Machine learning in trading: theory, models, practice and algo-trading - page 648
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
There are at least two time series in cointegration.
But these are not all of them.
These series are not stationary.
But that's not all.
These non-stationary series must be connected in such a way that the result is stationary.
Trading decisions are based on this STATIONARY series and thus guarantee the very possibility of the forecast.
Well, what about me? :) I'm not sure about stationarity, I haven't made any tests... but it looks like
i think the Arch is present... do you know how to build a fast separation in MT-ha and look at the tails? P is not an suggestion ))
or just with Dickey-Fuller?
Well, what about me? :) about the stationarity, however, I'm not sure, I have not done any tests... but it looks like
arch seems to be present... do you know how to build quick splits in MT-ha and look at the tails? P is not an suggestion ))
Or just dickey-fuller?
Read what I wrote again: TWO rows.
DickeyFuller, of course.
I'm only discussing R's. I have respect for python. Everything else is a sandbox game - I have no time for it.
Here, it means we need to determine if there is memory... is it determined by tails?
I do not know how to tail, this is a specialty of those who know how to garch. There are only a couple of people on this forum who understand it at all :)
You can take your favorite model (for example gbm) and try to find model parameters so that training with k-fold crossvalidation would give results R2>0. Take a couple dozen of past values, using them to predict the next (target), and by sliding window to make a whole training table.
At any random data the result of R2 will never be greater than zero. If it's positive, it means you can predict next values from past ones, you have memory.
I only discuss R. I have respect for python. Everything else is a sandbox game - I don't have time for it.
))
Read what I wrote again: TWO rows.
DickeyFooler, of course.
I'm only discussing the R's. I have respect for python. Everything else is a sandbox game - I don't have time for it.
There are 2 rows, I wrote that cointegrated rows (2 currency pairs)
not distracting then ))Tails I can not, this is a specialty of those who know how to garch. There are only a couple of people on this forum who understand it at all :)
You can take your favorite model (for example gbm) and try to find model parameters so that training with k-fold crossvalidation would give results R2>0. Take a couple dozens of past values, predict the next one using them (target), and use sliding window to make a whole training table.
At any random data the result of R2 will never be greater than zero. If it is possible to get in the plus - then by past values it is possible to predict the next ones, we have memory.
But if there's an Arch effect, it's not sure that it will return right away... and dispersion is variable... but it's okay, we'll analyze it...)
There are 2 rows, I wrote that cointegrated rows (2 currency pairs)
I am not distracting you then ))I judge by the picture, there is only eurusd, what is the second one?
I judge by the picture, there is only eurusd, what is the other one?
gpbusd in this case, but in fact any, makes no difference
So the property of noise is that it does not need to be predicted) it deviates from the average - so it will soon return... but if there is an arch effect, it is not certain that it will return immediately... and the variance is variable, but okay, we'll figure it out)
Why? For the substantiation of an algorithm, which guarantees that it will come back.
It's necessary to construct two or more time series (two pairs) in a special way.
gpbusd in this case, but in fact any, makes no difference
Wonderful discussion, like a fig in your pocket.
And how did you put these two pairs together that there is such a residue?