Machine learning in trading: theory, models, practice and algo-trading - page 2728
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
to the choice of training (optimisation) intervals
we can start from the classical Random-Walk model, from which we have not gone far and in which deviations are ~2*sqrt(T). (two is the remnants of trends)
It is remarkable in that
(a) it is scalable (exactly the same parabolas, the same lines as on the screenshot will remain when switching to the lower TFs and ticks),
(b) we will compare the results directly and directly with it anyway
(c) it has characteristic intervals. The same focus of the same parabola.
these are the intervals to be orientated on.
Technically quite solvable, I guess. The question is how to interpret the results of such an experiment.
Matstat has many tests to check the homogeneity of samples, for example. If I understand your terminology correctly, of course.
And what is wrong with these criteria?
I have saved but not implemented link1, link2, link3.
And what's wrong with these criteria?
I have kept but not implemented ref1, ref2, ref3.
The criteria are quite good. You can add to them the Kolmogorov-Smirnov test as the most popular one.
This is about the principles of forming comparable samples, which have nothing to do with the criteria themselves.
The criteria are quite good. We can add Kolmogorov-Smirnov test as the most popular one.
This is about the principles of forming comparable samples, which have nothing to do with the criteria themselves.
How is it irrelevant? If the criteria change with the size of the sample, then the sample must be forced into the necessary criteria. I don't understand something...
to the selection of training (optimisation) intervals
we can start from the classical Random-Walk model, from which we have not gone far and in which deviations are ~2*sqrt(T). (two is the remnants of trends)
It is remarkable in that
(a) it is scalable (exactly the same parabolas, the same lines as on the screenshot will remain when switching to the lower TFs and ticks),
(b) we will compare the results directly and directly with it anyway
(c) it has characteristic intervals. The same trick of the same parabola
these are the intervals to be orientated on.
For any point we can specify the value of d, at which it will become the focus) I'm not even talking about the dependence of the focus on the ratio of scales in price and time).
How is that irrelevant? If the criteria change with the size of the sample, then the sample has to be loaded into the right criteria. I don't understand.
A criterion is basically just a formula into which you put the samples you are interested in in order to compare them. You decide which samples to compare, not the criterion.
For any point you can specify the value of d, at which it becomes the focus) About the dependence of the focus on the ratio of scales in terms of price and time I do not even stutter)
this is a question of metrics, how many minutes (ticks) corresponds to 1 point, what to consider as the pi/4 angle and why; but this way you can slip to the unloved by TC Ghana :-)
and so - all currencies exactly correspond to the law deviation=2*sqrt(T). Which is not surprising
A criterion is essentially just a formula into which you need to substitute the samples of interest in order to compare them. You decide which samples to compare, not the criterion.
You don't think the Criterion makes sense? Take ten samples of different sizes and compare them - choose the one with the best performance on several indicators that are responsible for the similarity/similarity/homogeneity of the samples.
Do you think the Criterion doesn't make sense? We take ten samples of different sizes and compare them - choose the one with the best performance on several indicators responsible for the similarity/similarity/homogeneity of the samples.
The meaning is/is not in the criteria, but in the way they are used. Your way of using them is not clear at all - what you are comparing with what and for what purpose).
The meaning is/is not in the criteria, but in the way they are used. Your way of using them is not clear at all - what you compare with what and for what purpose).
Okay, since you're being clever, you're not willing to think in that direction. I won't bother.