Obtaining a stationary BP from a price BP - page 4
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 start a thread and discuss what you are interested in.
I'm sorry for being absurd, but I'd like to get some use out of the forums.
Stationary BPs are known to be predictable if they are not white noise.
What is the relationship between predictability and the type of spectral density? So with pink noise, everything is already predictable?
Hmmm... What does predictability even mean in this context, the ability to use in a profitable TS or the desire to predict price movement?
As we know, stationary BPs are predictable if they are not white noise.
You see, here's the thing:
1). they are predictable, if at all, then they are MEDIUMIUM predictable.
2). if any time series are called "stationary", they are thereby assumed to be REMARKABLE, i.e. without any spectrum there.
3). the "forecasting" error, acceptable for normal economics (5%...10%), is killer for marginal (amplified, leverage) trading.
If you ignore these features, all other reasoning will lead nowhere.
Do a little more thinking :) Just because the MO of a random variable=0 doesn't mean that CB itself can be replaced by zero, as you deftly do :) :)
I always rack my brains, instead of taking all sorts of nerd nonsense for granted. It's better to double-check it than leak it.
In probabilistic models it is quite adequate to substitute known expected payoff instead of unknown value.
Of course, it won't be 0 and 0 is the expected value. Suppose we want to get a more down-to-earth model. In this case we have white noise with variance = constant, MO = 0. Okay. A pseudo white noise generator is no problem. We adjust the variance. We get BP(x) = rnd(x).
Substitute it into the formula. We obtain:
forecast(x) = price_appr(x) + rnd(x) = fit + noise = bullshit, not a forecast
The botanical method is ahine. I suggest that we don't even come back to this question, as all paths lead nowhere (just calculate the end result).
The question is why do we need to take noise as a model when it's more adequate to take the residuals as a stationary BP? After all, if the residuals are delta(x) stationary and are not noise, then they are by definition predictable. And consequently, you can get a mathematical model of these very residues by extrapolation: delta_appr(x) ~ delta(x). Only in this case the mathematical model will correct errors of the fitting - price_appr(x) - in the extrapolation. It may not be 100% correct, but it will be correct.
Open[time + i + j] ~ forecast(time + i + j) = price_appr(time + i + j) + delta_appr(time + i + j)
2). if any time series are called "stationary", they are thereby assumed to be LOCAL, i.e. without any spectrum.
A spectrum is a type of function, so there is always a spectrum.
I enclose for H1 several spectra for 10 days with a shift of one day, during which: there are major trends and minor trends;
The periodicity of them and of the others is changing; the trends appear and disappear - and this is during the day.
How can it be transformed into something stationary? And at the expense of noise separation. Noise has nothing to do with it. We can't deal with the trends during the day.
The question is, why do we need to take noise as a model when it is more adequate to take the residuals as stationary BP? After all, if the residuals - delta(x) are stationary and are not noise, then they are by definition predictable. And consequently, by extrapolation we can get a mathematical model of these very residues - delta_appr(x) ~ delta(x). Only in this case the mathematical model will correct errors of the fitting - price_appr(x) - in the extrapolation. It may not be 100%, but it will be.
What are we singling out from BP? Everything is fine for the chaff: the FFT - we may even predict the target, but where will the market go? Don't be lazy and open to the previous post. You can clearly see what we are dealing with.
The question is why do we need to take noise as a model,
On the one with 1:100 leverage you are working exactly with noise - those fluctuations that big market participants-banks working with 1:10 leverage consider to be SHOOT (from their point of view). And you can't change your sitting point (your sitting point determines your point of view), it's not to your advantage.
On the one with 1:100 leverage you are working exactly with noise - with the fluctuations that the big market participants-banks working with 1:10 leverage consider to be SHOOT (from their point of view). And you cannot change your "point of view" (your point of view determines your standpoint).
You are a pipsqueak.
A spectrum is a type of function, so there is always a spectrum.
What kind of nonsense is that, where did that come from? Uncle Vasya's definitions were the only thing missing here.
A spectrum is the interpolation of a segment of a function by a finite set (sum) of sinusoids.
What kind of nonsense is that, where did that come from? Uncle Vasya's definitions were the only thing missing here.
A spectrum is the interpolation of a segment of a function by a finite set of sinusoids.
That's pretty much what I meant. Sinusoids are for Fourier, but there are other functions, but that's not the point. Not everything is Fourier decomposable and that's the problem for us, as forex BP is not representable by Fourier.