First sacred cow: "If the trend started, it will continue" - page 43

 

They can both be right and wrong, as they both use non-universal definitions of a trend.

 

Yes ss, from page 36 textbook phrases, how about the opinion that there is no trend

Some say no fibo levels, fibo and forex are mystical.

I think there is no trend, there is a price movement that many (most) try to classify

they are trying to define, find sources, and identify the effect, describe, calculate and fit into a system

there is no systematic market (etc.), but price movements and processes that occur during price movements are not systematic.

otherwise people would not have discussed for 101 times what has already been discussed and mathematicians would not work ... and programmers would have written everything they could write in 2 years.

I know many say the system exists, but it's unstable, etc., but this is demagoguery.

Not one forecast wave level, etc. etc. no one can give an answer to the question TREND TURNS OUT

one cannot predict, much less predict what IS NOTHING

and the essence of prediction comes down to someone flipping a coin, someone to token from a slot machine, and someone staring at the bottom of a cup of coffee - nothing changes.

The price movement is always up or down. like roulette black or red - all further classifications sideways, etc. are groundless - open M1 - up and down

 
baltik >>:

Да сс-с, с 36 страницы учебниковые фразы, а как вам мнение что тренда не существует

Кто -то говорит нет уровней фибо, фибы и форекс это мистика.

я считаю что нет тренда, есть движение цены которые многие (большинство) пытаются класссифицировать

дать определение найти источники и выявить следствие описать, обсчитать и запихнуть в СИСТЕМУ

НЕВЫДЕТ - так как системен рынок (и тп.), а движение цены и процессы которые происходят когда происходит движение цены не имееют системности

ИНАЧЕ бы люди не обсуждали по 101 разу то что уже обсуждали и математики не работали бы... а програмисты бы за 2 года написали бы все что можно написать

я знаю что многие скажуть что система существует- но она изменчива и т.д. но это уже демагогия

Не один прогноз волновой машечный уровневый и т.д.и.тп. НИКТО не может дать ответ на вопрос РАЗВЕРНУЛСЯ ТРЕНД

нельзя предсказать а тем более прогнозировать то чего НЕТ

и суть прогноза сводится к тому кто что подбрасывает монетку кто-то жетон от игрового автомата, а кто на дно чашк с кофе пялится - ни чего не меняется.

Движение цены всегда вверх или вниз. как рулетка черное или красное - все дальнейшие классификации вбок и т.д. безпочвенны - откройте М1 -вверх и вниз

Introduce a simplification model and convert this haphazardness into a trend system :) It is possible. Another thing is that they all turn out differently, so no uniform terminology can be introduced. But this is the problem that the author of the topic wants to solve. Everything is very simple.

 

For a general start: Pastukhov did not say anything about trends, there is no such word in his text at all.


baltik >>:

Да сс-с, с 36 страницы учебниковые фразы, а как вам мнение что тренда не существует

Кто -то говорит нет уровней фибо, фибы и форекс это мистика.

я считаю что нет тренда, есть движение цены которые многие (большинство) пытаются класссифицировать

The idea is not original at all, but difficult for the average person to grasp. Peter Phillips, not the last man in the world of econometrics and time series analysis, says almost the same thing (the lecture "Exploring the Mysteries of Trends and Bubbles", unfortunately, I do not have the link).

A trend, for any time series, can only be constructed if there is a theory that explains it. If there is no theory, then any trend is self-defeating. For financial markets, there is no such theory, except for the effect of inflation, which provides a weak positive drift for the stock market. No theory, no trend.

Anyone who doubts, I invite you to study trends, and at the same time fibs, waves, etc., on a random walk chart.

 
timbo писал(а) >>

For a general start: Pastukhov did not say anything about trends, there is no such word in his text at all.

The idea is not original at all, but difficult for the average person to grasp. Peter Phillips, not the last man in the world of econometrics and time series analysis, says almost the same thing (the lecture "Exploring the Mysteries of Trends and Bubbles", unfortunately, I do not have the link).

A trend, for any time series, can only be constructed if there is a theory that explains it. If there is no theory, then any trend is self-defeating. For financial markets, there is no such theory, except for the effect of inflation, which provides a weak positive drift for the stock market. No theory, no trend.

Anyone who doubts, I invite you to study trends, and at the same time fibs, waves, etc., on a random walk chart.

H-volatility is an estimate of the instrument's trendiness/flatness, and the trading methods proposed by Pastukhov and Shiryaev based on this estimate with Kagi or Renko are elementary trend and flat system. If the instrument is trending R(H)>2, we buy after the upside brick and sell after the downside brick. If a flat R(H)<2, then it is vice versa

A picture from Shiryaev's lecture

Perhaps they didn't use the word itself, but in fact H-volatility contains the definition of trendiness and its estimation

P.S. The lecture with pictures is available at https://www.mql5.com/go?link=http://club.investo.ru/viewtopic.php?f=14&t=127600&start=15

 
Avals >>:

Возможно само слово они не применяли, но по сути H-волатильность содержит в себе определение трендовости и ее оценку

P.S. Лекция с картинками выложена в https://www.mql5.com/go?link=http://club.investo.ru/viewtopic.php?f=14&t=127600&start=15

"Basically" is your attempt to fit the problem to the answer. There is no word trend because it has nothing to do with a trend, because there are no trends.

The paper is theoretically oriented, assuming that... . Realistically, there are no continuous series with a swing greater than 2H. This is exactly what Pastukhov did not say. 2H is a random walk, and the distribution of increments can be anything, not just normal as Pastukhov's, as long as the third moment is equal to zero, which is not a problem. More than 2H would be a sub-martingale, but there are no sub-martingales in financial markets. Less than 2H is mean-reverting. That is, the market is an alternation of random wandering with periods of pseudo-stationarity, no trends. H-volatility can be a criterion of separation of these periods, but unfortunately only after the fact, when the train has already left.

P.S. The full text of Pastukhov's dissertation is on the spider.

P.P.S. The already mentioned Peter Phillips moves the idea of soft explosions as a possibility of sub-martingales in financial markets leading to financial bubbles. But that's another story altogether.

 
timbo писал(а) >>

"Basically" is your attempt to fit the problem to the answer. There is no word trend because it has nothing to do with a trend, because there are no trends.

This is my opinion of such a little formalized concept as a trend. And the opinion that Pastukhov, too, analyzed trends was formed not only by me and has not changed for five years since the publication and discussion of his dissertation. This is why I do not adjust to the current discussion of the trend.

timbo wrote >>.

The work has a theoretical orientation. . Realistically, there are no continuous rows with a swing greater than 2H. This is exactly what Pastukhov did not say. 2H is a random walk, and the distribution of increments can be anything, not just normal as Pastukhov's, as long as the third moment is equal to zero, which is not a problem. More than 2H would be a sub-martingale, but there are no sub-martingales in financial markets. Less than 2H is mean-reverting. That is, the market is an alternation of random wandering with periods of pseudo-stationarity, no trends. H-volatility can be a criterion for separating these periods, but unfortunately only after the fact, when the train has already left.

No, he was just giving examples of real series, where by his calculations h-volatility is consistently different from two. Which I doubt, and apparently so do you. For example they calculated 1.58 for RAO UES.

P.S. Actually h-volatility is similar to Hurst index and fractal dimension, though of course it is calculated differently. The fact that persistence is a recognized measure of trendiness is kind of obvious. For example Naiman writes http://capital-times.com.ua/dobavit-novost/view-28.html

"Calculation of the Hearst exponent for the purpose of identifying trendiness (persistence)". He is a PhD in economics.

 
Avals >>:

нет, он как раз приводил примеры реальных рядов, где по его рассчетам h-волатильность устойчиво отличается от двойки. В чем я сомневаюсь, как и вы видимо. Для РАО ЕЭС они например подсчитали 1,58

I say there's no such thing as more than 2. Anything less is fine. In other words, there is no trend according to Pastukhov either.

However, one must keep in mind that Pastukhov talks about continuous series, while in financial markets we always deal only with discrete series. Two problems follow from this:

1. In reality, the multiple of the swing depends on the choice of H; in theory, the size of the swing expressed in H is always one.

2. The dimensionality of the increments greatly affects the multiplicity of the swing size. If H is comparable in size to the size of the increments, you may see 3H or more, but the "trend" strategy will still not work.

 
timbo писал(а) >>

I say there's no such thing as more than 2. Anything less is fine. In other words, there is no trend according to Pastukhov either.

However, one must keep in mind that Pastukhov talks about continuous series, while in financial markets we always deal only with discrete series. Two problems follow from this:

1. In reality, the multiple of the swing depends on the choice of H; in theory, the size of the swing expressed in H is always one.

2. The dimensionality of the increments greatly affects the multiplicity of the swing size. If H is comparable in size to the size of the increments, you can see 3H or more, but the "trend" strategy still won't work.

I agree :)

I'm not suggesting we use their methods. Just noted as part of the discussion that persistence and h-volatility are indicators of trendiness. As far as I understand you do not agree with the latter and this is your right, as Pastukhov obviously does not pronounce the word "trend" and "trendiness".

 

It is useful to think of a time series in general terms as a mixture of four components:
1. trend or long-term movement;
2. more or less regular fluctuations relative to the trend;
3. seasonal component;
4. residual or unsystematic random effect, white noise.

The series is conveniently represented as a sum of these four components, and one of the purposes of the analysis
is to decompose the series into its components for separate study.

We will now be interested in 1-2pc.

Also, we need to clarify the concept of "stationarity",

the time series stationarity (quotes of currency pairs),

in a broad sense, stationarity means that the average X = 0 and

this property should not depend on the time point, as applied to

let's say for EURUSD, whose quotes we examine over a year, then

the average X should not depend on the current month, week, day, etc.

i.e. the average in June should be equal to the average in October,

and in December...obviously this is not the case and therefore the

EURUSD is not a stationary time series in the broad sense.

But it is possible that on some time interval, say a week,

with some arbitrary timeframe (M5,M15,...), stationarity will be observed

with an average value X-B=0, where B is an arbitrary constant, say, equal to 1.3656;

If instead of B we substitute linear regression equation X-(Ax-B)=0,

If an angle coefficient A=0, then we are dealing with a flat,

If A<>0, we have a trend.

In this case the problem of distinguishing a flat from a trend reduces to

In such a case the problem of distinguishing a flat and a trend is reduced to testing of the hypothesis of A being equal to zero (non-zero).