How can I tell the difference between a FOREX chart and a PRNG? - page 14
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
So by correlation you mean any dependencies at all and different ways of evaluating them? Usually correlation is quite specific types of dependencies, and the presence of any dependencies between members of a series is called non-markness. You don't like the term, so what the hell with it, although there is no sectarianism there))
About the terver - I didn't suggest using it. The pricing model is primary, the methods of using it are secondary. If the model is real, it usually does not require much mathematics to use it, nor does it have practical applications in other areas.
Not really. You said that ON THE MOMENT and on the condition that there is a model. Any model is built with the approximate expectation of the methods by which it will be solved. No one needs models that are solved by sliding multiple integration over the surface of a semicircular topological spherical on semigroup horse in a vakuum, do they?
And this is where the sacred question arises. Please take it quite mathematically - seriously. I first heard this question in a Hollywood movie, where an old Mafioso asks another Mafioso, his friend:
"You say 'not sure'?! Jesus, what in the world can you be sure of?
My dear! The only thing you can be sure of is your daughter's love for you! That's what you can be sure of!"
Can't remember the name of the movie.
So, building a trading market model it would be nice if a mathematician asked himself this question - with respect to known methods he is going to use to solve the model and make further predictions.
Sergey, there is essential difference between tasks of random/unknown signal detection in radiolocation and in quotes, with all almost identical letters: for radars the calculation delay of an order of a pulse duration 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.
I am glad that we are communicating in the same language, and not inventing new definitions and nonsense. The fact that it takes infinite time in theory is well known in radiolocation. It's not acceptable in practice either. Air battle is rather short-lived, there is no time to think there. The two threshold Wald detector was developed for this case. Somewhere on this forum I posted a book about this detector. I checked it out and it works quite well. I guarantee that it is much better than the intersection of the two wagons)).
Z.I. I absolutely agree that theory should be applied wisely. Just surprisingly the tasks that are solved in radiolocation are very close to those that are solved in the market.
may be useful for those who know what correlation and autocorrelation are
https://www.mql5.com/ru/code/8295
it will soon be 5 years old.
might come in handy.
https://www.mql5.com/ru/code/8295
It'll soon be five years since he's been lying down.
It can't. It won't come in handy. Privalov, when you write formulas, you should at least unify the notation in your two lines, and/or write a full description of the variables. I read somewhere that it is so in mathematics, I do not remember exactly where I read it.
Privalov, what is mu-small?
By whom and when and HOW is the Mu-big trend "removed" ? Before or after the ACF is built? What is "x" there ?
It says "standard deviation", is it the deviation of what and from what ? And in what area - in all available, or in the window on i, or on m, or maybe on the sliding window?
Privalov, where did you learn to write formulas like that?
Where is the formula for the model itself, the analytical function that is shown in blue on the graph? What exactly is it modelling in there somewhere - the time series itself, or its ACF? Who can understand it but yourself?
Oh, sorry, I didn't realise straight away that it was written for my own people, "for those who speak the same language", for physicists or radio operators, for whom everything is "trivial" and "obvious".
The next questions are: how to distinguish the chart from the random drift and whether it is possible to earn from this kind of movement?
Simple - take a series of increments and count the MO)) If the drift is constant (MO=const), then no problem. But where do you get such a series? If the drift changes, the question is how fast. Tracking often tries to exploit situations of slow drift change. The question then is the size of the sliding window or the reference point of drift detection (trend). A moving window of a fixed size is appropriate when the trend component changes slowly without sharp changes, or if we are not able to identify the moments of these changes in time. If we can, then the reference points, as the moments of abrupt changes of the trend component, are relevant.
This is all irrelevant for modern markets and especially for fx. imha.
I am glad that we are communicating in the same language, rather than inventing new definitions and making up nonsense. The fact that in theory you need infinite time is known in radiolocation. It's not acceptable in practice either. Air battle is rather fleeting, there is no time to think there. The two threshold Wald detector was developed for this case. Somewhere on this forum I posted a book about this detector. I checked it out and it works quite well. I guarantee that it is much better than the intersection of the two mashka)).
Found it, thanks, I'll read it.
Z.U. The fact that it is wisely to apply the theory, I absolutely agree. It's just that surprisingly the problems that are solved in radar are very close to those that are solved in the market.
In five years you were the first to ask. Thank you. The rest of us just went by. But an attentive reader like you didn't go by and started asking questions. Because not everything is simple there. Half of the forum throws smart words, correlation, autocorrelation, etc., but they cannot calculate correctly. Now in order of the questions received.
By whom and when and HOW is the Mu-big trend "removed"? Before or after the ACF is built? What is "x" ?
is deleted before ACF is built. (You don't have to remove it, the shape (ACF) will not change). Mu is the straight line equation y=a*x+b it is this straight line that is the trend. It is recommended (but not obligatory) to remove the trend from the sample. the coefficients a and b are determined by the method of least squares. the indicator should have this procedure there. "x" is the time series being analysed. The clues that go to the input of the algorithm.
It says "standard deviation", is this the deviation of what and from what? And in what area - in all available, or in the window by i, or by m, or maybe in the sliding window?
I don't understand the question a bit, but I'll try to answer it. There is an analyzed data array. It doesn't matter which one it is (let's say the height of students in the class). From it you can always calculate the RMS (average height) and RMS (standard deviation of growth). Standard formulas. In our case, if we analyze N samples, this is the RMS of these samples.
Privalov, where did you learn to write formulas like this?
In high school, a little bit in school. What you see is a screenshot of MathCad. That's how formulas are written in MathCad (international standard of sorts...). If you enter it into matcad exactly like that, it will be calculated. I'm sorry again, but only what is written in MQL can be posted here. I should have probably posted it as calculated in Mathcad and shown how the same is calculated in MQL.
If i'm not mistaken, it's the same as in MQL. mu -v is small it's often called MOL. this formula has a different name for ACF but it's wikipedia, i have a mathad that does the math calculations. And as far as I remember before I posted the indicator, I checked all calculations with matcad (to make sure everything coincides).
Where is the formula of the model itself, the analytical function that is shown in blue on the chart? What exactly is it modelling somewhere in there - the time series itself or its ACF? Who can understand it but yourself?
This is the most important, the most delicious and correct question. At least for me during all the time I've been on this forum. I'll try to explain. As you see the model, that blue one, practically completely repeats ACF of a real process (process that we analyze). I.e. we have a quote stream at input whose ACF has the form shown in red. And there is a formula (model) whose form almost exactly repeats the input process..... and since there is a model, using it (the model) we can predict how the object will behave next...what we are all looking for here, an opportunity to predict and try to make money on it.
Not to write too much, I will refer to this forum again where I posted Kalman filter and said that I put in it the model of inertial-link of the 2nd order. It is this ACF that is shown in the figure with the blue line. Look for the formula here somewhere or find the book by Tikhonov Vladimir Ivanovich. "Transformations of Random Processes". This link is described there in detail.
That's it. Thanks again for your questions, I'm glad someone could benefit from my work.
Found it, thanks, I'll read it.
If it is not a military secret, what is the radar pulse duration on modern aircraft? It's interesting to estimate how much it is in meters, say, at 10 Mach.there are radars with nano pulses.http://forums.airbase.ru/2002/01/t3307,10--fiziko-matematicheskie-osnovy-radiolokatsii-prodolzhenie.html
the specifications of these radars are roughly like this http://wikimapia.org/6807622/ru/%D0%A0%D0%9B%D0%A1-%C2%AB%D0%94%D0%BE%D0%BD-2%D0%9D%C2%BB