Trading Strategies Based On Digital Filters - page 56
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
NOXA CSSA results
Here they are, pretty similar to MESA. Lets wait for Goertzel results.
Krzysztof
So the idea is:
The degree is like a multiplier of the length value.
The lenght value is how much bars the indicator uses for smooth the data, more bars, more smooth, and the indicator becomes less adaptative.
Is correct this?Yes, the more bars, the less adaptive the indicator.
No, the degree is a very sensible setting, as it is the one and only real parameter of MESA.
The relationship between degree and length is that degree must be less than length, otherwise the math underneath MESA fails.
A good thumb rule is that for currencies you keep degree at 150 and length at 200 at the least.
As you shorten length you must lower the degree and keep it under 75% od length.
There is not a general rule. It is useful to study (and I would like that the most curious of you extensively do it) different combinations of length and degree. One way is to have a signal with a well known spectrum and change settings to see what happens.
Hereafter you can see the spectrum of the last 512 bars of the signal posted by Krzysztof as measured in different conditions.
The first picture is with default setting for degree (150). You can see the spectrum as I have used to set the espert advisor that gave the above trading result. I can see peaks at 20 and 50 (to see the peak at 100 we need a much longer window), so I have set 15 for minimum fatl period, 45 for minimum satl period and it was the only fitting (I need to tell my indicators in which windows to estimate best peaks, it is an implicit action when we look at DFG spectrum).
The second and third pictures are relative to values of 100 and 50 for degree parameters. You can see that 50 gives a very untrue spectrum for the lower frequencies.
If we apply a detrending (linear) to this particular signal, things get better, as it is possible to have a good reprentation of the spectrum even with 100 (fig.4) and an acceptable one for 50(fig.5).
You have to think in terms of bars. How many bars is your timeseries from 2008.07.01 till 2009.01.01? Obviously it depends on timeframe. For a daily timeframe it should be something like 120-140 bars (which is few). For a H4 it should be something around 480-500, wich is fine. Put the number of bars in 'length' parameter.
ok
, other variant..
0 bar is current time (for example 2009.03.15 23:00)
i want make spectr
from 2008.31.12 23:00 - is 1140 bar
till 2008.07.01 23:00 - is 4220 bar
now tell me which parametr i need set for it calculation ?
ps. calculation on H1 timeframe
ok
0 bar is current time (for example 2009.03.15 23:00)
i want make spectr
from 2008.31.12 23:00 - is 1140 bar
till 2008.07.01 23:00 - is 4220 bar
now tell me which parametr i need set for it calculation ?
ps. calculation on H1 timeframeHi keekkenen,
you have to put 1140 in 'backwardBars' parameter and
(4220-1140)=3080 in 'length' parameter.
detrending
Yes, the more bars, the less adaptive the indicator.
No, the degree is a very sensible setting, as it is the one and only real parameter of MESA.
The relationship between degree and length is that degree must be less than length, otherwise the math underneath MESA fails.
A good thumb rule is that for currencies you keep degree at 150 and length at 200 at the least.
As you shorten length you must lower the degree and keep it under 75% od length.
There is not a general rule. It is useful to study (and I would like that the most curious of you extensively do it) different combinations of length and degree. One way is to have a signal with a well known spectrum and change settings to see what happens.
Hereafter you can see the spectrum of the last 512 bars of the signal posted by Krzysztof as measured in different conditions.
The first picture is with default setting for degree (150). You can see the spectrum as I have used to set the espert advisor that gave the above trading result. I can see peaks at 20 and 50 (to see the peak at 100 we need a much longer window), so I have set 15 for minimum fatl period, 45 for minimum satl period and it was the only fitting (I need to tell my indicators in which windows to estimate best peaks, it is an implicit action when we look at DFG spectrum).
The second and third pictures are relative to values of 100 and 50 for degree parameters. You can see that 50 gives a very untrue spectrum for the lower frequencies.
If we apply a detrending (linear) to this particular signal, things get better, as it is possible to have a good reprentation of the spectrum even with 100 (fig.4) and an acceptable one for 50(fig.5).And money wise ?? Did detrend help ?? I made some simulations in MATLAB and for Goertzel/FFT seems to be a key to do it.
Regarding CSSA results. I think detrending and denosing is already done for this but I will double check manually if it is possible to get better results by adjusting depth groups of eigenvectors, otherwise there is no room for improvement there.
I think I will create more advanced signals otherwise I feel that comments from post 549 can apply here a bit.
Krzysztof
Simba !!!!
You asked me not to share your indicator but not info about your group and your results.
As you said at the beginning we were just engaged not married. After just one week I realised that none of you has enough competence for cooperation as you were not able to answer basic functional questions.
Just answer simple questions
are you an engineer ??
are you able to read code ??
do you have any expirence from Research and Developement of any sort ??
Because if the answers are NO there is a big mismatch here and years of trading and zylions of posts will not help here.
My impact was very usefull - i pointed to faults in your indicator and your methodology.
If you don't want to post the results than don't post, all this secrecy is just funny for me, hiding method shown by Meyers in 2002....
KrzysztofThanks for your kind comments.
The answers are NO,YES and YES...I am not an engineer, just a trader ,it just happens to be that I hold a degree in Economics and a MBA by one of the top ten european business schools,but,IMO this,as being an engineer is irrelevant for trading,what is relevant is being able to conceptualize and adapt the tool to the task.
Thinking that being an engineer is necessary to use digital filters is like thinking that prerequisite is necessary to use a Formula One,and last time I checked,neither Fernando Alonso,nor Kimi Raikkonen nor Lewis Hamilton held that degree...but I would bet a castle against a house that the three of them have a conceptual grasp of the inner workings of their machines,in terms of driving results, that no engineer can get,unless he expends thousands of hours driving them.
There is an excellent book by Malcolm Gladwell,just published a few weeks ago..OUTLIERS,basically he studies those people that are several sigma deviations from the "achievement norm" in their respective endeavours(like Bill Gates,Mozart,Kasparov...)and his conclusion is extremely interesting...what really gave them the edge was a combination of luck(being the right age at the right time or being there when their field of knowledge expanded exponentially),PLUS family and society support and,KEY FACTOR, getting 10 thousand hours of expertise faster than their competitors...So,in regards to digital filters what is important is not being able to write a full textbook on them,but just for how long have you been trading with them..of course you need a minimal conceptual grasp,but to know the differences between a FIR and an IIR or what is zero padding and why MESA works better on high SNR series,you just need a modicum of intelligence and enough interest.
By the way,Bill Gates wasn`t an engineer either when he founded Microsoft and he knew a few things about coding
.
If you think the above is disgressing then you haven`t realized that ANY successful trader is approximately a 2 sigma deviation from the norm
Thanks for being so kind to respect our agreement,even if you think it is funny.
Regards
Simba
Hi Richcap,
According to what Simba said, do you have some denoising on your indicators? I remember you told us at the beginning that there was no denoising, but I saw a lot of posts about denoising and I don't know if now there is some kind of denoising in your indicator.
I don't see anything when I put your indicators on a chart. Only the spectrum and cutoff frequencies indicators are all fine. Am I missing some library?
What are the parameter resolution, write-N-peaks, WaveA, WaveL, Filter Period, Filter mode (what happens if we choose NonLagMA) ? for the spectrum indicator? For minPeriod, what is the Nyquist frequency?
Cheers