![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
The idea of non/weakly stretch-sensitive pattern searching on the history was chewed up at 4 about half a year ago. I didn't get involved, but I looked into it. They seemed to have found some catching features there.
And what's wrong with Renko preprocessing? It's supposed to fix the pattern in the context of the channel.
--
And besides, you were waving the cognitron in public. I remember! It's got a twist and a variable distance. Go for it... :)
// But it is possible and necessary to steal ideas from there (from the cognitron).
Well, the price array is a "smooth function". Non-linear distortions can also be approximated by piecewise linear, we will get coincidence of 90-95%, but it's not so far from fitting, we need reasonable balance. In general, there are a lot of thoughts, I'm itching to code it all, but where to get 128 hours in a day?
Yes, and none of the network specialists here have not tried to recognize channels before?
Well, the price array is a "smooth function". Non-linear distortions can also be approximated by piecewise linear, we will get coincidence of 90-95%, but it's not so far from fitting, we need reasonable balance. In general there are a lot of ideas, I'm itching to code it all, but where to get 128 hours in a day?
Yeah, and haven't any of the networking experts here tried to recognise channels before?
Once again for the especially gifted: recognizing patterns is not a problem, because there are a bunch of algorithms for that. The problem is that you can recognize a pattern only after it's rendered.
Nobody needs history patterns, waves, levels, candlestick combinations and other stuff because the past is already known, regardless of whether it is recognized and broken down into patterns or not. The trader needs to know what the next pattern will be. And no algorithms will recognize the next pattern, because it hasn't existed yet.
>Recognising patterns is not a problem
That's a given, the question was about channels.
>The problem is that a pattern can only be recognised after it has been drawn.
Where's the problem if it's the non-tradable part of the pattern that needs to be recognized?
>Recognising patterns is not a problem
That's a given, the question was about channels.
>The problem is that a pattern can only be recognised after it has been drawn.
Where's the problem if it's the non-tradable part of the pattern that needs to be recognized?
Here is a puzzle for you: suppose you have detected some non-trading part of the pattern. You come in your trousers for joy, share the news with the international community through a blog or a forum, wash your hands of the news and get drunk like a pig. And then what? This is what happens next:
gpwr:
...
I've noticed this pattern, the longer the pattern and the greater the number of them in the dictionary, the higher the bucktest profit, which is understandable by overtraining. But anyway, with different N and L, the forward test looks chattering around zero profit. Starting to get frustrated with the patterns. Apparently they are not constant in forex, or in other words forex has no memory for patterns - new ones are created every time.
It's one thing to recognise, for example, symbols or a piece of speech to digitise, or an enemy plane to shoot down. It's another thing to do botany, i.e. recognition for the sake of recognition.
There are no problems with recognition. The problem is monetising what is recognised.
Once again, for the very gifted: recognising patterns is not a problem, because there are plenty of algorithms to do this. The problem is that a pattern can only be recognised after it has been drawn.
Nobody needs history patterns, waves, levels, candlestick combinations and other stuff, because the past is already known, whether it is recognized and broken down into patterns or not. The trader needs to know what the next pattern will be. And no algorithms will recognize the next pattern, because it hasn't existed yet.
I showed you on the first page the prediction of the incomplete pattern, the point 3 (EURUSD = 1.300), that worked out very accurately as well as the rebounce after that touch. That's a profitable trade for you. I now predict a breakout of the upper boundary. Buy. Notice that all these predictions are in real time. What's wrong with that? I've already asked you to post your predictions. Let's compare them. Maybe we can combine several successful strategies into one, and something interesting will come out. In the future I will show you step by step how a new channel is formed and where the orders are placed.
I showed you on the first page the prediction of the incomplete pattern, point 3 (EURUSD = 1.300). It worked out with great accuracy, as well as the rebound after this touch. That's a profitable trade for you. I now predict a breakout of the upper boundary. Buy. Notice that all these predictions are in real time. What's wrong with that?
In the future I will show you step by step how a new channel is formed and where the orders are placed.
A puzzle question: suppose you recognise a non-tradable part of the pattern. You come in your trousers for joy, share the news with the international community through a blog or a forum, wash your hands of the news and get drunk like a pig. And then what? Here's what happens next:
This is bullshit! All we have to do to recognise a dog is to look at its face and make a pretty good guess about what the rest of its body looks like.
Why are you so bitter? Are the neurons killing you? Or have you already donated your own neurons as artificial ones don't work?
Once again, for the very gifted: recognising patterns is not a problem, because there are plenty of algorithms to do this. The problem is that a pattern can only be recognised after it has been rendered.
Recognition of the whole by its part is a task that neural networks (especially multilayers) handle quite well.
Nobody needs history patterns, waves, levels, candlestick combinations and other stuff, because the past is already known, no matter if it is recognized and broken down into patterns or not. The trader needs to know what the next pattern will be. And no algorithms will recognize the next pattern, because it hasn't existed yet.
Besides, there are [relatively] stable pattern sequences. One of them is exactly what the author is promoting.
--Yura, don't exaggerate.
Recognition of the whole by its part is a task that neural nets (especially multilayers) are quite good at.
Besides, there are [relatively] stable sequences of patterns.
--Yura, don't exaggerate, the approach is quite workable.