a trading strategy based on Elliott Wave Theory - page 81

New comment
 
Just in case you're not doing it right - "How to insert pictures in this forum (explainer)"
 
OK, maybe it's my hands, but it didn't work.
 
As would be expected - there is correlation between the currency pairs. On the Eura, the average probability is close to the New Zealander.

 
Made the output in the indicator of the two variants. At the same time, I got an idea.



Here just values, and here the difference




I think, during weekend (when markets close) it would be nice to compare StdDev calculations, in case someone has error in calculations. I can post it as an Excel file, with bar numbers and values on that bar.
 
I think at the weekend (when markets close), it would be a good idea to compare StdDev calculations with each other, in case someone has an error in the calculations. I can post it as an Excel file, with bar numbers and values on that bar.
Then you also need an identical way of selecting channels. As a matter of fact, my pictures of StdDev and StdDev23 have been present for some time :). About the stability zones: and outside of them consider that there is no channel? Those Matlab pictures show that despite the variability in characteristics, the channels have continuous trajectories and hence exist outside the stability zones as well. It seems to me that there are more and less good times for calculation of the channel characteristics. That is it is necessary to define somehow the point in the past where calculation gave the most adequate characteristics.
 
I want to ask a childish question. Using a conditional or normal distribution, we find the value of the confidence interval expressed in RMS. For example, we have found an interval with 90% probability, which means that the SV will be within this interval with 90% probability. In other words, being at the border of this range we can say that the price with 90% probability will go back inside and with 10% probability it will go further. If my reasoning is correct, it means that approaching the midline, 90% probability will become 0 and 10% probability will become 1, which is obviously not true as at the midline the probability is 50%. Where is my reasoning wrong?
 
2 Rosh and others

Don't you guys think you're getting too hung up on the RMS23>SCO condition. This is essentially just a convergence criterion, and convergence by itself can neither be a channel identification condition, nor a tipping point condition.
Moreover, the channel selection algorithm that has been investigated here for several pages assumes that recalculation is performed on every new bar. The result, as you have all seen, is that the channels float. So it turns out that you are applying RMS23>RMSCO criterion every time to other channels. But at the same time you compare them to each other. I think this is just an example of not very correct optimization. We all know that whichever piece of history we take, we can always find parameters with which even a bad EA will be profitable.
I believe the correct setting of the problem should be as follows. 1. Formation of channels using a separate stable algorithm. This means that a relatively small shift on the history should not cause the change of channels. 2. Selection of 3 or 4 channels that are related to significantly different services. RMS23>SCO can also be a selection criterion. But this condition alone is not sufficient. 3. Monitoring of fixed channels, including with the help of RMS23>RMS. Note, however, that the criteria are applied to a fixed channel and do not result in a realignment of that channel. In this case, a breach of the criteria leads to the conclusion that the channel is collapsing. If we tweak it all the time, however, we are thereby misleading ourselves. 4. Once destruction of a channel has been confirmed (reliably punctured, for example), a search for new channels takes place at that location. This search is repeated on every new bar until ALL channel criteria are satisfied.
Everything from the first point onwards.
IMHO
 
Approaching the midline, the 90% probability will become 0, and the 10% probability will become 1.


Exactly ! At the midline, the probability that the price moves even closer to the midline =0 (which I hope is obvious :)
And the probability that it will start to move away from the midline =1 (which I think is also obvious :))
 
I want to ask a childish question. Using a conditional or normal distribution, we find the value of the confidence interval expressed in RMS. For example, we have found an interval with 90% probability, which means that the SV will be within this interval with 90% probability. In other words, being at the border of this range we can say that the price with 90% probability will go back inside and with 10% probability it will go further. If my reasoning is correct, it means that approaching the midline, 90% probability will become 0 and 10% probability will become 1, which is obviously not true as at the midline the probability is 50%. Where is my reasoning wrong?

You are a bit confused about the concepts. You are operating with a confidence interval talking about the probabilities of price movement at its boundaries, using numbers in a straightforward way.
You're actually making a mistake there. When price is at the 90% confidence level, that means there is a 10% probability of price moving up or down. The probability is calculated based on the interval width from the central regression line. Thus, the probability of price movement back to the interval will equal 90%+10%/2=95%, and further at the rate of 100%-95%=5% correspondingly. I.e., the interval of 10% belongs to 2 equal parts of 5% at the top and bottom of the 90% confidence level, therefore, we should divide 10% by 2 to obtain the data on these parts of probability.
Thus, when the price is on the centre line, the probability is 0%+100%/2=50%, 100%-50%=50%, i.e. we obtain equal probability of upward and downward movements.
 
Thank you, that makes sense now.
1...747576777879808182838485868788...309
New comment
Reason: