The market is a controlled dynamic system. - page 63
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
All this is true only if we consider a system that constantly predicts and makes trades. But it does not fit at all to variant when system detects optimal entry points, where in its opinion quality forecast is possible, and only then chooses direction of forecast. In practice there can be 2-5 entries in a week on a minute chart, i.e. number of made forecasts is less than 0.1% of number of octivation samples.
we can also stack up and down inside a trade on some shallow frame, if PF lag on returns is critical
Heh, that's tempting of course, but the necessary error-return happens AFTER we've bought. And we have to evaluate the criterion and determine the direction BEFORE entering. So, if we know BEFORE entering that an outlier to one side is more likely than an outlier to the other, we can simply factor that into our system, and use it from then on.
Also, I was wrong to erase the diagram along the way: there were TWO errors drawn on it: 1) internal modelling error, about which I said it should be normal and uncorrelated, as it is a criterion that the model adequately describes the system structure (econometrics has nothing to do with it), and 2) prediction error, which should not and will not be normal, as the input has those very unpredictable abnormal outliers. And this is even good, because otherwise even our potential earnings would probably be guaranteed to be 0.
Also, I was wrong to erase the diagram along the way: there were TWO errors drawn on it: 1) internal modelling error, about which I said it should be normal and uncorrelated, as it is a criterion of the model adequately describing the system structure (econometrics has nothing to do with it), and 2) prediction error, which should not and will not be normal, as the input is those unpredictable abnormal outliers. And this is even a good thing, because otherwise even our potential earnings would probably be guaranteed to be 0.
Yes, the scheme disappeared quickly - didn't have time to save either.
Alexey, what forecast horizon do you think might be optimal/possible?
Limitations should be there anyway - the error will grow if one tries to look too far...
Or is it a variable parameter and the system has to determine it somehow in the course of data arrival/accumulation after the start (until it enters the working mode)?
about internal modelling error - how is it counted?
The deterministic component of the input signal and the structure and parameters of the system (blind deconvolution problem) is estimated using some optimization method on a chosen interval using the chosen criteria, then the input estimate is run through the model; the difference between the obtained output and the real process, thus, is the noise estimate.
sergeyas:
Alexey, what forecast horizon do you think might be optimal/possible?
Limitations should be there anyway - the error will grow if you try to look too far...
Or is it a variable parameter and the system has to determine it somehow in the course of data arrival/accumulation after start (until it enters operating mode)?
Rather variable, it can be determined from the derived model parameters, roughly speaking there is always some characteristic relaxation time in the system, the horizon can be proportional to this figure.
So far, this is how it's worked out, with no automatic adjustments yet.
GBPUSD H4
.
GBPUSD Daily
Quotes are in the 5 digits.
Incidentally, the question of the forecast horizon, its possible optimality and variability, is not so simple at all.
Suppose some prediction system pp(n) is built, which makes a prediction for n steps ahead at the k-th step. Moreover, for different n the prediction error ep(n) will be different. Moreover, the prediction errorep(n) will change from step to step, i.e. it depends on k.
Let define Nep as the horizon giving minimal prediction error atthe k-th step, when predicting at(k-n)-step
We can clearly see the variability ofNep from step to step.
However, there is some dependence of this variability onNep for different parts of the process.
Here is a video giving a good visualisation of the variability ofNep
By the way, the question about the forecast horizon, its possible optimality and variability, is not so simple at all.
Suppose some prediction system pp(n) is built, which makes a prediction for n steps ahead at the k-th step. Moreover, for different n the prediction error ep(n) will be different. Moreover, the prediction errorep(n) will change from step to step, i.e. it depends on k.
Let us define Nep as the horizon giving minimum prediction error atthe k-th step, when the prediction is done at(k-n)-step
We can clearly see the variability ofNep from step to step.
However, there is some correlation between this variability for different parts of the process.
At first glance it looks very similar, but something tells me that it is not so much k that is responsible for variation in Nep, but rather the quality of the
forecasting.
It turns out that the model for some reason (maybe - incorrect assumptions, etc.) doesn't take into account some important factors, properties of the
of the process or insufficient observation history.
What is k in essence? Not the passage of time? If it is, then it is not correct to blame it (I think).
Something tells me that it is not so much k that is "to blame" for Nepal's variability, but the quality of the forecasting system itself.
It turns out that the model for some reason (perhaps incorrect assumptions, etc.) does not take into account some important factors or properties of
of the process or insufficient observation history.
What is k in essence? Not the passage of time? If it is, it is incorrect to blame it (I think).
No, of course,k is not itself "guilty" of the variability of Nep, butthe state of the process at timek makes it possible to predict the further development of the process - an external factor. And also affect the internal factors of the forecasting system - the failure to take into account some facts and properties of the process.