well, if these posts of mine will be deleted after the complaint of Dick, who takes any criticism, even in very restrained tones, as a personal insult, then my FI to the moderator who will do it.
Nikolai Semko #: I don't understand why you talk about hills and peaks as something static. The market is not static! After all, it is absolutely obvious, if you get a new FF surface every day, for example, when "optimising" two parameters, and then glue the obtained frames together, you will get something like this:
So what if you even caught the right hill? It's a hill of history, but not of the future. What's the point? So I agree with Dmitrievsky. Fitting to history is still fitting, even if you call it optimisation by the Macaca, Crane or Octopus method.
In a changing environment (in this case, the optimisation surface) it is not important where you stop at a given point in time, but where this point will go down or up at the next point in time. Moreover, and this is not important, but it is important whether it will move upwards with acceleration or with deceleration. I.e. if you choose a point on a growing hill, but at the next moment the speed of growth starts to slow down, it will be a worse decision than if you choose a point on a falling trough, but if at the next moment the speed of fall slows down. Without a predictive model with probability > 55% any strategy is nothing.
Back to the question. 50 found vertices should be run on OOS. And if a conditional InputsMax1 is found among them, we dig further. If not - throw it away (change the symbol).
It is too presumptuous to think to change a person's mind on the Internet, I have heard your point of view.
I am not in the mood to engage in dialogue in a disrespectful manner.
There was a championship of optimisation algorithms, where there was an opportunity to show your knowledge and skills, but somehow I didn't see people drawing poop there now.
It would probably be convenient to have a browser extension for each MQL5 branch, where you could ask not to show certain posts. I would like to see only constructive posts.
You started to write about gcp comparison - failure. You started explaining about robustness - only wasted people's time. You'd better do what you know how to do - rewriting optimisation algorithms in mql. Teacher.
If some regularity in prices can be reproduced by TC, then the sought set will correspond to some local top of FF. Some high tops will correspond to non-systematic white swans that were on Sample. Because of them, lower but potentially stable sets of inputs are missed on classical AOs.
A simple statement that is easily tested in practice. Take almost any TS with a large number of non-overlapping positions. For example, 10 such positions per day. Find the global maximum of InputsMax1 for the whole year 2023 and InputsMax2 for the summer of 2023. Obviously, in the summer of 2023, no AO will find InputsMax1 even close. But you will find that among the local vertices for summer 2023 there is one that is very close to InputsMax1.
Probably just a very similar experiment was just conducted in the last article. Indeed, the best results on the forward period were those sets that performed well but not the best on the main period.
Allow the use of cookies to log in to the MQL5.com website.
Please enable the necessary setting in your browser, otherwise you will not be able to log in.
You are signed in. Two-factor authentication has been disabled. Set it up again and scan a new QR code, or continue with two-factor authentication disabled.
I don't understand why you talk about hills and peaks as something static. The market is not static!
After all, it is absolutely obvious, if you get a new FF surface every day, for example, when "optimising" two parameters, and then glue the obtained frames together, you will get something like this:
So what if you even caught the right hill? It's a hill of history, but not of the future. What's the point?
So I agree with Dmitrievsky. Fitting to history is still fitting, even if you call it optimisation by the Macaca, Crane or Octopus method.
it has already been discussed
Forum on trading, automated trading systems and testing trading strategies.
Discussion in article "Population Optimisation Algorithms: Saplings Sowing and Growing up (SSG) Algorithm".
Nikolai Semko, 2023.03.24 00:01
In a changing environment (in this case, the optimisation surface) it is not important where you stop at a given point in time, but where this point will go down or up at the next point in time. Moreover, and this is not important, but it is important whether it will move upwards with acceleration or with deceleration.
I.e. if you choose a point on a growing hill, but at the next moment the speed of growth starts to slow down, it will be a worse decision than if you choose a point on a falling trough, but if at the next moment the speed of fall slows down.
Without a predictive model with probability > 55% any strategy is nothing.
it's already been discussed
Forum on trading, automated trading systems and testing trading strategies
Discussion of the article "The role of random number generator quality in the efficiency of optimisation algorithms"
fxsaber, 2024.04.08 23:19
Back to the question. 50 found vertices should be run on OOS. And if a conditional InputsMax1 is found among them, we dig further. If not - throw it away (change the symbol).
I am not in the mood to engage in dialogue in a disrespectful manner.
There was a championship of optimisation algorithms, where there was an opportunity to show your knowledge and skills, but somehow I didn't see people drawing poop there now.
It's too presumptuous to think to change someone's mind on the internet
You're right, but the inability to change someone's mind on the internet isn't the only problem.
messages will be deleted
It would probably be convenient to have a browser extension for each MQL5 branch, where you could ask not to show certain posts. I would like to see only constructive posts.
If some regularity in prices can be reproduced by TC, then the sought set will correspond to some local top of FF. Some high tops will correspond to non-systematic white swans that were on Sample. Because of them, lower but potentially stable sets of inputs are missed on classical AOs.
A simple statement that is easily tested in practice. Take almost any TS with a large number of non-overlapping positions. For example, 10 such positions per day. Find the global maximum of InputsMax1 for the whole year 2023 and InputsMax2 for the summer of 2023. Obviously, in the summer of 2023, no AO will find InputsMax1 even close. But you will find that among the local vertices for summer 2023 there is one that is very close to InputsMax1.
Probably just a very similar experiment was just conducted in the last article. Indeed, the best results on the forward period were those sets that performed well but not the best on the main period.