a trading strategy based on Elliott Wave Theory - page 209
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 above arguments, for me (I emphasize that for me) do not prove the absence of trends as such, but only convince of the limited knowledge of Nature. (a bit philosophical :o) Any modification of autocorrelation estimates only a conditional correlation between samples, and this figure by itself does not, of course, yet indicate presence or absence of a trend.
You're wrong!
A trend can be defined as the excess of the number of co-directional price jumps, over multi-directional jumps. Otherwise, if you subtract the sum of all counter-directional price jumps from the sum of co-directional jumps and take the obtained difference to the total sum of all jumps, then the obtained number will, by definition, lie in the range from -1 to 1. If the value is close to 1, then we can talk about a predominantly directed price movement. If it is close to -1, it will indicate the counter directional nature of the series. If it is equal to zero, the process is of random nature.
Sergey, does such definition of a trend remind you of anything? That's right! This is how the autocorrelation coefficient is defined :
r=SUM{(Open[i]-Open[i-1])*(Open[i+1]-Open[i])}/SUM{(Open[i]-Open[i-1])*(Open[i]-Open[i-1])}, or
r=SUM{sign((Open[i]-Open[i-1])*(Open[i+1]-Open[i]))}/N, where N is the total number of jumps, sign is the sign number. Or by regrouping the terms under the sign of the sum in the numerator:
r=(SUM{n+}-SUM{n-})/N, where n+ is the number of co-directional jumps, n- the number of counter-directional price jumps.
Thus, the autocorrelation coefficient, precisely says whether a deterministic trend is present or not. The last point is very important because a trend can be stochastic. And there should be no way of detecting it in nature. Otherwise, the whole slender building of the picture of the world will crumble:-) I'm talking about a violation of causality. We are not fantastists and sick people who would try to violate any laws of nature...
Yes!
It is too early to derive a criterion for detecting a deterministic trend. We need to build a holistic and, if possible, internally non-contradictory picture of price formation, only then will the ways to build an optimal predictive model become clear.
For example:
1. The market moves in such a way as to provide maximum liquidity. I.e. the price goes to the place where the maximum number of participants is ready to trade. For example SL - also trading and hiking for stops is one of the main features of the market to provide maximum liquidity. The main objective of the market is to provide the maximum number of willing traders.
2. In zero sum markets/levels the price moves so as to offset the profits of some with the losses of others. A rough example: the profit of breakers in case of their good luck is made up of the losses of breakers. Therefore, at times of strong moves there are usually two opposing trading ideas (although they can combine a number of others). Price is driven by a conflict of methods/trade ideas. In the long run, these methods counterbalance each other.
North Wind If you haven't figured it out or figured it out yourself, why defend the method so much? :)))
Hasty conclusions again.
Keep in mind that I have not said anywhere that I do not understand exactly what they did and how they did it, i.e. what mathematical methods were used. I only said that I did not understand the interpretation of the results, which is something else entirely.
I couldn't agree more. It's still a microscopic approach. Even if you integrate the whole process inside a candle, forming candles by average position holding time or whatever, by doing so you are simply moving to another fractal level. From my point of view, this approach is good for model evaluations but not for dynamic market analysis. I see two disadvantages in it. 1. The process stochastics are hidden inside the candle and therefore can't be a source of information. Whereas actual measurements of those macro variables that can describe market conditions should dynamically rely on that stochasticity. 2. By shaping the candlesticks in this way you are imposing a certain period on the market. But you stated yourself that the spectrum has no stationary components. So such candlesticks will not light up anything and the market will not like it. :-))
Yes, this is a microscopic approach. But maybe together we organise a marginal shift to a macroscopic view of the market...
On the first point: I haven't said anything about a decision-making mechanism for market entry. Of course, to make a decision to open a position I analyze each price TIC within the current synthetic bar. So, the "Stochastic Process is not hidden inside a candle".
On the second point I agree.
Neutron
It's too early to derive a criterion for detecting a deterministic trend. We need to build a coherent and, if possible, internally consistent picture of price formation, only then will it become clear how to build an optimal predictive model.
For example:
1. The market moves in such a way as to provide maximum liquidity. I.e. the price goes to the place where the maximum number of participants is ready to trade. For example SL - also trading and hiking for stops is one of the main features of the market to provide maximum liquidity. The main objective of the market is to provide the maximum number of willing traders.
2. In zero sum markets/levels the price moves so as to compensate for the profits of some with the losses of others. A rough example: the profit of breakers in case of their good luck is made up of the losses of breakers. Therefore, at times of strong moves there are usually two opposing trading ideas (although they can combine a number of others). Price is driven by a conflict of methods/trade ideas. In the long run, these methods balance each other out.
As correctly noted by grans, the initial function can be decomposed in different ways. The question of which method is better is one of the most significant problems. In particular, the speed and accuracy of the resulting solution directly depends on it.
I don't know the best way. The variant that I propose is covered in scientific papers on the subject.
...
Thanks for the link. And the topic is interesting.
I don't understand why people are so weird there. The coin thread's been drowned in flubber. Why?
Seems like the topic doesn't interest many people, but they just want to scratch their tongues.
...
I've been thinking for a long time about how to phrase the answer more precisely, and I think I've come up with this. :)
For example here, the forum as a means of technical support program, it imposes features on the contingent. Just as ascetic forum engine also contributes to the same. There's a DT forum where 95% of visitors are guessing "Where do you think the price will go? When they get bored guessing, they start looking around and get annoyed when they don't understand something. As I've noticed, this topic has also survived a couple or three such forays.
North Wind you seem not long in this thread and therefore allow myself to give you advice, Alex Niroba here most champion of this kind of conversation that you are now leading with him. "He's proved everything to himself" don't prove anything to him, don't, you're littering the thread unnecessarily, it's happened to all of us more than once... This is one of those cases where it's better to remain silent.
North Wind you seem to be not long in this thread and that's why I take the liberty to give you some advice, Alex Niroba here is the champion of this kind of conversations that you are having with him now. "He's proved it all to himself" don't prove anything to him or you're littering the thread, it's happened to all of us more than once... This is one of those cases where it is better to remain silent.
OK. Fair enough said, I'll bear that in mind.
As grans rightly pointed out the original function can be decomposed in different ways. The question of which method is better is one of the most important problems. In particular, the speed and accuracy of the solution depends directly on it.
I don't know what the best way is. The variant that I propose is covered in scientific papers on the subject.
The requirements for the model: it must logically derive the specific entry levels, goals, sl and / or position rules. The multifactor model you cited is described elsewhere, but it is of only theoretical interest, because it does not have the above conclusions necessary for trading. And to put it simply, there are so many factors and they change so often that it is impossible to predict anything (in the right sense) with this model.
The model that I tried to describe is based on sentiment and trying to find discrete points of time, when there are few basic factors and others become noise in this model. Here is another rough example: when a breakdown is traded, the main trade ideas are related to it and to the counterride, while trades executed due to other considerations can be reasonably considered as noise with the MO=0. This model allows to numerically estimate the zones in which this or that group will drive the price in a certain direction and find signs of timely connection to the necessary group and numerically define the goals and the system sl.
...Приведенные аргументы, для меня (подчеркиваю, что именно для меня) совсем не доказывают отсутствие трендов как таковых, а лишь убеждают в ограниченности знаний о Природе. (немного философски получилось :о)
Любая модификация автокорреляции оценивает лишь условную взаимосвязь между отсчетами, а эта цифра сама по себе, конечно, еще не говорит о наличии или отсутствии тренда.
You're wrong!
A trend can be defined as the excess of the number of co-directional price jumps over the counter-directional ones. Otherwise, if you subtract the sum of all counter-directional price jumps from the sum of co-directional jumps and take the resulting difference to the total sum of all jumps, the resulting number will, by definition, lie in the range from -1 to 1. If the value is close to 1, then we can talk about a predominantly directional price movement. If it is close to -1, it will indicate the counter directional nature of the series. If it is equal to zero, the process is of random nature.
Sergey, does such definition of a trend remind you of anything? That's right! This is how the autocorrelation coefficient is defined:
r=SUM{(Open[i]-Open[i-1])*(Open[i+1]-Open[i])}/SUM{(Open[i]-Open[i-1])}, or
r=SUM{sign((Open[i]-Open[i-1])*(Open[i+1]-Open[i]))}/N, where N is the total number of jumps, sign is the number sign. Or by regrouping the terms under the sign of the sum in the numerator:
r=(SUM{n+}-SUM{n-})/N, where n+ is the number of co-directional jumps, n- the number of counter-directional price jumps.
Thus, the autocorrelation coefficient, precisely says whether a deterministic trend is present or not. The last point is very important because a trend can be stochastic. And there should be no way of detecting it in nature. Otherwise, the whole slender building of the picture of the world will crumble:-) I'm talking about a violation of causality. We are not fantastists or sick people who would try to violate any of the laws of nature...
Sergey, you are surprisingly flexible (or I don't understand anything)! Before that you wrote, verbatim, "It will definitely not converge to a trend...", and previously you convinced yourself of that, and with this post you refuted the earlier statement and expanded on mine. It may well be that I haven't yet fully understood where it is and where it isn't.
In general I agree, if one chooses just such a definition of trend as "counting multidirectional deviations" and tweaks the calculation of autocorrelation to it, i.e. replacing y(i) and y(i+1) by their respective differences one can get to its estimate.
My understanding of autocorrelation (hereinafter referred to as AC), mostly derived from DSP, is slightly different. AC, revealing the internal structure of the signal, gives only a quantitative and moreover relative value of the interrelation of samples, but in general it is not small. And AK alone will not be enough, still an additional criterion will be needed, for example: how many transitions there must be, how many values greater than 0.5 must be, to recognize the presence of a trend.