How can I tell the difference between a FOREX chart and a PRNG? - page 12

[[Deleted]]

Demi:

Yeah....

Listen up:

What you write in your post is an attempt to interpret fractal geometry passing it off as the fruit of your own thoughts. THANK YOU, BUT DON'T.

Well, don't, so don't.

Bye-bye.

[Дмитрий]

Why invent conspiracy theories and prove there is no obvious presence effect etc. when the whole world uses the supply and demand theory for all markets? Well, I've written about this before.

I was taught in probability theory that uncorrelated series are virtually non-existent in nature. To pick two series with correlation coefficient = 0 is a "must do".

[Дмитрий]

AlexEro:

Well, don't, so don't.

Bye.

I won't, but it really is fractal geometry.

[Avals]

AlexEro:

And here's how: on page 8 of this thread

https://forum.mql4.com/ru/53661/page8

ALSU gave definitions, but "forgot" to specify what role autocorrelation of series and correlation between consecutive random variables play there (these are somewhat different things, but that's not what we are talking about now).

So to start with - one should consider that correlation between supposedly random price quotes HAS to be present, and then proceed from that.

Why is it there - in the pricing themes.

Why it should be taken into account - well-ooooo, mate, in probability theory VERY ALL conclusions start with ".... random uncorrelated-values.....".

Does your understanding of pricing suggest that there is autocorrelation between price increments? Exactly correlation, not memory effects in general (non-marking)?

Are these assumptions enough for you and you don't need anything else from pricing? I.e. like in the TA premise - "the price remembers everything" and go on with the mathematics-statistics-cryptography))?

[[Deleted]]

Avals:

does your understanding of pricing suggest that there is autocorrelation between price increments? Exactly correlation, not memory effects in general (non-marking)?

Are these assumptions enough for you and you don't need anything else from pricing? I.e. like in the TA premise - "the price remembers everything" and go on applying mathematics-statistics-cryptography)?

I sense a certain unfriendliness in your questions. I usually stop the discussion immediately. But for you I will make an exception and answer:

1. The correlation between two bars of the M15 timeframe is the autocorrelation of a series which is between fifteen bars of M1. The correlation of the bars of the larger timeframe is the autocorrelation of the bars of the smaller timeframe, its microstructure. Add here the fact that the quotes received by you are already filtered, i.e. they already have the Slutsky Effect. Perhaps, that's why Privalov wanted unfiltered tick quotes so vehemently, and got banned for it (I'm more relaxed about the tick problem).

2. I don't know what "non-marking" is. Tautologies of flawed flawed speculative scholastic mathematical theories have never interested me.

3. not enough. Something else is needed.

4. TA. I can repeat again (see above) what is written in almost any probabilistic inference, and why POSTLY there are no methods for handling highly correlated "random" series in theor-ver (it's just exactly a belief, like in a cult). Professor Orlov (a well-known practitioner of probability theory, author of many articles, journal editor, and author of books) also writes about this, clearly warning about the dangers of applying statistics to economics.

[Alexey Subbotin]

Prival:

There is such a notion asa Wiener process and there is a filter that monitors this process. It is called a Wiener Filter

The technology is simple. Feed the process to be analysed to the filter input, and watch the output. If the filter jingles (jargon) then the analyzed process is not Wiener's, but different from it...then it's the turn of statistical radio engineering..... There are a lot of letters, who is interested, I hope he will follow the links and at least read them.

Z.Y. We used to solve such problems with cadets at practical classes in radiolocation. The standard problem is to distinguish a noise at the radar input from a mixture of noise + signal...

Sergey, between tasks of random/unknown signal detection in radars and in quotes, with all letters being almost identical, there is an essential difference: for radars the calculation delay of an order of a pulse length itself is absolutely uncritical (not to mention that Wiener Filter ideally needs infinite observation time and strict stationarity of the system), but for trading it is almost a disaster. Therefore, the second problem is an order of magnitude more difficult, and not every radio engineering cadet will be able to cope with it.

[Alexey Subbotin]

AlexEro:

... "forgot" to specify what role autocorrelation of series and correlation between consecutive random variables plays there...

I didn't "forget", I just got tired of typing))

[[Deleted]]

The author didn't seem to ask to approach the issue from a pricing point of view. What does this have to do with ... There are 2 ranges of gpsh and kotirs (no specific sampling depth), some have no internal fluctuations, others do. May be based on this. + Presence of thick tails. Thick tails after all indicate that in some places the narrowing dispersion may last longer/shorter in number of samples than in the normal gpsh distribution, hence the discharge in the cotier sometimes occurs by shots/presence of thick tails, in consequence of these prolonged narrowings of the dispersion. In which row is it more common to see the "body" of the candle reduced by, say, 5-6 times in a row? Roughly speaking, the "average" discharging in the gpsc is happened in 3-4 candles, while quotes may have more. But in fractalization, the rarity dependence of the "pattern" of dispersion narrowing on its appearance and triggering (breakdown of the dispersion narrowing) will have an exponential dependence on the set of statistics of such patterns. And for a quotient on different fractal structure, the "bias" or prolonged narrowing of the variance, may be shifted relative to the fractal levels. We obtain in kotirs in some places (at superposition of increasing fractal structures by layers) compaction by dispersion and in some places a more rarefied state. In gpsc such densifications and their occurrence will also be pseudo-random and to achieve a periodicity at least of some kind (naturally, not constant in time) one will have to previously reduce to it - by MM or, if we speak of CP, by coefficients, to such cases. Somewhere to hold the variance narrowing, somewhere to accelerate it.

[[Deleted]]

alsu:

I didn't "forget", I just got tired of typing))

Well, clearly, I hadn't really forgotten, I'd just simplified as much as I could, to the point of hilarity. That's why I put it in quotes.

[Дмитрий]

AlexEro:

4. TA. I can repeat again (see above) what is written in almost any probabilistic output, and why POSTLY there are no methods for handling strongly correlated "random" series in theor-ver (it is just exactly a cult-like belief). Professor Orlov (a well-known practitioner of probability theory, author of many articles, journal editor, and author of books) also writes about this, clearly warning about the dangers of applying statistics to economics.

I don't, of course, dare to speak for all theorists and matstats, but in correlation and regressionanalysis the problem of multicollinearity of factors is being dealt with, and quite successfully. Why "danger"? Not danger, but checking and transforming.