Dr.Fx corner - page 5
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 am skeptical of all indicators, that is why I do not use them in Expert Advisors. But it is interesting to compare your filter with something. As you can see, the break has not responded to your wish.
I have said and written before. You cannot compare the quality of filtering by opening/closing positions. TF has other characteristics. Here is a link 8 years ago we compared Kalman filter and Butterworth filter, topicstarter can take files (matcad I see he knows) and compare. Might get a good comparison
http://forum.mql4.com/ru/9321/page22#53541
Here is the criterion by which to (possibly) compare the quality of filtering. The phrase of a respected member of the 4th forum below
If to calculate the sum of squares of differences of initial series and series constructed by Kalman filters (FC) and Butterworth (FB), then the greatest approximation to initial BP is given by FC, see figure of differences:
The red line is FB minus the original series, the blue line is FC. Thus, FC copes splendidly with the task using a priori data about laws that describe behavior of the studied object.
It is a pity that many drawings have disappeared over time. Hopefully at least the codes have survived. So you can compare them all yourself, you don't need me for that...
This is a good question. But this effect is apparent. The local extrema are preserved there and there. However, when you look at the price, you think the high to the left is more important. When in fact, the high to the right is more important. It marks a change in the trend. The high on the left is just a local extreme.
For our trading purposes, what matters is minimal delay and minimal spikes when the pulse is fed. Things that are mutually exclusive ))
There is no mutual disconnection.
Feed a square wave to a broadband repeater and the output
you get a square wave with minimal delay and spikes :)
I have said and written before. You cannot compare the quality of filtering by opening/closing positions. TF has other characteristics. Here is a link 8 years ago we compared Kalman filter and Butterworth filter, topistarter can take files (matcad I see he knows) and compare. Might get a good comparison
http://forum.mql4.com/ru/9321/page22#53541
... However, I disagree that you can't assess the quality of the filter by opening/closing positions. Not only you can't, but it's the most qualitative method of evaluation...
I hope I understood correctly the notation in the Matcad file from Prival. Exactly, that V+a = model as is, YY = model + noise, VO = Kalman, shown all on the same graph + that filter which everywhere above is labelled Fs.
For clarity I do not show filtered input signal = Model&Noise, I only show Model, and compare Kalman and W1, ... W5, applying different smoothing similar to that shown above for USDCAD, W5 and there is denoted differently Fs on other charts.
the comments are curious. I see that W1-W4 keep the maximum position where Kalman is, and W5 is slightly to the right, which I think is more objective. someone will say it is a lag. but it should not appear, it mathematically has nowhere to appear, because the algorithm itself was very cleverly designed for that - we only see visually significant effect of transients, which become "longer" as smoothing increases, but transients and lag are of different nature. If the original (noisy) signal were a price chart, I would sell all the time. By testing doubts in the high area (just before the 300th bar). I would have ended up with all the maximum possible profit. On Kalman with those settings as they were in the Privat file I would not have been able to trade so successfully.
USDJPY may be getting a buy signal... I am closing one of the previously opened sell trades. Possibly I will close the second one soon and start buying.
On the euro yen is a strong sell. USDCAD is also a confident sell.
What remains is to calculate what we did withNeutron
We need tocalculate the sum of squares of differences of the original series and the series constructed by Kalman (FC), Butterworth (FB), and your filter. As far as I remember exactly this criterion you suggested in personal correspondence.
If that amount is less than the other two, then very great. Even great.
And as for using the quality criterion to filter open/close trades, that's not quite right. Having only one SMA (you can also use your curve) you can build dozens of trading systems, and all of them will trade differently, some TS will be better and some worse.
USDJPY may be getting a buy signal... I am closing one of the previously opened sell trades. Possibly I will close the second one soon and start buying.
On the euro yen is a strong sell. USDCAD is also a confident sell.