Machine learning in trading: theory, models, practice and algo-trading - page 1109
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
That's why he compared it to Fermat's theorem, which they couldn't prove for 300 years.)
All of us are looking for some singularities, we read texts of radio engineers - why waste time on all this?
The behavior of the dispersion of increments in the prices of financial assets is very intricate and thick tails are the most primitive. More than 100(!) GARCH models have been created to model all these curiosities in the variance (and before that in the mean) - they're about that, aren't they, again some Ferma....
If about variance (non-stationarity), so there is no systematically, model by model GARCH, through the tester, to see something. No, it's much more interesting to bullshit.
We all look for some singularity, read texts of radio engineers - why waste time on all this?
The behavior of the dispersion of the increments in the prices of financial assets is very intricate and thick tails are the most primitive. More than 100(!) GARCH models have been created to model all these curiosities in variance (and before that in mean) - they're about that, aren't they, again some Ferma....
If about variance (non-stationarity), so there is no systematically, model by model GARCH, through the tester, to see something. So no, it is much more interesting to bullshit.
I don't know, I just read what they write about sometimes. I think he's an asset manager, but I don't know.
I do other things myself.
We all look for some singularity, read texts of radio engineers - why waste time on all this?
The behavior of the dispersion of the increments in the prices of financial assets is very intricate and thick tails are the most primitive. More than 100(!) GARCH models have been created to model all these curiosities in variance (and before that in mean) - they're about that, aren't they, again some Ferma....
If about variance (non-stationarity), so there is no systematically, model by model GARCH, through the tester, to see something. So no, it is much more interesting to bullshit.
In my opinion, the main reason to search for such singularities is a significant non-stationarity of price (increments) that cannot be reduced to stationarity by standard methods. Any regression, in one way or another, reduces everything to stationarity.
Perhaps there are some adequate approaches to building a non-stationary regression - with coefficients that depend on time. Probably, this can somehow be done by generalizing the concept of stationarity.
That's why he compared it to Fermat's theorem, which they couldn't prove for 300 years.)
There is a significant difference) Fermat's Theorem was originally formulated mathematically correct. This does not exist in our field and is unlikely to be possible.
And this, in fact, is correct:
And therefore there are no similarities between different timeframes. And what does it mean? And the fact that with "one measure" to the five-minute and daily periods should be approached with great caution and with distrust to those who formulate it as an axiom.
Well, that too, by the way:
I'll add to what I wrote in my commentary: an important gnoseological conclusion: if we build trading methods on past daily or even hourly rates (if there are many ticks in the hours), then it is meaningless to "dig deeper" the functions of the second degree nonlinearity of price increments or price logarithms (prices in the text easily change to price logarithms without losing the essence).
The second conclusion is not very clear, it is clear that logarithmic increments like returnees are good, but the first conclusion, IMHO - nonsense.
Maybe, there is nothing theoretically to substantiate the similarity of timeframes but I was virtually convinced that it exists and, moreover, it appears even in instruments coinciding in one of the currencies.
I also do not know how to explain it mathematically, maybe it is a psychological factor, because most people use the same charts from the standard set of timeframes for analyzing an instrument and making decisions.
Let's go on a date. Come alone,
Without a blacksmith, we don't need a blacksmith)))
radikal.ru/video/uz7qxNNhyO8
The second conclusion is not very clear, it is clear that logarithmic increments such as returnees are good, and the first conclusion, IMHO - nonsense.
Maybe there is nothing theoretically to substantiate the similarity of timeframes, but I was virtually convinced that it exists and, moreover, it appears even in the instruments coinciding in one of the currencies.
I also do not know how to explain it mathematically, maybe it is a psychological factor, because most people use the same charts from the standard set of timeframes for analyzing an instrument and making decisions.
The first conclusion is just good because from smaller TFs to larger ones statistical characteristics float.
It turns out that what is good for statistics is death for trading:)
Do you know Gorchakov? Again he writes some thoughts on smradlab
https://smart-lab.ru/blog/499678.php
Even if we assume the applicability of the CPT here (for example, due to gaps there are doubts about the limited variance of xi in the population), then due to non-stationarity of the increments we do not know their expectation in the future and, accordingly, the expectation of their sum will also be unknown to us. In other words, the price will be distributed according to the normal law, but with unknown parameters. It's not quite clear what good it will do.
PS. The author of the article wrote about it in the commentsFrom smaller TFs to larger ones the stat characteristics float.
They generally float, in all dimensions, in linear time and in its scale, from instrument to instrument, etc. The main question is HOW IT GOES, what is the functionality of changes in statistics, in particular how regular are the change functions, if statistics are not changing continuously (at least piecewise constantly) and either regularly, then trouble, then only insiders are left to shake off the market.
But the regularity of BP statistical characteristics change in time is present, the regularity of dispersion is obvious to everybody, it is caught even linearly, the higher moments are also relatively predictable, although they are less useful for our purposes, but the best thing is a sign of future increase, it is bad, on the verge of noise and it is sad, yachts and islands are postponed.