a trading strategy based on Elliott Wave Theory - page 77

 
It was my mistake after all.

. Now I'm definitely done for the day.
 
Naturally, I immediately tried to test this assumption in practice and share the results in this link https://c.mql5.com/mql4/forum/2006/07/min_poten_energy.zip

What do we see in these charts? We can see the linear regression channel plotted using the method of minimum RMS in a series, which I use in my Expert Advisor at the current moment. On a separate graph, we draw a graph of the algebraic (i.e. I carry out summation of gradients taking into account their signs (+/-)) sum of gradients of ball potential with respect to the parabola line constructed for the current sample. For simplicity of understanding I can say that absolutely the same approach is used here which is used for determination of Hurst coefficient with the difference that instead of approximating linear regression channel a parabola (quadratic function) is taken, and also parabola itself is constructed for all current sample, including calculated bar for which the potential gradient is searched. Just in case, let me remind you that in order to find the Hurst coefficient we took a sample that does not include the current bar.
The blue vertical line indicates the local minimum for the algebraic sum of the potential gradients for the channel starting at this very bar. If you look carefully at the EURUSD and USDCHF charts, you can see that the local lows exactly (+/-1 bar) coincide with the local maximums/minimums of price. I attribute the error of 1 bar to the peculiarities of sample preparation. I choose (O+H+L+C)/4 bar values as a sample. Probably, in this case we should selectively choose either High or Low that would definitely eliminate an error of determining the channel with the minimum potential energy. And then the channels in the series will coincide exactly with those shown in the picture shown earlier by Vladislav for USDCHF. Well, I will try to implement this assumption in my EA. I think it is really close to the method of channel selection used by Vladislav and shared by him.



I have also decided to become a copy artist :)

 
...And then the channels in the series will coincide exactly with the channels shown in the picture that Vladislav cited earlier for the USDCHF...


I guess it's a guess because if you apply the 2/3SCO>SCO condition you cannot get the image as shown by Vladislav. But if we refuse from it and take into account condition of not exceeding the last third of confidence interval of the channel plot by 2/3 of sample + selection in each group with minimum of skewness then we get quite similar results.

And in general it seems to me we are trying to get what Vladislav showed not much not the method he described.

Vladislav 26.06.06 10:50



Jhonny 16.06.06 16:48
....
1) Level of importance of channel selection criteria is the same (i.e. you find an optimal combination of these criteria), or there is a consecutive selection from more important to less important criteria....


No it is not - each one has its own weighting coefficient.



That is channels are not selected sequentially like I did, first we looked at 2/3 that didn't fall out of last 1/3, then we select those channels that have STR2/3>SCR (about this criterion, Solandr, please tell me in what post Vladislav mentioned it, because I couldn't find it today), then from these groups we select by minimum STR. Then for each channel weights are taken, which probably is a product or sum of values obtained RMS, Hurst, PE functional minimum (and others may be what I missed), then best of them are selected (though there are a lot of questions because Vladislav many times said that he selects those that satisfy in extreme way (for each criterion) and this means completely different :( ) then maybe something similar will work out.

But to be honest I thought my question about the level of significance of criteria Vladislav didn't quite get it right as it wasn't quite correctly asked. The significance level is different in the two given possibilities (and not the same as I said for the first case). In the first one, we consider all channels but with a combination of their criteria accounted for with different weighting coefficients, and in the second one, we just select from more important to less important (as I do now). In short I'll deal with what Solandr wrote in the last post...

ZS Hearst in the figure do not look function has written a long time ago, probably a lot of errors, and therefore the values there may be wrong, and even a glitch that popped up with two identical channels.
 
Tried entering RMS(2/3) > RMS and it turns out that this almost never happens at all. Either a bug in the code (can't find it yet), or a consequence of not randomly distributing the outliers.
 
Well I have seen it quite often.
There are 2 pictures on this page "trading strategy based on Elliott Wave Theory" where groups of channels were selected with this condition, i.e. the min SCO condition was turned off
 
Yes, there was a mistake. Still, I don't like the condition of RMS(2/3) > RMS yet. Below is my current picture, without it. As far as I understand, in principle it coincides with the pictures of other participants. As the saying goes, an extra check won't hurt.
 
I messed around with the objects a bit. I took OBJ_FIBOCHANNEL as an arbitrary one. I managed to put it and then read 5 double numbers, but found limitations for OBJPROP_SCALE (two decimal places, but 7 in front) and for OBJPROP_ANGLE (only one decimal place and two in front, it does not accept minus).
         pr1 = -9678912.312345; pr2 = 91.3; ObjectSet(obj_name, OBJPROP_SCALE, pr1); ObjectSet(obj_name, OBJPROP_ANGLE, pr2); pr1 = ObjectGet(obj_name, OBJPROP_PRICE1);
          pr2 = ObjectGet(obj_name, OBJPROP_PRICE2); pr3 = ObjectGet(obj_name, OBJPROP_PRICE3); pr4 = ObjectGet(obj_name, OBJPROP_SCALE);
          pr5 = ObjectGet(obj_name, OBJPROP_ANGLE); Print("Prop1 = ",DoubleToStr(pr1,10)," Prop2 = ",DoubleToStr(pr2,10),"  Prop3 = ",DoubleToStr(pr3,10)," Prop4 = ",DoubleToStr(pr4,10)," Prop5 = ",DoubleToStr(pr5,10));


        Prop1 = 1.27149825 Prop2 = 1.27632842 Prop3 = 0.00084162 Prop4 = -9678912.31000000 Prop5 = 91.30000000
 
solandr
If possible, post your picture on the Eura Clock. Both the channels themselves and the probabilities are of interest (I have a normal distribution) - I would like to compare.

 
For Candid

 
I'll compare myself for the company as well :)
I'm not really looking for channels with 45 bar but 100 bar, I think they break too fast, short channels do not all run in a row but in 5 so selected ones have multiples of 5 bar.