a trading strategy based on Elliott Wave Theory - page 78

New comment
 
solandr<br / translate="no"> If possible, post your picture on the Eura Clock. Both the channels themselves and the probabilities are of interest (I have for normal distribution) - I would like to compare.

Выложил https://c.mql5.com/mql4/forum/2006/07/10_07_2006.zip
 
.
 
I have pretty much the same on H1 as on M30.
I looked at the performance of the algorithm on the last hundred bars. In the picture below the RMS of the selected channels is what the algorithm would give on a rising history, past right. You can see that channels with low RMS can be unstable.
 
<br / translate="no">Rosh Rosh 10.07.06 00:11


solandr
If possible, put your picture by Eura clock. Interested and channels themselves and the probabilities (I have for the normal distribution) - I would like to compare.
...


Rosh for a full comparison you would calculate the average probability, I would calculate it myself using channel lengths and probability, but it's not always visible what sign it has(Solandr uses Up and Down, I put + or - sign, while you have no such information, I think if channels are not exactly the same, but average probabilities are close, then we are moving in the same direction)
 
The previous picture was actually RMS 2/3 . And here is the RMS on a thousand bars. The axes are abracadabra - it's the matlab that's glitching, but the first picture should be clear.
 
<br/ translate="no"> Rosh for a full comparison you would calculate the average probability, otherwise I would calculate it myself using channel lengths and probability, but it is not clear what sign it has(Solandr uses Up and Down concepts, I put + or - sign, while you have no such information, I think if channels do not coincide completely, but average probabilities are close then we are moving in the same direction)


Yesterday I forgot about probabilities, today I have added them. The first value is counted as solandr' a, the second one by "log averaging" cxbnf- instead of channel length takes the logarithm of this length, i.e. yeso coefficients = MathLog(channel length).

My values can be interpreted as stochastic, the closer to 100% the more likely I am to sell, the closer to 0% the more likely I am to buy. I took this as an indicator for the future in order to calculate the average probability on the history.


 
I'm posting it now with an extended search period (while I have not solved the problem of automatic detection of the start of analysis, I saw that Solandr has a large channel took instead of 90 days 190 and it was also detected).


I have started to test my forex Expert Advisor and I'm getting a feeling that it may be difficult for me to use it.

I have started to test the Expert Advisor, it's too early to show anything yet, as positions are poorly managed (although Sondr's profit went straight away without management), i.e. i still have to bend and bend.
 
I'm posting it now with an extended search period (until I solved the problem of automatic detection of the start of analysis, I saw that Solandr a has a big channel took 190 days instead of 90 and it was detected too).



I should probably add the possibility to choose the type of price (and its spread) - I have a different channel length on the o'clock.



By the way, the size of depth does not matter for forced algorithms

2006.07.10 13:02:29 VG_Ch2 EURUSD,H1: Let's open channel 3
2006.07.10 13:02:29 VG_Ch2 EURUSD,H1: Let's open channel number 2
2006.07.10 13:02:29 VG_Ch2 EURUSD,H1: Display channel number 1
2006.07.10 13:02:29 VG_Ch2 EURUSD,H1: Display channel number 0
2006.07.10 13:02:29 VG_Ch2 EURUSD,H1: Time of the optimized algorithm 81 ms
2006.07.10 13:02:29 VG_Ch2 EURUSD,H1: a=-0.0001 b=1.281 lastBar1 firstBar=74 StDev=0.0019
2006.07.10 13:02:26 VG_Ch2 EURUSD,H1: initialized
 
<br / translate="no">.
By the way, with boosted algorithms the depth size does not matter



I was thinking something else at the time of the discussion, will have to read it because with 190 days on Celeron 2.2 (yes at 5 bar) takes about 3 seconds per hour tester.

For now I have decided the more accurate the better.
 
Lazy to make a script =)<br / translate="no"> But I posted the utility:
https://c.mql5.com/mql4/forum/2006/07/FXTEdit1.zip
https://c.mql5.com/mql4/forum/2006/07/FXTEdit2.zip

This utility doesn't work correctly for me. I uploaded the missing files to the system directory.
The utility runs, but when opening the fxt file it says it can't open the file. The utility loads some of the parameters in the window including the maximum lot size, which does ravel 50. But the utility doesn't allow to change anything. All in all it didn't help me.
1...717273747576777879808182838485...309
New comment
Reason: