Trading Strategies Based On Digital Filters - page 39
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
peak 0.22
Yes, most likely shoothing. FFT for non smoothed gaus here
damiani against non smothed gauss noise
no change, a lot of buy signals.....
Don't ask me why it shows like this. Obviously smooth of gauss noise changed
results. For clear gauss noise only whittle estimator shown exact 0.5 however others are close.
I will try those wavelet indicators from this site but I think you agree that the most important is to know when we play roullette and when we really trading
/KrzysztofHi,
Interesting...raw gauss noise as a H=0.5 so absolutely random,while smoothed gaussian noise has a H=0.9XXX..So,practically a persistent trend...that is why Damiani`s volatimeter gave good signals with the smoothed noise and,yes you are right,it gave "random" signals where there should be none when plotted against the raw noise..Additional conclusion should be to take care with the smoothing,since it can transform a random series into a persistent one..I don`t know why..do you have any idea?
Well,we always play roulette since there is an element of chance in every trade ,I would say that we have to know if we are the casino or the sucker and play only when we have an edge...so,yes,we agree on that....we have to find,ideally,periods of low noise stationarity ,then trade them.
Sidequestion...from the 7 methods you used is Whittler a preferred choice? I had the idea that R/S range analysis was the more robust choice,but your example with gaussian noise made me wonder.
Regards
Simba
reply
Sorry I didn't reply yesterday but i just went to bed.
I think the way to look into those results is to take the median value of all
(so the worst for us) and ingore the rest. Next days I will investigate this deeper.
Gauss was smoothed so FIR filter response is in the result.....
Damiani is implemented as diffterence betwen ATR-STDev and this formula is not able to distingush between random noise and real signal. So it will work
like everything else....sometimes....
Did you try Goerzel against my GOLD files. I just wonder if it will find what it suppouse find in real time. During next days I will try to generate non stationary data than we can play.....
Krzysztof
gold1-600 bars of gauss noise
Sorry I didn't reply yesterday but i just went to bed.
I think the way to look into those results is to take the median value of all
(so the worst for us) and ingore the rest. Next days I will investigate this deeper.
Gauss was smoothed so FIR filter response is in the result.....
Damiani is implemented as diffterence betwen ATR-STDev and this formula is not able to distingush between random noise and real signal. So it will work
like everything else....sometimes....
Did you try Goerzel against my GOLD files. I just wonder if it will find what it suppouse find in real time. During next days I will try to generate non stationary data than we can play.....
KrzysztofHi,
I didn`t apply Goertzel real time(will explain why in a few lines),I started by applying synthetic vix,standard settings,not even cycle adjusted,to gold1 file,which is,or should be 600 bars of raw gaussian noise..Result:I was able to predict approximately 80% of the tops(I could do the same with bottoms,just a question of using the inverted syntvix)by just entering when syntvix rebounded on the 0 line...See attached...I didn`t chose the period to show off,I just open a chart and applied the syntvix,I am fairly sure that your complete file will get a prediction percentage between 70 to 85%....if you can generate a similar file with 3k observations,I AM fairly sure I would get similar results,so theory is good in theory but in practice,practice is better,predicting the turn point with near 80% accuracy and reward higher than risk means that we should start trading Gauss noise Futures
,please check your file,IMO it exhibits a very clear cyclic behaviour and would say that it is stationary...now,on Goertzel.
Sorry,I don`t know how to import your file-pure-into mt4 terminal,I resorted to importing it with the continuous gold contract equivalent timeframe,so,the first 600 bars in the chart were your gold1 ,then I had next(with a terrible gap,obviously)the original Gold continuous contract...this allows me to use syntvix but not Goertzel since Goertzel just uses 3*maxperiod data from end point...so,if you explain me how to import your hst files as standalone(to the terminal,not to the DFG),I will do the analysis.
Regards
Simba
Gaussian noise 1 and 2
I did a SSA smooth of gold1 data,then used MESA on DFG and it shows a very clear cyclic peak at 20(real cycle IMO)..attached "gaussian noise 1"..then I just did the Mesa analysis on your gold1 file and it shows a small peak at 14(faulty cycle IMO),see attached gauss noise 2...
Check the time range of samples analyzed,they are the same(1 bar difference) from 7th of Oct at 08:00 to 24th of October at 01:00,they are the same files,one SSA smoothing applied to your raw gaussian noise,the other just raw gaussian noise.
I am fairly sure that any kind of serious smoothing +MESA will uncover the real cycle,if any,behind the apparently random fluctuations...That is why syntethic vix with standard period=22(very similar to single cycle period=20)works so well to "trade" your raw gauss noise file..the file has an embedded cyclic structure.
Regards
Simba
gaussian noise imput
Hi,
Just returned from my tennis...
https://www.mql5.com/en/forum/173071/page26
This is FFT of raw gaussian noise (GOLD M1) and there is no cyclic components for sure and that's the input. I think that any type of transformation of this signal like SSA smoth, MA smooth etc can introduce
new components which can be read later by indicators as cyclic components
but it does not change fact that input was random. (it was confirmed by change of values of Hurst exponent after smoothing of gauss noise)
Regarding importing. I just delete .hst files than replace with mine and work
offline or with trading simulator. Otherwise if MT is online those files will be updated. The best is to switch MT chart to line display not a candle sticks.
Sigview is a shareware, you can download it and generate files yourself, beside gaussian noise you can generate white and pink noise also. It exports XY file than it is just enough to make export .csv file from MT to excel, replace data columns with Y from Sigview and import again. Just be carefull to use proper delimeter during this operations. MT must be offline and .HST
files must be deleted during import. Than when you close MT it will generate .HST files just with your data.
I think the way forward is to implement the latest Ehler S/N ratio indicator
for MT and check what it can do
Traders Tips - November 2008
Unfortunatelly there is no code for MT there.
Recently he is using filter banks to extract dominant cycle, this method is described in 'skinning the cat' document ===> if we can not extract the
cycle and we dont have a trend than we have random/noisy data. But most likelt we will be abe to extract something.....maybe any genius MT coder
can help
other possibility to investigate and design some filter which will filter noise. Those filters are mentioned in Codebreaker thread in FF. For this I believe the best platform for all those researches is mathlab, it has different
toolboxes and it's easy make data manipulations there.
Krzysztof
Regarding importing. I just delete .hst files than replace with mine and work
offline or with trading simulator. Otherwise if MT is online those files will be updated.You can simply open your HST via File > Open Offline
Cycles in gaussian noise
"This is FFT of raw gaussian noise (GOLD M1) and there is no cyclic components for sure and that's the input. I think that any type of transformation of this signal like SSA smoth, MA smooth etc can introduce
new components which can be read later by indicators as cyclic components
but it does not change fact that input was random. (it was confirmed by change of values of Hurst exponent after smoothing of gauss noise)"
Ok,so we can trade gaussian noise by smoothing it?Because this is what I did,show you that by SSA smoothing plus applying a synthetic vix you can trade Gaussian noise ....if you still doubt it ,send me a 3k hst gaussian noise file,and I will do the same,smooth and filter+apply a volatility oscillator..
"Regarding importing. I just delete .hst files than replace with mine and work
offline or with trading simulator. Otherwise if MT is online those files will be updated. The best is to switch MT chart to line display not a candle sticks."
Thanks ,will try and post my results.
"Sigview is a shareware, you can download it and generate files yourself, beside gaussian noise you can generate white and pink noise also. It exports XY file than it is just enough to make export .csv file from MT to excel, replace data columns with Y from Sigview and import again. Just be carefull to use proper delimeter during this operations. MT must be offline and .HST
files must be deleted during import. Than when you close MT it will generate .HST files just with your data."
Thanks I worked with sigview a year ago or so,probably discarded prematurely,will retry.
"I think the way forward is to implement the latest Ehler S/N ratio indicator
for MT and check what it can do"
99%% agree with you+1%let`s try other S/N options too...first we need the coefficients,then w ecan implement them on mt4.
Best Regards
Simba
PLease show
You can simply open your HST via File > Open Offline
Thanks,I have been trying to do so,no appearance of gold1 hst...then I pasted and copied gold1 .hst on History files..when I try to access it trough File>Open Offline...I can`t.
Can you explain in detail how to do it?
Regards
Simba