The market is a controlled dynamic system. - page 22
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
Do you want to pinch?
And I presume you possess not only the terminology but also the knowledge? :D)))))))))
: just a lot of crazy people's bullshit.
here's the thing...
If the 2nd order AR model, which was, so to speak, in all seriousness, tried on the existing realities, is the top of econometrics achievements... what can I say... when applied to forex...
.
Such elementary models work satisfactorily for linear objects of unsophisticated dynamics ;)
If such a model is closed by feedback, it can be applied to weakly nonlinear objects as well.
But it seems that all this is unknown to you...
By the way, do you know the difference between stationary object and non-stationary object?
If that 2nd order AR model, which was, so to speak, in all seriousness, tried on the existing realities, is the top of the econometric achievement... what can I say... in the application to forex...
It's just that AR model is widely known for being applied by the US government. Bryukov applied it to forex in his book, claims to be successful, I have not been able to apply it. But it is not the only model. There are a number of classes of models in econometrics and it's not about them, it's about us. Some have succeeded in building a robust model, and some haven't.
While we are stressing here and trying to test others, people are taking a ready-made toolkit that has been tested for decades, and riveting models. That's what I'm saying. You don't have to reinvent the wheel. You have to learn a ready-made toolkit that is open, and hone it for the rest of your life
But it also depends on the algorithms.
I can't get away from the idea that it's impossible to create a model for centuries on quotients as they exist in nature.
The plan is simpler: create a model that meets the criteria: no autocorrelation and no heteroscedasticity. Make a forecast 1 step ahead (one step is M1 or D1). Then we repeat it. So to say, adaptation.
Seconded. I believe that the trend should be viewed in relation to the hindsight within which it is defined. In the general case, it seems to be the angle of slope of the line describing the price matrix for the period in question. In order to avoid discrepancies in the methods of determining the numerical value of the coefficient which defines the trend, it should probably be determined by the method of nominal Gauss quadr. Gauss, which was not the case with the scatter around the straight line in the retrospective considered
Here I am demonstrating the futility of a linear trend in Forex trading.
Nevertheless, to be fair, it should be noted that any attempt to develop a quotient model at the initial stages should always include the trend and a constant in the regression equation. Moreover, a number of tests provide for automatic inclusion of a trend and a constant. But then the question needs to be answered: is the use of these values significant in the model? The answer is given by the value of the probability that the coefficient at trend and constant are equal to zero. In my modest practice in the final model neither a trend nor a constant remained due to this proved reason.
.... without any smoothing or application of "filters", while application of other methods will lead to chaos and inconsistencies.
Application of filters is not an end in itself and I never said so in my posts. You must first define the purpose of modelling. I don't have an aim to "define the trend" but to give credence to the calculated error of the forecast. And you can only trust a forecast if its error, acceptable in magnitude, changes little, and you can only get that if there is no trend in the error (or rather no dependence between bar errors) and none of the types of heteroscedasticity.
Unfortunately, this is how econometrics works - it is a pile of tools with practically no instructions on how to use them. My presence on the forums is due to the fact that I have spawned a huge number of models "from the ground up", but I have not managed to get a single one sustainable. Development of any model even for one currency pair is very time consuming and dull process and I am looking for people on this forum who will join me in building and analyzing models. All of us "old timers" should remember that thousands of people graduate from institutes every year with a degree in econometrics and these people are not inclined to pedal their own bikes.
I am waiting to hear back from you, or a rant, do you agree with my definition of a trend?
Trend is not important in my concept.
I am doing a statistical analysis of a quotient, and it is possible if the quotient does not contain a deterministic component, as it muffles the noise. So I'm trying to isolate the deterministic component, which, by the way, is extrapolated beyond the sample named forecast. The next question is: what is the error of this prediction? If the residual after removing the deterministic component i.i.i., then there is no problem, but (1) this is not the case in financial markets and (2) even if it is, there is no guarantee that at the next forecast step we will not get something else instead of i.i.i., which will not allow us to extrapolate the error, since it will be variable.
This proves nothing: the main goal of the US government is not its own profit, but the economic growth of the state. It is not the same as the personal profits of a single trader.
This proves nothing: the main goal of the US government is not its own profit, but the economic growth of the state. It is not the same as the personal profits of a single trader.