Econometrics: one step ahead forecast - page 71
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
I haven't seen any publications on NS with R^2. That's what I mean. What does regression analysis have to do with it?
RA has nothing to do with it, it's the measure (minutiae) that changes. Straights, curves and obliques are shifted.
As they say - to the nearest constant.
Take this and calculate for yourself what equity will be at the 13th bar after you start buying. There is no netting, we are trading in DCs!
I really sell the muv and really take care that the open short positions in total correspond to the sold muv (well, of course, with a factor of 13), making "escorts".
About the period build-up - that's the next idea, very sensible, by the way. But for now we need to understand the basic one.
I don't argue with that and completely agree. On the 13th bar the price of your average cumulative position will correspond exactly to the value of the muving with the averaging period of 13. Netting will show this fact even more clearly. The problem is that on bar 14, you can no longer make your position equal to the current average with a period of 13, the moving average will be gone and your average entry price will remain the same. The only thing you can do is to average again and use the mouwing with period 14, on bar 15 you will have to use the mouwing 15, on bar 16, and so on to infinity. In the limit, the moving average will become so big that the price will not return to it in the foreseeable future. I.e. it is not possible to do any "accompaniment".
Tomorrow I will write out my thought in the table to make it clear.
Closing old micropositions should not be allowed as the overall result of the Y operation will be distorted by the realised profit/loss.
It's simpler, just old micropositions will be closed at the current price of zero bar, and in order to maintain the period old micropositions should be closed at their opening prices 13 bars ago, which is not possible. But the muwink kind of closes the old values at the old prices, it can do it, because it is an indicator.
to:faa
While we've been discussing and discussing, the freak regiment has arrived:
Predicting time series using exponential smoothing
Predicting time series using exponential smoothing (end)
The candidates are a perfect match for your method.
to:faa
While we've been discussing and discussing, the freak regiment has arrived:
Time series forecasting with exponential smoothing
Predicting time series using exponential smoothing (end)
The candidates are a perfect match for your method.
I did. He declined. I was interested in how to adjust the smoothing parameters based on the prediction error. This is part of the problem for me.
There's another problem. Earlier I posted simulation results for one model. Now I'm posting it for another one:
kotir hp1(-1 to -2) hp1_d(-1 to -1) eq1_hp2(-1 to -3) eq1_hp2_d(-1 to -4)
where HP smoothes the 1/DX quotient, i.e. the inverse of the dollar index.
Here is the result:
Very good model. lends itself to optimization by LM ACF and max Prob C
And here are the depressing results:
When forecasting inside the sample I have a fantastic profit factor, especially please pay attention to the profit factor in the observations. But outside the sample ..... Why are such rosy results not extended one step further? I can't understand it.
Vladimir: SanSanych's outlook is not narrow, but the task is specific, it seems to me. imho, of course.
And a bulldog's grip...
Usually the makers of such models quickly run them in the tester, make sure they are draining, and move on to new models. But here the starter shows daily predictions in real time expecting a miracle - masochism of sorts.
When predicting inside the sample I have a fantastic profit factor, especially please note the profit factor in the observations. But outside the sample ..... Why are such rosy results not extended one step further? I can't understand it.
Finally the adherent of the cult, has revealed the main secret of the religious trick!
Elementary, Watson! Because they are non-stationary. Stationarity is when dispersion and expectation are constants and do not depend on the sample, on which they are measured. I.e. in any other independent sample, we should get approximately the same constants. If we don't, then the stationarity hypothesis is disproved.
The stationarity hypothesis can be tested in another way by increasing the sample dimension. In the case of stationarity both variance and expectation should also remain constants.
I haven't seen any publications on NS with R^2.
In any algorithm you can use any error...and p-Q in NS as well...
At last the adherent of the cult, has revealed the main secret of the religious trick!
Elementary, Watson! Because they are non-stationary. Stationarity is when dispersion and expectation are constants and do not depend on the sample, on which they are measured. I.e. in any other independent sample, we should get approximately the same constants. If we don't, then the stationarity hypothesis is disproved.
The stationarity hypothesis can be tested in another way by increasing the sample size . In the case of stationarity both variance and expectation should also remain constants.
the fuck it should in theory...but in practice it won't...and it will depend not only on the size of the total sample but also on what's in it...