Yoghurt systems and canned systems or The relationship between trading tactics and the reliability of historical test results - page 15
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
Let's make it clear: to tell in a topic about patterns used in TC or to train a network is not the same thing !
What happens when you train a network? The network uses the data you gave it to look for patterns.
P.S. Really depends on what kind of network.....
I agree on clarification:
Regularities identified from manual trading practice lose reproducibility (credibility) below H4.
I'm also looking for an MTS with fast reactions to work on the minutes. But so far - alas.
P.S. i.e. compensate for ignorance with quick reactions.
I agree for clarification:
The patterns identified from manual trading lose reproducibility (validity) below H4.
I'm also looking for an MTS with fast reaction time to work on minutes. But so far - alas.
I wrote above, you may have missed it or not understood it -
These are all patterns that are visible to the human eye and accessible to human understanding. But there are patterns that cannot be detected and understood by man himself. That is, they can be found only by experimenting with optimizer and indicators, including a neural network. These patterns are the most stable in my opinion.
Pictures are coming :)
But tell me - is this a pattern or a "how"? ..... where 8, 14, 16
EURUSD has been studied. The period from 2005 to the present.
Vertically, it shows the number of zigzag breaks built on TF 1440, 720, 480, 360, 240, 60.
Horizontally - time of day. And if yes, how it can be used? :)
I wrote above, you may have missed it or not understood it -
These are all patterns that are visible to the human eye and available for human understanding. But there are patterns that cannot be detected and understood by man himself. That is, they can be found only by experimenting with optimizer and indicators, including a neural network. These patterns are the most stable in my opinion.
If the EA is not NN then it is usually a rearrangement of the manual TS. The question on this thread was, as I understand it, about patterns that are visible, known and can be applied to MTS.
Therefore I will explain what I said above about H4
- manual TS are effective when analyzing H4-D1, after the analysis we look for/wait for an entry point at M30/H1.
When such a TS is transferred to MTS, then by a natural human reaction it is forced to work on M5-M15 and even M1, and of course it is dumped.
However NN MTC developers do not extract rules from the trained network and do not analyze them, that is why the question about "invisible to the eye" regularities remains open.
I, in particular, do not believe there are any!
Pictures are coming :)
But tell me - is this a pattern or a "how"? ..... where 8, 14, 16
EURUSD has been studied. The period from 2005 to the present.
Vertically, it shows the number of zigzag breaks built on TF 1440, 720, 480, 360, 240, 60.
Horizontally - time of day. And if yes, how it can be used? :)
forbid an Expert Advisor to trade at a non-dark time)))
>>))) advisor to trade at unchildish time - forbidden))
"no question about it..... but HOW to allow it in children's time(?) :))))
There are short waves at NOT baby time, shortening the period of the indicators does not work, except to invert the entry rules.
P.S. Sorry - thought about it and didn't #follow# the answer))
The first column is the sum of the MACD histogram from the signal ( Main[2]<Main[1] and Main[2]<=Main[3] ) counted from 1 to the previous opposite signal.
The others are named. The distribution is done according to the maximum profit (by High) from the point where we entered to the opposite signal.
As a result, distribution of profit interval fraction depending on the sum of histograms, there are small positive res....
This is an old work, so I don't remember some things from it, but if you wish it may be restored...
By the way only Buy was considered...
Glad there are such researchers. It will be nice to communicate, but a bit different, though interesting too, if I understand the idea tested there correctly. Didn't analyse the figures.
The point is that to test the hypothesis put forward MACD is not suitable. Suppose we want to analyse the opening of the Americans at 16:00, we are not interested in the data one second before 16:00 (MACD is not helpful here), what is important is where the movement goes in the interval from 16:00 to 16:02 (upwards move +1, downwards move -1) and to analyse further movements we put -1, if the upwards move was larger than the downwards move during the session from the time 16:00, if vice versa - +1. We get two arrays that we check for coincidences and accept or reject - coincidences + with + and (or) - with - are random or not.
And only then we decide when to enter, where to enter and if it is worth doing :), i.e. we start forming TS.
Short waves are at the baby time, shortening the period of indicators does not give any result, unless we invert the entry rules.
I.e., if I understand correctly, it is possible without any additional indicators - we look at the previous breakdown (Up, Dn) and open, respectively - long or short, just by market order or using stop orders - that was my idea :)
The hardest thing is with the exit :)
and this is probably already offtop )