Trading Strategies Based On Digital Filters - page 41
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
Wavelet denosing
Here is the example of possibility of denoisig using matlab and simulink.
Pictures are self explonatory.
Krzysztof
trading gauss noise with mean reversion
Here is a definition
The values of the Hurst exponent range between 0 and 1. A value of 0.5 indicates a true random walk (a Brownian time series).
In a random walk there is no correlation between any element and a future element. A Hurst exponent value H, 0.5 < H < 1 indicates "persistent behavior"
(e.g., a positive autocorrelation). If there is an increase from time step ti-1 to ti there will probably be an increase from ti to ti+1.
The same is true of decreases, where a decrease will tend to follow a decrease. A Hurst exponet value 0 < H < 0.5 will exist for a time series with
"anti-persistent behavior" (or negative autocorrelation). Here an increase will tend to be followed by a decrease. Or a decrease will be followed
by an increase. This behavior is sometimes called "mean reversion".
from this site
Estimating the Hurst Exponent
and picture below from pdf attached. So we have to be very carefull trying to trade gauss noise with mean reversion. It must have hurst exponent < 0.5,
otherwise it won't mean revert. I think picutre confirms this. So without good well tested indicator we can be f.....up.
http://www.columbia.edu/~ad3217/fbm/thesis.pdf
So my gauss noise smoothed has a H aroud 1 so it should not mean revert
but I'm not sure if it is still gauss noise after smoothing.
Krzysztof
Stone
Yes..but,how to use it?We can carry as many stones as we want...still no solution
..It doesn`t work for trading.
Hogen, a Chinese Zen teacher, lived alone in a small temple in the country. One day four traveling monks appeared and asked if they might make a fire in his yard to warm themselves.
While they were building the fire, Hogen heard them arguing about subjectivity and objectivity. He joined them and said: "There is a big stone. Do you consider it to be inside or outside your mind?"
One of the monks replied: "From the Buddhist viewpoint everything is an objectification of mind, so I would say that the stone is inside my mind."
"Your head must feel very heavy," observed Hogen, "if you are carrying around a stone like that in your mind."
Fractal index indi test
I put the reference data for Brownian Motion here
https://www.mql5.com/en/forum/178285
With this data it is possible to test one avaialble indicator for MT4 which calculates Fractal index, I think it is called LT_FDI2. From my side I will try to
make an indicator in MATLAB based on function wbfmestiwhich measures Hurs exponent to see if it is usuable in real time. It calculates H pretty well based on Brownian series generated in MATLAB
Krzysztof
Hurst Exponent
I put the reference data for Brownian Motion here
https://www.mql5.com/en/forum/178285
With this data it is possible to test one avaialble indicator for MT4 which calculates Fractal index, I think it is called LT_FDI2. From my side I will try to
make an indicator in MATLAB based on function wbfmestiwhich measures Hurs exponent to see if it is usuable in real time. It calculates H pretty well based on Brownian series generated in MATLAB
KrzysztofKrzysztof,
1-I tried the LT_FDI2(100 periods) on your raw gaussian noise series(Gold1)..end result is that it gave a result ranging from 1.9 to 2.0...so,no autocorrelation at all,just the opposite(Hurst exponent near zero)....IMO it should have given a result nearing 1.5(H=0.5)
2-I tried it on your gold15 file,2 cycles 100 and 20 periods ,and LT_FDI2gave a result between 1 and 1.03(Hurst exponent near 1),so,it was indicating a trend...IMO,it should have given a result nearing 2(H=0),since the chart is clearly antipersistent.
Probably the mean reversion of Gaussian noise has distorted the results.
Regards
Simba
Hurst extimation and Synapse
https://www.mql5.com/en/forum/173071
Average H from selfis was 0.41 for gauss noise so it seems that indicator is not acurate, it should show like 1.5 for gauss noise (Fractal index value)
I just made two plots now for GOLD15 (0105sincos), ACF function and all Hursts. Difficult to say how to interpret this because H values are varying between 3 (??) and 0.28. ACF seems be positive and negative but mean value
will be positive I think, but how to interpret Hursts is a key here.
I include raw data which was an input for GOLD files so you can read it by SELFIS.
Synapse Peltarion --- cool, tool very good and very good tutorials, I already made two of them. So did you make any models for Synapse for FOREX ??
I think the key is to feed NN with proper data for training. I think in case of Synapse the problem will be with interaction with MT also, it can export
MS NET dll but I'm not sure how MT should communicate with it. Do you know
maybe ?? In case of using MATLAB it's easy there are articles which describe it.
Some articles
Forecasting Financial Time-Series - MQL4 Articles
Price Forecasting Using Neural Networks - MQL4 Articles
and that one how they made money on it.
News - Automated Trading Championship 2008
Krzysztof
Hurst exponent and trading strategies
Hi,
Very quiet, no action here. So something to diccuss. I've got a new toy called Neuroshell and it has a few interesting indicators. Have a look. Three charts with fractal indiicators first, EURUSD, gauss noise and clear signal
Krzysztof
test strategies results
Stochastic test strategies result for above. Help for indicators attached.
Krzysztof
Krzysztof,
On your GOLD15 Stochastic test , how did you combined Hurst and FDI in your stochastic strategy?I mean
did you come to conclusion that GOLD15 was the most noise-free timeframe and applied stochastic strategy?İt is not quiet clear to me.
stochastic tests
Hi,
GOLD15 and GOLD1 those are artificial signals generated by me for test purposes. Stochastic strategy is a standard strategy build in NS and don't use fractal indicators, I just wanted to show the relation between Hurst exponent value for different series and efficiency of standard strategy. Strategy was optimized with genetic algorithm in all cases. For GOLD1 (gauss noise) H=0.4-0.5, for GOLD15 (no noise) H is around 1 and for EURUSD it is when you measure I guess. Simple conclusion is when H goes down in direction of 'Random walk' (H=0.5) you are making less and less money. Other explanation from DSP === S/N goes down and more errors during making trading decisions.
In the next post I will try to evaluate significiance of Hurst parameters and S/N ratio in comparision to other parameters in testing strategy using the same genetic alg. Have a look to enclosed PDFs also.
Krzysztof