Volumes, volatility and Hearst index - page 29

New comment
 
faa1947:

In posts earlier, in another thread, I tried to prove that quoting one timeframe by resonance frequencies is one thing and quoting another timeframe is quite another.

You still haven't figured out in what units that program measures periods?
 
joo:
.....

It's a bit rambling, but I don't have another definition, but the principles I adhere to. In my opinion, Paterns, as I have defined them, cannot be investigated by correlation and other statistical methods, and in general it is impossible to derive analytically formulas of characteristic Paterns, because they continuously appear and disappear, flowing into each other, and, as I said, in each TF a different Paternon, which do not depend on each other. Different combinations of PATTERNs in different TFs give different but moment-specific Investigative PATTERNs. It is like a kaleidoscope or snowflake pattern, although the patterns are infinitely many, but exclude the appearance of "impossible" patterns. That is, there is some set other than the set of Patterns.

It follows from all this that it is necessary to analyse Paterns simultaneously on different TFs. It is not the same as the Three Screens Method, which only gives discrete signals. The Method of Flowing Patterns (well, there's finally a name for my method) gives continuous (with the smallest possible discretization that is possible on the BP under study) signals in time.

....

Oh, and one more thing. It is similar to human figurative thinking. Each individual image does not play any significant role in thinking, besides, images of the same concepts are different for all people. However, it is the totality of all images or groups of images that allows thinking, generation of new images, and obtaining new information. One can say that there is a certain minimal level of images, at which generation of new images adequate to reality is possible. This is how new knowledge/discoveries appear in mankind, only on the basis of already existing knowledge/images.
 
Candid:
You still haven't figured out in what units that programme measures periods?

I even gave you a trailer. I don't need to figure it out - minutes. There are no frequencies in this programme as they are not used in forex.
 
Farnsworth:
to HideYourRichessto faa1947

But if you go deeper into FA, after poking around with all sorts of correlation integrals, information dimensions, entropies, singularities, etc. (that's me, as you noticed - "crushing" intellect :o)))) + some optimism, then one can come to one very important conclusion. Quoting is an extremely complex process, but not random (!!!!). The process is not noisy, it is as we see it - but it is very complex(!!!)

If you reinvent your own bikes, then yes.

There is a fairly widely accepted market model - trend + wave (maybe) + noise. It's called ARPSS (1976!). The model works on non-stationary quotients, but it's not universal. So there are sections of the quotidian where no model can be identified. But on those sections where it is possible to identify, then predictions can be made. In my opinion, the right way is to try to extend this model to areas where it does not work. This too was done in 1984 and is called GARCH with many modifications thereafter.

In fact ARPSS together with GARCH also looks for patterns, as they were looked for in the past ("head and shoulders"), now it is looked for with TC (it is often impossible to describe in words what TC looks for). But the point is the same - to shift the probability in favour of winning, and optimism very soon turns into pessimism

 

It is truly amazing the persistence with which many try to interpret similarity solely as geometric similarity. Despite the perfectly specific example of similarity given, I am referring to the statistical ratio of High-Low and |Close-Open|. That is the real similarity. By the way, Yuri, your ZZ example might be even better, but it seems to be from a personal account, so I don't bring it here.

Another wonderful example of incomprehensible stubbornness is the requirement to have ideal fractals in real rows.

By the way, perhaps the patterns are just segments of "almost undisturbed" fractal development. Which, of course, cannot last long.

I also do not think it is correct to compare minutes with days. In euro minutes I have almost 4 million bars for example. On the days I have 3316. I'm just sure I can find quite a few very similar spots on the minute history.

Even the recent off-topic with the pullback distribution is actually not an off-topic at all, but an example of a real similarity. The price passed 100 pips, rolled back 23%, then passed another 50 (150 in total) and rolled back 23% again - isn't this a similarity?

I suggest that arguments like "here real trees are different from fractal trees, therefore we do not need the science of fractals" should not be considered any more.


Another question is that it is not very clear how to make money out of such a similarity. So it is suggested to think about it, to look for maybe more suitable characteristics.

 
faa1947:

I even gave you a trailer. I don't need to figure it out - minutes. There are no frequencies in this programme as they are not applied in forex.
No they don't, they are measured in pcs of bars. I wrote that in that thread, but you seem to have just missed that post.
 
faa1947:

There is a fairly widely accepted market model - trend + wave (maybe) + noise. It is called ARPSS (1976!). The model works on non-stationary quotients, but it is not universal. So there are sections of the quotidian where no model can be identified. But on those sections where it is possible to identify, then predictions can be made. In my opinion, the right way is to try to extend this model to areas where it does not work. This too was done in 1984 and is called GARCH with many modifications thereafter.

Well you can't isolate noise on a quote - apparently you don't understand that because you haven't tried it. And no ARPSS will help you on quotes and you will never find these plots. We'd be such clever millionaires walking around here in droves - the island and the castles would not be enough for all. :о) To isolate noise means to find an adequate model.

In fact ARPSS together with GARCH also looks for patterns, as they were looked for in the past ("head and shoulders"), now it is looked for with TC (it is often impossible to describe in words what TC looks for). But the point is the same - shift the probability in favour of winning, and optimism very soon turns into pessimism

Oh, no! That's no comment at all.

PS: on non-stationary series and AR works, only there are significant limitations for AR, ARPSS, GARCH and similar. These models do not work and for them to work I need some optimism :o) By the way, I use some of the models listed above as models for random structures. That's not the whole question:

In my opinion, the right way is to try to extend this model in areas where it doesn't work.

This is essentially yours:

If inventing your own bikes, then yes.

The question is to find such a phase space on which these models start to work. And that's all there is to it. And reinventing bicycles a bit, how can you do without it :o)

 
FreeLance:

I am of course "wildly sorry", but explain to me "inexperienced" the reasons for the Slutsky-Yule paradox/effect.

Otherwise I cannot understand the addition of random variables.

Especially your reasoning on the subject of self-similarities.

Forgot to answer. I think the Slutsky-Yule effect is explained very simply. Let's look sequentially: at moment t1 - the sliding window (w) fixes a certain length of time series, which limits the sample under study. At moment t2 the same window shifts by one count, but how does the "filling" of the window change?


Not much :o). In a new sample of length w, only one sample is shifted by one count. That is, when you move one step, the whole sample (w-1) is retained except for one value. Take a random sample of length w, shift and you get two loosely distinguishable samples. All statistical characteristics will differ insignificantly. I.e. there will be correlations and pseudo-cycles that don't really exist. You can try it on a completely random series, you will get this effect in full glory.


PS: In this regard, I strongly recommend not to use MA. It's fooled by randomness :o)))

Gentlemen Scientists!

I am not a scientist.

 
Farnsworth:

You cannot isolate noise on quotes - apparently you do not understand it, because you have not tried. And no ARPSS will help you on quotes and you will never find these plots. If only there were more of us millionaire smart guys walking around here - the island and the castles would not be enough for all. :о) Toisolate noise means to find an adequate model.

If you use ARPSS, I don't get it. The premise of ARPSS is: trend + wave + noise.

PS: AR also works on non-stationary series, but there are significant limitations for AR, ARPSS, GARCH and the like. These models do not work, but for them to work I need some optimism

Or qualification, qualification first.

The question is to find the phase space where these models start working.

I've been thinking a lot about this, but nothing. maybe you could share your results?

 
Candid:
No they are not, they are measured in bar pieces. I wrote that in that thread, but you seem to have missed it.

Indeed.
1...22232425262728293031323334353637
New comment