Not Mashka's business! - page 8
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
Naturally!
After all, if you can predict one bar ahead, you can also predict two bars using recursion, and there by induction. But the forecast error will grow exponentially with the growing horizon, that's why we're not interested in searching for relationship between the accuracy of the elementary forecast (for one bar) and the confidence interval width as f-fi of the forecast horizon. Let amateurs do that. You and I will study the quality of the forecast basis itself - 1 BAR ahead and that's it! True, we will gather statistics by predicting each time by 1 bar and making one step forward, and so on 10,000 times. Just to be sure.
Yes we can do that, for starters. Aren't you interested in setting a boundary for the visibility of your model?
PS: also, with a maximum forecast of 1 bar you're very wrong. This is a limitation of only the models under study - yours and Burg's. And here you are generally right, the statistical proof of the size of the horizon I have,
- not so much, although sometimes there are long forecasts.But on the other hand look at the error graph - it's not bad at all, the error varies by 1 point.
But returning to our philosophical conversations, you should not limit yourself, because you limit yourself, but not nature. So man would never take to the air or sink to the seabed. By the way, about the prediction, I think I'll give you conclusive proof that you can predict quite far ahead relatively soon, but let's not be in a hurry. We'll get there in time :o)
But back to our philosophical conversations, you shouldn't limit yourself, because you limit yourself, not nature. So man would never take to the air or sink to the seabed. By the way, about the prediction, I think I'll give you conclusive proof that you can predict quite far ahead relatively soon, but let's not be in a hurry. We'll all make it :o)
I don't believe it, but I'll be glad to know I was wrong :-)
But back to our philosophical conversations, you shouldn't limit yourself, because you limit yourself, not nature. So man would never take to the air or sink to the seabed. By the way, about the prediction, I think I'll give you conclusive proof that you can predict quite far ahead relatively soon, but let's not be in a hurry. We'll all make it :o)
I don't believe it, but I'll be glad to know I was wrong :-)
It's not faith, it's knowledge:o)
Sergey, who is this Burg?
Serious guy, invented another prediction method. In MathCAD his algorithm is implemented as predict() and burg() functions. You can look it up in the help. Berg is probably correct, but more often it is called burg. I even received some kind of prize for this method.
As you can see, it does a pretty good job of predicting, but I need to define the parameters. So here Seryoga, first handle my little brother :o)))))))) Just kidding of course :o))))
PS: That would be a description of this method to find ... Maybe North Wind knows where to look? :о)))
It is important, as it seems to me, and therefore I am writing a separate post. Seryoga, you are dealing with NS, try to train a neural network to forecast MA (it's possible), and using the forecast MA you may easily restore the future price series. The same adders :o)
What do you think I've been doing all this time? That's right! I've been studying the predictive properties of the underlying NS. It's a dead end. The thing is that constructing MAs, you essentially integrate the initial BP without adding anything new there. It is clear that the analysis of MA by NS or even by the devil, nothing new that is in the initial BP can give us. Therefore the prognosis series is sure to crumble as the event horizon approaches. It is already a medical fact. You can improve it noticeably if we require n-fold differentiability of MA. Then it is possible to break through the event horizon, though it looks like pulling a splinter through a splinter. The same effect can be achieved in a simpler way.
As for Burg's algorithm, whatever it is, it will definitely lose to NS, which is based on Kolmogorov's theorem on optimal approximation of a differentiable function on a d-dimensional cube. So I don't even bother with the other one anymore.
What do you think I've been doing all this time? That's right! I've been studying the predictive properties of the underlying NS. It's a dead end. The thing is that constructing MAs, you essentially integrate the initial BP without adding anything new there. It is clear that the analysis of MA by NS or even by the devil, nothing new that is in the initial BP can give us. Therefore the prognosis series is sure to crumble as the event horizon approaches. It is already a medical fact. You can improve it noticeably if we require n-fold differentiability of MA. Then it is possible to break through the event horizon but it is like pulling a splinter through a splinter. The same effect can be achieved in a simpler way.
As for Burg's algorithm, whatever it may be, it will by definition lose to NS, which is based on Kolmogorov's theorem on optimal approximation of a differentiable function on a d-dimensional cube. So I don't even bother with the other one anymore.
Wait, let's go through it sequentially. Do you think the maximum MA forecast horizon regardless of the method chosen (NS, Burg or whatever) is how many counts (or bars)? One or is there a difference?
Mashka has a forecast horizon of more than one bar! This value depends on the smoothness of the series constructed by the LPF, which in turn depends on the averaging window and the multiplicity of differentiation in the averaging algorithm. For conventional MAs you can put a single multiple of differentiation, for EMAs the second derivative is continuous, so it allows for a farther horizon.
Further, when I ask you to forecast one bar ahead, you should be aware of the fact that you are forecasting 1 bar + the value of the group delay of your MA.
Further, when I ask you to provide a forecast one bar ahead, you should be aware that for you it will result in a 1 bar forecast + the group lag value of your MA.
What does lag have to do with it? Neither Burg's method, NS nor your method knows anything about MA lag. For these methods it's just BP and nothing more. And then if you accurately calculate the future value of MA, you will also accurately calculate the future price value. What does lag have to do with it??????