Machine learning in trading: theory, models, practice and algo-trading - page 1595
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
.....
again we can wait for the return
It is possible to wait for the return, but in the real TS it makes no sense.
Formulate TS and you will see that to "wait" we will have to reserve such a part of capital that the profitability of the strategy on the stationary section will be miserable
adf_test in R
That's not it, Dickey-Fuller only works for autoregressive processes.
That's not it, Dickey-Fuller only works for autoregressive processes.
it's his problem to recognize the process, we don't put him a cotier, but a synthetic, and not really synthetic
If adf_test gives 0.01, then you can go smoke bamboo, no?
that's his problem - not mine
to determine this
you don't need to trade changing modes, you need to change strategies when they change. If it is scalping, there will be hundreds of trades for each one. The task is to switch the strategy in time, i.e. to determine the mode change as early as possible, or even predict it.
If you solve this problem, the Grail is certainly in your pocketMode change is a very controversial question
I've already asked the question - how to consider stationarity - as a moving window or an elongating row?
but let's assume it doesn't matter and we got "nonstationarity" in both cases
no one seems to have ever managed to create TS on nonstationarity
another thing is that we can try to make the forest (while we are still stationary) determine "enter or not", but I think it will be stupid or guess
We can switch the TC "backwards" when we get out of stationary, but there are no reliable statistics
although we can apply a technique where if we know we will go back to zero, we can reverse the drawdown
it's certainly not easy, but it can be done... thank you for the idea of using the old billet.
If it works, of course.
But we need to think about it
mode change is a highly debatable question
I have already asked the question - how to consider stationarity - a shifting window or an elongating row? there are no answers
but let's assume that it does not matter and we got "nonstationarity" in both cases
no one has ever managed to create TS on nonstationarity
another thing is that we can try to make the forest (while we are still stationary) determine "enter or not", but I think it will be stupid or guess
We can switch the TC "backwards" when we get out of stationary, but there are no reliable statistics
although we can apply a technique where if we know we will go back to zero, we can reverse the drawdown
it's certainly not easy, but it can be done... thank you for the idea of using the old billet.
if it works, of course.
but I need to think about it
I prefer to count stationarity from the significant incremental lag, as in the last article. For example, lag ~24 for hourly increments is robust. Then there is no uncertainty in the choice of window.
Forest is not forest, here, for example, boosting (everything is already done) and it works until the mode is changed. When average increments happen, the model crashes, which is natural. It's amazing how long it took to figure out how much the bias of the mean (not even variance) affects tc. Dumbass.
What's missing has already been thrown around. Here is a simple example of clustering the increments (read: determining modes) and testing them on new data (3 modes found). I chose an easier one on purpose, I haven't experimented yet.
https://www.quantnews.com/k-means-clustering-creating-simple-trading-rule-smoother-returns/
In other words, for each cluster a separate model is fetched. The current cluster (mode) is defined, models are switched accordingly.
Nothing else to think about, I will have to do. K-means is not the best option, but it'll do as a test case.
If adf_test gives 0.01, then you can go smoke bamboo, can't you?
he rejects very special cases of nonstationarity (autoregressive, SB type), and nonstationarity can be much more diverse.
The point is that, according to the Wold theorem, any stationary process can be considered autoregressive, but there are very few autoregressives among nonstationary ones.
he rejects very special cases of nonstationarity (autoregressive, SB type), and nonstationarity can be much more diverse.
The point is that, according to the Wold theorem, any stationary process can be considered autoregressive, but there are very few autoregressive processes among nonstationary ones.
so what? let's say he rejects
and what-2?
So what? Let's say he does.
and what-2?
What do you mean, "Suppose rejects"? Do you even know what a statistical test is, a statistical hypothesis?
What do you mean, "Suppose rejects"? Do you even know what a statistical test, a statistical hypothesis is?
Can you answer questions without asking counter questions?