a trading strategy based on Elliott Wave Theory - 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
https://c.mql5.com/mql4/forum/2006/06/sverkaM30.zip
Ok. So I do not need to bother with images. I delete the previous post ;).
Good luck and good trends.
https://c.mql5.com/mql4/forum/2006/06/sverkaM30.zip
Does it mean that the channel is calculated on one timeframe (by SCO<=SCO2/3) and Hurst is calculated on another timeframe?
I just assumed not much different. I take the minimum number of bars for sampling as 45 (30 for 2/3) and look for channels satisfying the criterion. Quite often such a channel appears to be equal to the very 45 bars that indicates impossibility to draw the channel from the current bar. In this case, I was planning to insert the sliding indicator deep into the history to find such a channel. For example, the channel on the watches that I posted yesterday has already broken today and is trying to adjust to the last 45 bars - that shows the unsuitability of that channel (in other words - fitting that method in this case).
And with Hearst - here we have a pointy-tip and dumb-tip situation. For some reason unknown to me the algorithm of calculation is substituted. I admit that it is in fact a new method of calculating the new criterion, but not the Hurst index. That is, I do not deny that this method works, but I cannot yet understand the physical (or mathematical) meaning of its reading. That is, the relationship between the oscillation relative to some zero line and the maximum absolute channel of that oscillation, which takes into account the accumulated error of displacement over N measurements.
Forget about the formula itself for a moment. Just do the followingu There is the RMS of the errors, calculated relative to the regression line, which goes back and forth a bit when calculated by Vladislava's method (regression on the previous sample, not including the calculated bar). There's also the overall spread of the ALL price sample of the Hi-Low. Take and analyse the ratio between these values. If you have an approximately equal ratio, then you can say that the channel may have been picked at random and will disappear in the very near future. If the ratio between these values is high then it is said that the channel is not random and will continue in the future. I think a certain analogy can be drawn here between the Hearst ratio and the coefficient of determination (Bulashev) if it makes more sense to you. That is, the higher the ratio, the less likely it is that the channel is solidly in error.
Forget about the formula itself for a moment. Just do the following. There is an error RMS calculated relative to the regression line, which goes back and forth a bit when calculated using Vladislava's methodology (regression on the previous sample, which does not include the calculated bar). There's also the overall spread of the ALL price sample of the Hi-Low. Take and analyse the ratio between these values. If you have an approximately equal ratio, then you can say that the channel may have been picked at random and will disappear in the very near future. If the ratio between these values is high then it is said that the channel is not random and will continue in the future. I think a certain analogy can be drawn here between the Hearst ratio and the coefficient of determination (Bulashev) if it makes more sense to you. That is, the greater the ratio the less likely the channel is to be solidly in error.
That's right. I originally wrote that I was considering the R\S statistic, which is also commonly referred to as the Hurst ratio. In this ratio, S is RMS and R is the sample spread. For horizontal channels it is unambiguous; for sloping channels there are several ways to calculate the spread. The general idea is the same as for the Hurst index - to get an estimation of the degree of determinacy (local persistence, if in terms of the Hurst index).
Good luck and good luck with trends.
According to Peters the Hurst index uses the RMS({Log(Close[i]/Close[i+1]}) (i is the number of bar in MT)
It is also possible to use the RMS({Close[i]-Close[i+1]}).
You use, as Solandr explained to us, RMS({Close[i]-Approx[i]}), where Approx[i] is the forecast by approximation of all bars from the selected bar.
The difference of successive Close (the logarithm of the ratio is also suitable) is the same value that serves as the basis for cumulating the spread.
But the value of Close[i]-Approx[i] does not form the basis of the accumulated spread; it represents the regression prediction error. That is, the ratio of the spread to the RMS of this value should indicate the quality of the approximation.
However, the accumulation of prediction error by regressions is formed by another quantity, namely (Close[i]-Approx[i]) - (Close[i+1]-Approx[i+1]) which, imho, will give us RMS of original series reduced by "prediction ability" of approximations. And then, imho, we should take the error spread, not the spread of the original price series.
Then the use of exactly these RMS values and margins for R/S statistics allows to estimate the quality of the price series with the regression-excluded trend, and comparison with similar values for the original price series, respectively, allows to estimate the quality of the approximation.
Is there an error in this reasoning? Can the resulting comparison be applied to the problem you have set? Why?
Thank you in advance.
I.e. you throw it on the chart - it draws everything you need, but you can't grab the channel with the mouse and drag it.
Then Ctrl-B -> LR -> Properties , change one of the dates, ok, close
After that, everything falls into place.
Bild pre194.
The regression was drawn using OBJ_TREND. I didn't use the regular regression.
However, how does this script locks up the computer! And for a long time I couldn't figure out where resources go in the terminal, I had to go through all the windows.