a trading strategy based on Elliott Wave Theory - page 203
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
Work on it?
A lyrical digression.
One of our famous volleyball players was famous for his ability to guess almost unmistakably the direction of the opponent's shot and put a block. Of course, you can't react when the ball is flying. When he retired from sport, in one of his interviews he said that he guessed the direction of the future blow quite simply: before striking the opponent turned his head to the side opposite to that of the kick - it was easier to kick in this way. Later, when it became known to everyone, it stopped working - volleyball players eliminated this technical defect in the attack.
Actually, I do not like the description of the market in timeframes as it is used nowadays. It's more important to answer the question not "how much time has passed" but "has price passed a significant number of pips". In many forums, such as MQL4: The Self-learning Expert Advisor, there is a discussion about a very interesting fighter. Yes, he's far from perfect, and he's also losing money. But his conception of market data presentation is very interesting. He builds "patterns". A pattern is an integer number, for example a 10-bit number. And each pattern has a price step - delta - associated with it. The way it works is this:
1) At the initial moment, we reset all bits of the template to 0 for example.
2) Fix the current value of the price.
3) Wait for the price to change either up or down by at least delta.
4) Shift the pattern to the right by 1 bit.
5) If price changes by + delta, write 1 to the high bit, if - delta, write 0.
6) Return to (2).
We get a set of pattern numbers describing history for different delta. It seems to me better to apply not binary, but ternary equilibrium logic here. I.e. at the initial moment the pattern is reset to 0. When price changes by +delta - we write 1 in the high position. On -delta - we write -1. It did not change within the time interval - 0. If we are speaking about compression of market data, I find the method very good. With high values of delta even a very old history is saved. With low values of delta even the newer history is saved. And only essential information is saved - in which direction the price moved a significant number of points. This is an example of a truly unconventional approach to analysis. Maybe we should think in this direction as well?
While TA is technical analysis, FA is fundamental.
To be honest, I still don't understand how you are going to use this criterion for trend identification.
I think you should first refer to Neutron on this issue. In his post read:
I took the liberty of simplifying this sentence to make its meaning clearer. In particular, it implies that in mathematics there are notions of "stochastic trend" and "deterministic trend". Since we have given up the role of housewives, we should first of all refer to science and ask Sergey, in addition to what he has already written, to present mathematical definitions of these two notions. So, "stochastic trend" and "deterministic trend" in the studio ! :-)))
We may not be comfortable with these definitions, but it's a good, proper start in the search.
And one more thing. There is a danger of confusing the definition (in the sense of defining the concept) of a trend and identifying it. I think these two things need to be separated from the start. A definition is given so that there is something to work with, so that the phenomenon in question can be captured. Without it, we will get a mixture of different phenomena under the microscope, and we will not be able to identify something that is essential and characteristic of what we are investigating.
Identification is practical identification. And it can be done post factum or at an early stage. We are all interested in early identification of a trend. But it is possible only if some regularities are found in the process of studying the phenomenon as such, i.e. in the process of researching history of all trends which can be identified using the definition of a trend. This is so, a lyrical digression about the methodology of the scientific approach.
As I wrote earlier, I used autocorrelation to determine a reliable forecast sample according to the rule: the slower the autocorrelation series converges, the more reliable the sample for the forecast (strong relationships are present). This is where the difficulty arises in defining and quantifying the convergence itself.
Of course, let's do some work:
(1) Let us start, perhaps, by defining a deterministic trend. What conditions must be satisfied for such trends? And what is a deterministic trend?
(2) For the CCS, everything is clear. There are:
Statistics: R(n) - number of transitions through zero
Criterion: R1(alpha, n)<R(n)<R2(alpha, n), in the end determines presence or absence of trend (For Yuri - if number of transitions through zero falls into this range, then series is considered a random one. R1 and R2 are calculated for each reference n)
What would be the statistics and criterion for the autocorrelation method?
PS: As far as science is concerned, I have yet to come across a coherent definition of neither deterministic trend nor stochastic trend. I am even inclined to argue that these concepts are strongly tied to a specific application area
The signal processing technology is called kagy\renko among traders, and delta-modulation among admirers of the signal processing theory. The Expert Advisor's own technology - search for patterns, in the general case can give about 2-3% stat advantage.
In general, understanding (I stress understanding, not a concept) implies the presence of some kind of regularities in a series of numbers, the presence of a "steady" process. The concept of a trend, as far as I know, is not articulated anywhere.
To be honest, I still don't understand how you are going to use this criterion to identify the trend.
Yurixx, it's very simple. From the current datum, in increments of +1 or greater, for example, samples are taken and statistics and criteria are analysed. There may be several variants: a trend may be found after the first iteration, and we should find the sample interval, in which the trend disappears or the trend is not found. In the second case one should not get frustrated, but continue going through the history until it is discovered and detected.
Northwind.
In general, understanding (I stress understanding, not a concept) implies the presence of some kind of pattern in a series of numbers, the presence of a "steady" process. The concept of a trend, as far as I know, is not articulated anywhere.
That's what I'm saying, it's all philosophy, until we get to an applied understanding of the tasks at hand.
Yeah, that's about it for now.
On the subject, in what was written before, there are clear features of at least two "not for housewives" topics. Quality control of the process and the challenge of the breakdown. Good luck.
Yes, so far so good.
On the subject, there are at least two "not for housewives" themes in what was written earlier. Quality control of the process and the challenge of the disassembly. Good luck.
I don't get it. What has quality control (if we are to add ISO 9000 to complete happiness) and the challenge of decoupling have to do with it? Explain, please.
OK. A little later, I'll formulate it.