Author's dialogue. Alexander Smirnov. - page 12
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
While Mr. Smirnov is thinking, I allow myself to go back a little bit.
By the way, a linear regression indicator (no channels; just a prediction of the next point along a straight line drawn through some number of previous LWMAs) is simply a linear combination of two dashes with the same periods:
LRMA = 3*LWMA - 2*MA
I think I will post this result to the Code Base, so that there would be no illusions about the fundamental difference between linear regression and wipes. But the proof has to be found or remembered...
The proof would be interesting. I think there is a difference (although I seriously doubt it now that you're telling me so). Do I take linear regression for 100 bars and MA for 100 bars and they will match in one bullet?
Once upon a time, when I experimented with LR, I also invented LRMA. Probably almost everyone went through it, as well as ZigZag. As I never was into wipes, I took a look, noted sensitivity and small lag for myself, and abandoned it. And now, when I saw this Mathemat'ik ratio, I couldn't believe it.
And it turns out that Mathemat is right. The relation LRMA = 3*LWMA - 2*MA is indeed true and can be proved quite easily. Only LRMA is not the prediction of the next point, but the LR value at the last (Nth) point, where N is the period of all three mashes.
For the proof it is only necessary to choose correctly the origin X in the regression Y=A*X+B, namely so that in the sliding window X takes values [1,2,...,N]. This can always be done, as Y values of the regression do not depend on the starting point of X variable. And then just add formulas for calculation of constants A and B by ANC to the regression equation. It should be taken into account that LWMA is the convolution of vectors X and Y with the appropriate normalizing factor and MA is the average of Y.
Thus, this relation is valid only because in LWMA linear weighting is performed with coefficients representing the sequence of numbers of a natural series, a very special case of linear weighting. If the coefficients in the LWMA implement a linear function but are not such a series, then the relation will not hold either.
Isn't there an SSA crawler algorithm in MT4? I can give you the link http://www.gistatgroup.com/gus/.Only this algorithm overdraws. And we need to invent some trick so it won't redraw. I think it is very promising.
Here is for example JMA and SSA with a period of 50. But I have CSSA based on SSA but not redrawing. Very fast. I recommend this algorithm .....
Isn't there an SSA crawler algorithm in MT4? I can give you the link http://www.gistatgroup.com/gus/.Only this algorithm overdraws. And we need to invent some trick so it won't redraw. I think it is very promising.
Where should I drop the dll or maybe the indicator doesn't work?
Well that's a bit off, in my opinion.....
You may not have noticed my 1st post in this thread. I would like to suggest that you again, at least post pictures. Where Jurik filters together with your filter, and test signals are applied (preferably several pictures which show all properties). Then at least you will have a visual evaluation. As a scientist you should know methods of quantitative evaluation, maybe I missed something but I did not see them in "VS" №01(75) 2006. Comparison of Djuric (and not begrudgingly his) with your algorithm.
Alexander, you can't do this. You came to the forum with questions. Let me remind you of them.
Your answers to these questions are important to me:
1. Whose algorithm is better: mine or Djurica's? How much better?
2. Do you have Djurica's algorithm?
3. How do they differ?
You were given a link to Juric's algorithm. There is a man who bought this algorithm for money and is willing to help you answer the questions you asked. But we are not magicians, we cannot compare the unknown, because we do not have your algorithm (indicator). And the answers to the questions of how and what you think are ignored by you.
To help YOU, we need to define the criterion of how to determine who is better. Suppose one indicator smoothes better, the second one lags less. Which indicator is better? We can argue about it till the second coming if we don't decide on the indicators and criterion. And the article contains not 2 indicators, but at least 4 (and some of them are not clear, especially how to calculate them).
At least do the following (since you keep your know-how and do not give it to anyone). Take a simple MA, and compare your indicator with it. Calculate and show in numbers how much better your indicator of a simple MA is (show what you claim in the article in words - in formulas and numbers).
Example layout
Post here an array of numbers by which you compare + figure. And the forum will help you - post the same calculations on the same data array and compare your results with their calculations and favorite indicators (Djuric also think to appear).
And your attacks on the members of this forum are ridiculous. You provide references and read them and say that we are "talking trash" here. All right, let them be, but you are a man and you keep your word. You said your indicator is better. Prove it in numbers and formulas (the article contains only words). You have to take responsibility for your "talk" :-). Give a comparison with the MA. See above for a design example.
Z.U. I hope this is a specific question or something needs to be clarified in the question ?
No... Definitely not the one. Definitely the wrong one... And ours is a little taller, too... ....
I must be doing something wrong. Decided to double-check. Here are the two indicators together. They don't seem to coincide at any point.
The straight line on MNC will always redraw, but LRMA doesn't seem to.
An ISC straight line will always be redrawn, but LRMA does not seem to be.
LRMA, which is actually plotted by MNA (not 3*LWMA - 2*MA), is the value of linear regression on the current bar, when the regression is plotted on N bars, including the current bar. It turns out that the current bar is the Nth bar in the sliding window, i.e. the last one. Therefore, although the regression line always changes its position, but only the last point is always taken from it for the indicator and therefore LRMA is not redrawn.
Yurixx, thank you very much for the unexpected support and valuable clarification. Yes, of course, when I started looking through my notes, I was convinced that it was exactly like that. I had forgotten, though - it's been over 2.5 years... There's something else left - about higher order regressions; it's all similar.
No, thank you. I, in my naivety, still believed that I had invented something original and of higher quality than traditional mashups. But it turns out to be just a linear combination of them. You learn for a long time, as the great Lenin bequeathed. :-)))