Discussion of article "Brute force approach to pattern search" - 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
Bruteforce and neural networks are a necessary thing to find patterns. Just some people use it incorrectly and look for the wrong place or the wrong thing, or are limited in resources, again due to the incorrect application of algorithms or programmes implementing them. For example, on mql4 or mql5, it is very difficult to implement it, no matter how much they praise this platform, it is designed initially not for these methods. But if you do it quite differently and search for the necessary data, and make the sample bigger than it is possible on mql, then these patterns are literally visible to the naked eye. For example, there is always a pattern between different currency pairs, you just need to look for them "correctly".
Bruteforce and neural networks are a necessary thing to find patterns. Just some people use it incorrectly and look for the wrong place or the wrong thing, or are limited in resources, again due to the incorrect application of algorithms or programmes implementing them. For example, on mql4 or mql5, it is very difficult to implement it, no matter how much they praise this platform, it is designed initially not for these methods. But if you do it quite differently and search for the necessary data, and make the sample bigger than it is possible on mql, then these patterns are literally visible to the naked eye. For example, there is always a pattern between different currency pairs, you just need to look for them "correctly".
I agree, I have just about to publish the third article on this branch ) just on this topic, and the fourth one is planned ). here you can even see for interest at what stage of development of the method ). this video in the fourth article will be only
https://www.youtube.com/watch?v=LhY_dN6d0lE
We need to make the machine learning to understand the profitability and the pattern relationship.And this is well explained in the article.We dont' need it to understand what the trend is.
My experience is neural networks dont' work in strong trends ,you can check it with bitcoin,crypto currency data series.
I just need to reply this article. Patterns are just a small piece of the overall picture.
The first degree polynomial looks better.
The ratio of optimisation period to profitability is 4 to 1, I have seen it not for the first time.
Half-waves, this is from probability theory, there are formulas for the maximum spread, the number of crossing 0, the probability of returning to 0.
Look at the link, by leaving only 1 degree you do linearisation, you can replace bruteforce with solving equations.
The first degree polynomial looks better.
The ratio of optimisation period to profitability is 4 to 1, I have seen it not for the first time.
Half-waves, this is from probability theory, there are formulas for the maximum spread, the number of crossing 0, the probability of returning to 0.
Look at the link, by leaving only 1 degree you do linearisation, you can replace bruteforce with solving equations.
leaving the second degree we work with the regression channel, with the 3rd degree we start to take into account deviations within the channel.
That's how it is, without details and fanaticism.
The first degree polynomial looks better.
The ratio of optimisation period to profitability is 4 to 1, I have seen it not for the first time.
Half-waves, this is from probability theory, there are formulas for the maximum spread, the number of crossing 0, the probability of returning to 0.
Look at the link, by leaving only 1 degree you do linearisation, you can replace bruteforce with solving equations.
The classical Taylor series is made for a function of one variable, I am using a version for a function with an unlimited number of dimensions. Although to be honest I certainly use only the first degree. simply the polynomial becomes the sum of the coefficients multiplied by the price shifts at each bar. In the end, the deciding factor is not the type of formula itself, but the frequency of the search. In general, it doesn't matter, it gives the result. Further you can roll out modifications, correct something.
Hi, your work and especially your thoughts are very interesting. You wrote, that both versions mt4 and mt5 are attached. I can't find a mt5 version! ?
Please add if possible.
Many thanks