FR H-Volatility - page 33
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
to Neutron
Serega, I'd like to borrow a random series from you for an experiment, the most random series you have ever generated. :о) Surely I have some lying around. Preferably, the number of samples should be no less than 50000.
PS: Seryoga, I found an equally cool phenomenon. And if it statistically will be confirmed, I'll create a theme like "market memory". There will be something to discuss. :о)
I say we swap without looking! - You put the idea out there without any stat validation and I'll put out a row.
I'm starting:-) Here's the "most" random of the fat-tailed ones. Hook on!
I say we swap without looking! - You put the idea out there without any stat validation and I'll put out a row.
I'm starting:-) Here's the "most" random of the fat-tailed ones. Cling!
Seryoga thank you! Definitely will share the idea, just give time to check everything, prepare, and present the material. But it's not a full idea yet, just a completely random phenomenon. Well, not completely accidentally found, but nevertheless. There is also an idea about it but it is too far away from practical realization. Or maybe it is just nonsense. But this nonsense, it seems to me, is associated with the memory of the market, and lies, so to speak - above what can be found by autocorrelation and other clever mathematical things, and therefore, just do not detect. "Theoretically", this phenomenon should not occur on a random series. We'll see, but if it does, I'll make a separate thread, simply because it deserves separate attention.
Yeah, I see - if you look at a portrait of the Giaconda up close, you see a bunch of bumps and dots, but if you take a step away, a miracle happens - you get a coherent and beautiful picture! It must have revealed to you where and to what extent you need to step back?
:о)))) Uh-huh. Very similar, how much to backtrack on that is the mystery, although you can get a statistical estimate. Roughly speaking, there seem to be structures that carry a very strong memory, or maybe there aren't any at all, and I'm messing with your head.
PS: and the row you have exactly at random? How did you generate it? A cursory analysis showed very strong correlations and in places a very high, almost 0.95 probability of finding a trend (confirmed), the fractal analysis also suggests the presence of "memory". At least literally five times (not enough of course) I calculated a probable price level with my model for random "current readings" - it worked. By the way, my model doesn't work properly on a Wiener process. But it takes a few days to generate it, even for my wheelbarrow. Once generated file was accidentally lost during transfer. :o( Ok, we'll figure it out.Random, random!
The measure of randomness is determined by the relationship between the increments. The autocorrelation coefficient (AC), which shows the relationship between adjacent increments, in this case does not exceed a few units per 10^-2, which is within the acceptable range of statistical variation for a series of this length. I generated the series by third-degree integration (to form "thick tails") of normally distributed SV with zero MO and sko=1. Below I have posted another implementation of the Wiener process (5*10^5 counts, ASK=2*10^-3). Check it out!
By the way, I don't need to explain to you what "memory" presence detected by this or that processing algorithm in such series means - it is a level of reliability of algorithm, not series - as you want it!
Accidental, accidental!
Accidental, accidental!
The measure of randomness is determined by the relationship between gradients. The autocorrelation coefficient (AC), which shows the relationship between adjacent gradients, in this case does not exceed a few units per 10^-2, which is within the acceptable range of statistical variation for a series of such length. I generated the series by third degree integration (to form "thick tails") of normally distributed SV with zero MO and sko=1. Below I have posted another implementation of the Wiener process (5*10^5 counts, ASK=2*10^-3). Check it out!
OK. Let's consider it random, besides, a few calculations of the likely price level don't tell us anything. But on occasion, I'll run my algorithm on your series, just for the sake of curiosity. But again, time... :о(
By the way, but you also got your row by some algorithm, which must also be determined by its validity, it did not fall to you from the sky. :о)))) My Wiener process, I generated as described here: https://www.mql5.com/ru/forum/50458 post "grasn 30.01.07 00:04", or rather, as it is described in books, but with a very huge N, hence the calculation time. That's where the model didn't work.
PS1: I'm not the only one who "wants it!", and I don't have any complaints about the series :o)))
PS2: Ok, a couple or three days I'll investigate the phenomenon. Thanks for the help, maybe I'll show up soon :o)
Oh! That's where the discussion has moved to. Greetings, everyone!
I see there's been an invasion of mathletes. :)
Your pictures are very different from what I was constructing, which is interesting. Here, for example, a picture from the Northwind branch, every 30th tick, was stolen (very characteristic branch, by the way, half of the obvious, but useful observations, and half - anti-scientific nonsense like "proof of possibility to earn by playing beagle"; and juggling of special terms in addition).
It seems to me that your claims are completely unfounded. Not only you make completely ridiculous conclusions without understanding in purposes and objectives of what was, but you also try to attribute to me what I never stated. That's not nice. "What's your hurry, Muscovite, rattling an empty holster?"
Tiki is from a well-known company, there is a link to it above the thread.