The ideal timeseries to trade - page 8
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
excellent, now if we could have this in mql4this would be awesome
Wavlets & Fractals
I belielve we should investigate this site
Wavelets and Signal Processing
There is a lot of sources for proper estimation of Hurst component
and Wavlets. It can be potential source of a new indicators.
Another site is John Conover: Home. It is a source of free Fractal tools for deep fractal analysis however installing this soft under linux is a pain.
Than potenitial indicator should be tested in MT against fictious random data
like a gause noise or Fractional Brownian Motion.
I already did some tests of digital filters FATL/SATL etc system. described at TSD and find out that that people are following semi blindly some system
without knowing under which conditions it is working. Perhaps I will publish those results under digital filters thread.
Krzysztof
Brownian Motion
As i promised in digital filters thread I generated 3 data series with Brownian motion process using Matlab. Series are for fractal index = 0.3, 0.5 and 0.8
and it can be as a reference signal to test MT4 indicators like a one included in this thread. They must be converted to .HST files with method described in digital filters thread.
Watching differen Brownian curves I realized that FOREX data is very similar to those trends so we should concentrate on analyzing them.
Krzysztof
0.5&0.8
Charts for fractal index 0.5 and 0.8
It's amazing how close these generated timeseries are to real fx timeseries.
That's why the literature on financial timeseries says they are fractional brownian motions.
But imagine that the literature is right, then all this is a big gambler's paradox... All forums, analysis, investment management,...
On the other hand, it's not because the series is random that that automatically means that you can't win from it, or is it?
what do you think?
randomness
They are sometimes random and sometimes not but always chaotic. This is the reason that technical analysis sometimes works and sometimes not.
You can try measure yourself how random FOREX is. Just enough to export
.csv files from MT4 (e.g. for EURUSD), read .csv by excel, extract closing prices and put to Hurst tool which I posted before. Tool measures Hurst exponent value. 0.5 means pure random.
Krzysztof
The Hurst Exponent Indicator
Does anyone here know how to read/use/trade using The Hurst Exponent Indicator ? (IVAR)
any news on the hurst indicator?
Hurst Indicator and Table Indicator
Attached is the Hurst Indicator as well as an indicator that calculate the Hurst value for different time frames and currencies displayed on a chart. If you display the indicator on a 1 minute chart it will update every minute etc.
Lyapunov exponent
Several weeks ago I used the software TISEAN for measuring the Lyapunov exponent for Forex EUR/USD. I made measurements for the 1 m, 5m and 15m time series.
I was greatly surprised. The Lyapunov exponent was not positive. In fact it was negative. What a surprise!
After I measurements, the 5 m. times series had the most negative Lyapunov exponent. The 15 m and 1m chart had equal characteristics.
Conclusion:
The 5m. chart was better than the 15m. and 1m. chart.
I am sorry I could not mount the pictures here, because I am a newcommer in this forum but check this:
forexforum.bg/viewtopic.php?f=17&t=668&start=20
There are pictures of the estimations of the Lyapunov exponent.
I am skeptical for he possibility to translate the algorithms into metatrader. A friend of mine has tried to code the algorithm of Rosentein and alia but gived up because it it too complicated. In fact there are a preliminary work to do to detrend the time series.
It is real fun how people are going in the same research path.
So if I can add:
1.Times series with low Lyapunov exponent (use the Tisean software works best under Linux, but since Ubuntu this has stopped to be a problem). It will tells how chaotic are the time series. (And I was surprised to discover the fact that sometimes it can be negative)
2. Hurst estimation: this is the second important parameter for long term dependencies.
3. We can check the estimation of Hurst with the technique of
randomized buckets methodology (this is included into the selfis software)
'How sensitive are the estimators to short-range
correlations? To address this question we employed
the randomized buckets methodology. Intuitively,
the estimations should not be affected after internal
randomization. Note that internal randomization
breaks the short-term correlations, while
preserving the long-term. On the contrary, external
randomization should significantly influence
estimations, since long-memory is distorted. Estimates
of the Hurst exponent of externally randomized
series should be close to 0.5 since long-range
dependence has been canceled.
Finally we can apply a preferred trading system you are comfortable with.
Whatever we consider working and we are comfortable with.
I think this is crucial stuff especially in the times of high frequency trading against us.
Reference:
iqnet.cz/dostal/CHA2.htm
Time Series and Chaos
For the newcomers in this domain please check this article for quick introduction right to the target. All you need to know for now.