![MQL5 - Language of trade strategies built-in the MetaTrader 5 client terminal](https://c.mql5.com/i/registerlandings/logo-2.png)
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 far as averaging goes.
Not really. By (order of MA - 1)/2. Although this is only true up to the 1st phase jump.
I'm in the process of redesigning an indicator for FIR filters. When I'm finished, I'll have some pictures of the dummies. If it works, the amount of calculations is frightening...
For now, a couple of pictures for comparison.
This is how AFR and IFR of MA5 looks like.
And this is the filter I'm thinking of using instead.
I propose to divide the logical and random price components at the stage of indicator line calculation. The random price fluctuations should not affect the regular indicator line, while the regular price movements should not affect the random line. A full view of the price should be maintained, so random fluctuations should not be discarded, as is common practice.
It looks like you can't do without a picture.
The green line is the close price chart. Blue - regular component, unmistakably extrapolated in this case by approximately 32 bars. Red - random price oscillations, hangs around zero, theoretically with unlimited amplitude, but practice shows that it is possible to draw lines that intersect very seldom. Blue line + red line = green at any point. At the same time the blue one can be seen 32 bars ahead. Don't you think you would have to try very hard to lose on this? It is the simplest way of identifying and using the patterns, so primitive that it is not a shame to show it to everyone.
I don't see anything useful in separating downward movement and upward movement. It's just another graphic representation of the same thing, while making it more complicated and less understandable, at least to me. You can change the coordinate system and display the price as a spiral or as a completely black spherical thing in a vacuum, what will it change? It is still unclear what to do with it. I also do not see anything good in excluding time from the chart. When I open a trade, I somehow do not really care if I close it in an hour or a year. So the timing has to be at least that's why. There are other reasons too.
1-Why is the red line a random component, and the blue line a legitimate one (what is legitimate about its construction). and where is the line between deciding this is a random component, and that is legitimate
2What is the sufficient data depth required for extrapolation for this case for 32 bars ahead (how many bars of history do I need)
first you need 32 bars for ma 32 - 32 bars + how many more bars of the built ma need to reach the starting point
Not really. By (order of MA - 1)/2. Although this is only true up to the 1st phase jump.
waving directly at the price without an offset is a waste of time
Not really. On (order of MA - 1)/2. Although this is only true up to the 1st phase jump.
I'm in the process of redesigning the indicator for FIR filters. When I'm finished, I'll have some pictures of the dummies. If it works, the amount of calculations is frightening...
For now a couple of pictures for comparison.
This is how AFR and IFR of MA5 looks like.
And this is a filter, which I am thinking of using instead.
I was wondering, having a regular set of arrays, say from MA1 to MA64, can you get similar results in lengthening lines into the future as you have for 64 ma, for 32 ma, ....? or you have no chances to get the same anticipation as you do and it's senseless to try to use other filters?
1-doubtthat there will be preemption with analysis of additional pairs as well. Even if it is possible to get a couple of pips preemptive in price and a couple of farts in time, it certainly won't be worth the effort. I think it's possible to get a much more serious anticipation in another way, I'm working on it now. I'm using 17 pairs with dollar. Crosses additionally do not give anything useful, if again do not get into ticky-pookey, gold and everything else not related to currencies is also in my opinion better to be excluded.
2- Basically only EURUSD data is enough for EURUSD analysis. And if you only trade EURUSD, why do you need anything else?
4-EURUSD=Ieur/Iusd. Knowing Iusd, you can calculate all the other indices.
1 - You say that the anticipation occurs with the data of one pair and you have to try hard to use it, and all pairs with dollar are just for the convenience of currency analysis and for some other things,
i.e. all these preemptions will be different, i.e. the preempting effect will be different for some pairs, and it will happen differently for every pair of the cluster and sometimes the preempting moment for some of them will be earlier, so the preempting moment for the currency will be bigger (earlier) than for other pairs ... why do you use only pairs with dollar to get indexes of other currencies, and do not build indexes of other giants using all its pairs in the cluster for these currencies as well, the anticipation will be better?
By the way, thanks a lot thanks to you the fact that multicurrency analysis is the second one which has to be done for further calculations .... though even now my brain struggles with the idea that anticipation is possible only for the analysis of all pairs and that the anticipation of a pair using its own data is impossible - only a pop-up effect is possible which should be used in the multicurrency analysis... I thought all pairs in a cluster and not only should have something in common, they seem to move differently but I always tried to find a common coordinate system where all pairs move together at the same frequency for all (not even the movement of pairs but the movement of the pair effects)
3- and I would like very much to hear your answer to my previous post
It was interesting to know, if I have a normal set of wands, say from MA1 to MA64 period, is it possible to achieve the same result in extending lines into the future as you have for 64m, for 32m, ....? or is there no chance to get the same anticipation as yours with a mask and it's senseless to try it?
First we need to note that all known extrapolation methods allow extrapolating a curve no further than a quarter of the period of the highest-frequency harmonic of the curve. Further on, with polynomial extrapolation, the result usually flies away to + or - infinity. My range is about the same. The difference is that the result goes smoothly into the horizontal line(as far as the LPF is concerned). This means that the influence of the past is over. DIGITAL FILTERS DO NOT NEED TO BE EXTRAPOLATED BY ANY THIRD PARTY FUNCTIONS. They are excellent extrapolators in their own right, if you think about it a bit.
Now let's look at the picture again.
Red circled are 2 problems of MA5. Accordingly MA89 will have 44 problems. The frequencies that should not pass through the filter, pass through with a factor of up to 0.25. This is a lot. Also, the phase in this area jumps and cannot be compensated by anything. You can try to extrapolate. Under certain conditions (guess which ones) satisfactory results may be obtained.
1 - You say the anticipation is obtained using the data of one pair and on it you need to try to lose, and all pairs with dollar you need only for the convenience of currency analysis and for some other things,
i.e. all these preemptions will be different, i.e. the preempting effect will be different for some pairs, and it will happen differently for every pair of the cluster and sometimes the preempting moment for some of them will be earlier, so the preempting moment for the currency will be bigger (earlier) than for other pairs ... why do you use only pairs with the dollar to obtain indices of other currencies, and do not use all pairs in their cluster for these currencies as well, the anticipation will be better?
I only use LINEAR operations. If I decompose something and then add it back together, I get the same thing. So in principle there can't be anything earlier or bigger than that, apart from sampling and quantisation errors, spreads, arbitrage situations and all sorts of other crap.
If I calculate a 0.1% growth for EURUSD, and it declines by 0.7%, because at the same time the EURJPY was forecast to decline by 1.5%, this could happen. I can't be more accurate, it happens to others. You have to watch what you trade. Many seemingly perfect forecasts are worth 2 kopecks because they are made based on analysis of the wrong data.
To begin with, it should be noted that all known extrapolation methods allow to extrapolate the curve no further than a quarter of the period of the highest-frequency harmonic of the curve. The polynomial extrapolation usually takes the result to + or - infinity. My range is about the same. The difference is that the result goes smoothly into the horizontal line (as far as the LPF is concerned). This means that the influence of the past is over. DIGITAL FILTERS DO NOT NEED TO BE EXTRAPOLATED BY ANY THIRD PARTY FUNCTIONS. They are excellent extrapolators in their own right, if you think about it a bit.
Now let's look at the picture again.
The MA5 has 2 problems circled in red. The MA89 will have 44 problems accordingly. Frequencies that should not pass through the filter pass through with a factor of up to 0.25. This is a lot. Also, the phase in this area jumps and cannot be compensated by anything. You can try to extrapolate. Under certain conditions (you can guess which ones) we may get satisfactory results.