![MQL5 - Language of trade strategies built-in the MetaTrader 5 client terminal](https://c.mql5.com/i/registerlandings/logo-2.png)
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
You can't see it.
It's not that simple function Z = cos(1.5*x)*cos(1.5*x) + sin(2.25*y) + cos(3*x*y); where X and Y are from -3 to +3 that I discussed in the previous article.
Not only that, but the author of that article actually built a huge bicycle (which is good for self-study, of course), but apparently he sharpened the search optimization for his task. This optimization will most likely cause problems (increase in amount of calculations) with other tasks.
There are also important metrics left out - how many passes were actually made in heuristic mode compared to a full overshoot. For example, in the MT5 example above we got 8,700 in genetics and 361,201 in brute force. I suspect that the author's own heuristic optimized variant actually took much more passes to complete the results.
The number of passes is very important, since rarely any strategy meets the deadline of seconds. The difference between our GA with 10,000 passes and another one with 30,000 passes results in waiting for an extra 20,000 passes * time of pass, which is very long. Our GA is specifically optimised for the fastest possible miscalculation. Usually ours is sufficient for 10,000 to 12,000 passes regardless of the total size of the search field. This means that any depth of search can be carried out roughly in 10,000 passes. The next step is to put your head in your hands and explore more accurately.
By the way, in MetaTrader 5 the author wouldn't have had to spend months writing his own engine, and could have obtained results at the click of a button. And in 2D/3D spin in different projections.
Please understand that we are not talking about the craftsmanship of the author of this article and not the author. We are talking about arguments he voiced as regards the logical way of seeing a solution to a heuristic algorithm. And those arguments are valid. I don't have any complaints about your GA, your platform and all that. I just suggested allowing to write a custom heuristic algorithm in MQL5. But you are praising almost every word and suggest trying to implement your own custom algorithm based on changes of your GA. Why the hell should I try to do it through my ass if I can write my own algorithm humanely, without crutches? If you do not give such an opportunity - just say that you do not see the feasibility. You've banned and banned everyone who tried to argue the necessity of what you are now introducing.
Well enlighten us, about economic feasibility(can you believe it, I'm so dumb I still don't get it).
Is it that from the area of the dollar exchange rate has changed? Like a year ago it was different, and now the work of programmers has become more economically viable (they need to be paid less)?
And that is why they have only now started to cook up code and provide such a feature, which has existed for decades in other trading platforms. Is that how you see it?
The answer is here: https://www.mql5.com/ru/forum/23 step by step.
There are so many suicide bombers and good advisers in the market that the one who does the right thing at the right time survives. Look at what we have built by consistent change over all this time.
All that matters is what's done, not promises, advice or ideas. We have implemented a lot of functionality rather than spitting at the ceiling.
Now it is time to reveal more interfaces, which we are doing. In addition to datafeeds, we will still unlock a large layer of features. Wait for the implementation, please.
You, unfortunately, are wrong.
For a quick look, just to understand the essence of optimization, try to understand the answers to these questions:
Dear Sir, I'll offer you not a hypothetical, but a real story of my practice of optimizing many TS that worked - brought incalculable profits.
Imagine I give you such a TS and say it works. And it is 100% true. Just imagine. But you have the task to find its working input parameters, which are not few, i.e. you need a heuristic optimization.
You launch your GA and find several thousand good combinations of such parameters. You chose among them according to some criteria of your own and launched them. But they all turned out to be a fit - well, it happens that way. After that you tell me that TC ***: not working.
But it is working, just that area of parameters, which really works, your GA simply ignored. It was producing random (non-systemic) extremums during the search rather than those where there was a pattern.
Such a situation has happened to me more than once. And improvement of GA through defining other optimization criteria did not lead to a significant improvement. This is why sometimes it made sense to create my own under-tester and run it through a full search in ten or two hours (in MT it can take months). And then in a huge matrix to dig out a workable variant. Or an alternative - a better heuristic algorithm.
joo:
Is it not clear that any functions, any features and technologies in any particular product appear when and only when it is economically feasible? - It is now, and it is for MT5 that such an economic necessity has arisen and will be implemented, not a year earlier and not a year later. We are doing it now. Yes, in some other products such possibilities have existed for a long time, but there were no other possibilities, that MT gives from time immemorial.
Understand that we are not talking about the author's work, nor about the author. We are talking about the arguments he voiced for the logical solution to the heuristic algorithm. And those arguments are valid. I don't have any complaints about your GA, your platform and all that. I just suggested allowing to write a custom heuristic algorithm in MQL5. But you are praising almost every word and suggest trying to implement your own custom algorithm based on changes of your GA. Why the hell should I work through my ass if I can manually write my own algorithm without crutches? If you do not give such an opportunity - just say that you do not see the feasibility. You've banned and banned everyone who tried to argue the necessity of what you are now introducing.
The arguments are theoretical. But most importantly in the narrow sub-topic of your particular strategy.
To make you understand my position - the beauty of going around every local extremum from all sides is fraught with a multiple of the required number of passes. The idea has a right to life, of course.
I must point out to you that it was you who made negative statements in our direction and then instantly switched to my personality. In fact instead of substantiating the claims against the GA you engaged in my persona. I take it this isn't the first time we've spoken in public.
Do not think about bans - this is not the Institute for Noble Maidens. If you are clearly inadequate and outright malicious, the result is predictable.
Imagine that I give you such a TS and say that it works. And it is 100% true. Just imagine. But you face the task of finding its working input parameters, which are not few, i.e. a heuristic optimization is required.
You run your GA and find several thousand good combinations of these parameters. You choose from them on the basis of some criteria as if they were working and run them. And they all turned out to be a fit - well, it happens that way. After that you declare to me that TC ***: not working.
So you think the GA should produce clean results?
That's a fundamentally flawed approach. GA should be used to quickly find possible clusters of good solutions, and then already look for a full enumeration inside a small field of each cluster.
But it works, it's just that the field of parameters, which really works, your GA simply ignored. It was producing random (non-systemic) extremums in the search rather than those where there was a pattern.
For this you need to run GA (as well as any other, no silver bullet) optimization several times. Random will allow you to jump out of periodically stuck local extrema.
And there are no such methods that reduce the counting field by orders of magnitude and "don't miss the gold mine". There are no such methods.
That is why all methods are approximately in the same conditions.
I have had this situation happen to me more than once. And improvement of GA through setting other optimization criteria did not lead to a noticeable improvement. That is why sometimes it was reasonable to improvise my own under-tester and run it through for a dozen or two hours (in MT it can take months). And then in a huge matrix to dig out a workable variant. Or an alternative - a better heuristic algorithm.
You run your GA and find several thousand good combinations of these parameters. You choose from them on the basis of some criteria as if they were working and run them. And they all turned out to be a fit - well, it happens that way. After that you tell me that TC ***: not working.
But it is working, just that area of parameters, which really works, your GA simply ignored. It was producing random (non-systemic) extremums during the search rather than those where there was a pattern.
Such a situation has happened to me more than once. And improvement of GA through defining other optimization criteria did not lead to a significant improvement. This is why sometimes it made sense to create my own under-tester and run it through a full search in ten or two hours (in MT it can take months). And then in a huge matrix to dig out a workable variant. Or an alternative, a more sophisticated heuristic algorithm.
1. How was the "working" out of the total number of complete brute force (what criteria)?
2. Why the same principles (criteria) as in p1. were not applied to optimization with GA?
Don't answer me, please. Answer these questions for yourself.
Undoubtedly. Just don't paint some people as fluffy and smooth, who have previously responded to requests to implement such things not with arguments about economic inexpediency, but with scathing, labeling and crushing granite "experience" that they know better than traders what they need. Those who do not agree are fools, fifth columnists, provocateurs, competitors and ... in short, a ban. Or have you lost your memory?
You're overreacting. Half of your posts in this thread are filled with insults and personal attitudes.
Not least for such whining and constant flag-pulling of political discontent people went to the ban.
Bah, it's hrenfx who asked for his account to be deleted. It all made sense right away.
The arguments are theoretical. But most importantly in the narrow sub-plot of your particular strategy.
To make you understand my position - the beauty of traversing each local extremum from all sides is fraught with a multiple of the number of passes required. The idea has a right to life, of course.
But that is not the truth! The arguments were of a general nature: for any TS.
I must point out to you that it was you who made negative statements in our direction and then instantly switched to my personality. In fact, instead of substantiating the claims against the GA, you engaged in my persona.
The answer is here: https://www.mql5.com/ru/forum/23 step by step.
There are so many suicide bombers and good advisers in the market that the one who does the right thing at the right time survives. Look at what we have built by consistent change over all this time.
All that matters is what's done, not promises, advice or ideas. We have implemented a lot of functionality rather than spitting at the ceiling.
Now it is time to reveal more interfaces, which is what we are doing. In addition to datafeeds, we will still unlock a large layer of features. Wait for the implementation, please.
That you have done a lot, I certainly agree, and there was a time when I thought MT was the best trading platform (there is no better), but gradually over time I have come across limitations, limitations that are not found in other platforms.
For example, in MT4 you could work with your own, and in MT5 you prohibited to do that, exactly prohibited. And by doing so you deprived a huge number of users of opportunities.
All that matters is what isdone, not promises, advice or ideas.
Yes, a lot has been done, and no one needs empty promises or stupid advice. But ideas, ideas are what counts, ideas are what stands at the beginning of the road. Not the copying of competitors' technological solutions (like we looked at theirs, waited a couple of years .... works, let's do the same in our software), so the competition won't survive. It is the idea that lies at the head of the corner - it is the idea (its implementation) that gives the competitive edge over others, and attracts new users.
Z.U. And ideas here on the forum and suggestions for improving the platform, laid out a lot, all kinds were both good and bad, but the lack of them certainly was not. Don't pass them by...