For those who have (are) seriously engaged in co-movement analysis of financial instruments (> 2) - page 36
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
As you can see, over 1000 reports the spread has not changed so radically. Given a recalculation of weights, for example every 10 reports, you can use the spread. commission of -10% of the convergence.
Explain what does "1000 reports" mean? What is the time duration of the sample on which the ratios are calculated, and what is the duration of the subsequent plot on which you investigated the obtained ratios. Your pictures do not indicate this, and we are not telepathic.
Two samples of 1,000 hourly closing prices.
Yes, I apologise for the hasty conclusions, things are much worse here than in indices.
I've looked at a long stretch of history, the spread from currencies often changes its behaviour, and there is very little "inertia", i.e. the structure changes very quickly.
I guess that optimization with a sliding window will not do any good here, maybe someone tried it? Although it's worth trying because in its "adjusted" state spread looks very interesting.
The author is right about the spread having to have a real economic relationship, otherwise it's just a fit.
How are the weights of traded instruments calculated? What do we buy, what do we sell?
Regression mainly.
Regression mainly.
How much more?
The currency pairs themselves are in trends and by trading spreads it is possible to enter the market against the trend and a move along the trend can easily overlap a deviation from the trend.
That is why I would like to see a specific regression with coefficients. And then in this details what are we trading?
Two samples of 1,000 hourly closing prices.
Yes, sorry for the hasty conclusions, things are much worse here than on indices.
I've looked at a long stretch of history, the spread from currencies often changes its behaviour, and there is very little "inertia", i.e. the structure changes very quickly.
I guess that optimization with a sliding window will not do any good here, maybe someone tried it? Although it's worth trying because in its "adjusted" state spread looks very interesting.
The author is right about the spread having to have a real economic relationship, otherwise it is just a fit.
Try Recycle(https://www.mql5.com/ru/code/10096), it has a sliding window.
because in its "fitted" state the spread looks very interesting.
The author is right about the spread having to have a real economic relationship, otherwise it is just a fit.
In a fitted state any bot looks very interesting. A fit is a fit. There will be no miracle on OOS. Or it will, but for a very short time.
Any bot looks very interesting in its trimmed state. A trim is a trim. There will be no miracle on OOS. Or it will, but for a very short time.
I think that's not quite true, and it's a bit incorrect to compare fitting a TC on a single instrument and creating a synthetic from multiple instruments that have a real economic connection.
A question to the author of this thread, if he is still looking at it.
Why do you think that multiple regression is "lopsided" and where do you see the advantage of the method implemented in Recycle?
The main argument you make is that recycle is a tool for finding market relationships. So with regression, using any optimisation model (each to each), you can go through a bunch of instruments and derive some indicators as a comparison parameter, for example, the same coefficient of determination.
I think this is not quite the case, and it is a bit incorrect to compare fitting a TS on one instrument and creating a synthetic of several instruments that have a real economic relationship.
And why on one instrument? You can run TC on many instruments, and adjust it individually on each of them. The total result will look great :) And you can come up with similar explanations, saying that it is all because of economic interrelations :)
So, fitting is still fitting. To assess the real performance of a system or synthetic, it is necessary to investigate it all in dynamics. I.e. in this case, using a sliding window, it is necessary to get a number of fitted values for each coefficient. And then assess the stability of these coefficients. To be more precise, their inertness.