikipedia
And I don't argue that this method of sifting is good for full optimization without genetics. As an alternative.Let's just separate optimisation and testing. Volking is "staggering". In my opinion, testing at non-optimized sections with a forward step is a typical stepwisetesting.
Moreover, the main idea of volking is the inertia of the market, or rather to check how much and for how long our EA is inertial in an unfamiliar environment. I.e. volking simulates the situation where we, just as in life, from time to time optimize our EA and run it in the roaring ocean of an unpredictable market.
The optimization can of course be different, including custom criteria, pulling from sets of sets, and whatever else traders dream up.
But the essence of testing is that we assume and reality disposes. No matter what we have done with the Expert Advisor - whether we have changed its code or picked a special set - testing simply shows how successful it is.
Therefore, the recipe is simple - perform a wolfing forward test for your method of set selection and compare it with the result of the wolfing forward test for another method of setting selection and everything will become clear at once.
Yuri, it would be nice if you could describe (and in general it would be cool if you could show the screenshots on Photoshop) - how do you think volking-forward testing should be done in MetaQuotes tester.
How to specify range and periods, output intermediate results, what to get at the output, etc.
I think it could save time on application review and remove some of the questions from forum readers.
Yuri, it would be nice if you could describe (and in general it would be cool if you could show the screenshots on Photoshop) - how do you think volking-forward testing should be done in MetaQuotes tester.
How to specify range and periods, output intermediate results, what to get at the output, etc.
I think it could save time on application review and remove some of the questions from forum readers.