Fractals, ZZ & the Pissing Dog | Page 14 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

ECN && STP && Scalping && Hedging - Broker Review - Oct 2009 24 replies

Alert on Fractals (MT) 2 replies

Zigzag & Fractals 10 replies

EA wanted for sound alert depending on fractals condition 2 replies

Fractals Edge Fx 3 replies

Attachments: Fractals, ZZ & the Pissing Dog
Exit Attachments

Fractals, ZZ & the Pissing Dog

  • #261
  • Jan 18, 2014 8:31am Jan 18, 2014 8:31am
Quoting emonts
{quote} Then we should need to know place and time
If you - aka the team - finally find the way to define clearly the turning points with your high maths skills, we will all meet in my place !
MT4/MT5 EAs/Indicators/Alerts coder
  • #262
  • Jan 18, 2014 8:44am Jan 18, 2014 8:44am
Quoting killerno
Hi, my Set's for 2011-2012: SET1: 16368 SET2: 7977 SET3: 2114 Carl, Coud you please review your probability doc file. The total length calculation is different in your doc compare how GG is calculating. (http://www.forexfactory.com/showthread.php?p=7206574#post7206574 ). Maybe we should go through the pdf and ask GG to confirm the steps. GG If you have time could you please post a newer 5min TF screenshot. Thanks for all! regards
killerno,

Here you go:

G.
Attached Image (click to enlarge)
Click to Enlarge Name: eurusd.m5.png Size: 33 KB
  • #263
  • Jan 18, 2014 9:00am Jan 18, 2014 9:00am
Quoting gg53
{quote} killerno, Here you go: G. {image}

Thank you!
  • #264
  • Jan 18, 2014 10:27am Jan 18, 2014 10:27am
Quoting killerno
Hi, my Set's for 2011-2012: SET1: 16368 SET2: 7977 SET3: 2114 Carl, Coud you please review your probability doc file. The total length calculation is different in your doc compare how GG is calculating.
killerno,
GG53 posted the document after I sent it to him to check.
He has verified the document.
It is how I am doing it. I don't know any other way.

Carl
  • #265
  • Jan 18, 2014 10:40am Jan 18, 2014 10:40am
Quoting crodzilla
{quote} killerno, GG53 posted the document after I sent it to him to check. He has verified the document. It it how I am doing it. I don't know any other way. Carl
Hi Carl,

I understand it, and your pdf is suberb, but the total length probability calculation is different regarding to Post 204. You compare the actual length -3 intervall occurance to actual length +3 intervall and GG is calculating like that e.g. MAX(p7, p8, p9, p10, p11, p12, p13) > (p14).
That's why I wrote that maybe better to go through the steps and GG can confirm it again, but it was just an idea.

regards
  • #266
  • Jan 18, 2014 11:54am Jan 18, 2014 11:54am
Quoting burnssss
we have the pattern [1 0] with length 10; length---------------occurs 7----------------------10 8----------------------7 9----------------------6 10--------------------5 11--------------------5 12--------------------4 13--------------------4 14--------------------3 maximum P (7,8,9,10,11,12,13) = 10 P (7,8,9,10,11,12,13)> P (14) Probability in length accepted Is this correct?? Many thanks gg53
G can you clarify the total length algo please? From burnssss I understand that the following is correct.

we have the pattern [1 0] with length 10;
length---------------occurs
7----------------------10
8----------------------7
9----------------------6
10--------------------5
11--------------------5
12--------------------4
13--------------------4
14--------------------3

maximum P (7,8,9,10,11,12,13) = 10
P (7,8,9,10,11,12,13)> P (14) = true

If we also had
15--------------------3
16--------------------19

do 15 and 16 get included or ignored? Eg does the equation become
P (7,8,9,10,11,12,13)> P (14,15,16)

In this case the calculation would be false

Cheers
  • #267
  • Jan 19, 2014 4:18am Jan 19, 2014 4:18am
Quoting FerruFx
{quote} Beers should be on us ... for all you are doing here with your "team"
+1
  • #268
  • Jan 19, 2014 8:01am Jan 19, 2014 8:01am
Quoting FerruFx
{quote} If you - aka the team - finally find the way to define clearly the turning points with your high maths skills, we will all meet in my place !
FerruFx,

What's on the menu? ;-}

G.
  • #269
  • Jan 19, 2014 8:14am Jan 19, 2014 8:14am
Quoting beachloup
{quote} G can you clarify the total length algo please? From burnssss I understand that the following is correct. we have the pattern [1 0] with length 10; length---------------occurs 7----------------------10 8----------------------7 9----------------------6 10--------------------5 11--------------------5 12--------------------4 13--------------------4 14--------------------3 maximum P (7,8,9,10,11,12,13) = 10 P (7,8,9,10,11,12,13)> P (14) = true If we also had 15--------------------3 16--------------------19 do 15 and 16 get included or ignored?...
beachloup,

Burnsss example and my response are the way I'm currently doing it.
I understand your logic - but please try to understand mine.
In that example: We currently have actual length of 10.
When checking length 11, 12, 13 we already "looking" into the "future" (within +3 tolerance).
So we should check our current length (+tolerance) against 1 step into the further "future" (#14), otherwise there is no end to it...

The whole issue of tolerance is to compensate for very small bars (causing longer ZZ bar count) or very large bars (causing shorter ZZ bar count).
The (-3/+3) tolerance numbers came after long simulations, in early stages of development, using MATLAB.
If you find other tolerance numbers to be more accurate, or some other way to compensate for very small or large bars - please share.

G.
  • #270
  • Jan 19, 2014 8:19am Jan 19, 2014 8:19am
Quoting gg53
{quote} FerruFx, What's on the menu? ;-} G.
I'm in paradise ... so everything is on the menu
MT4/MT5 EAs/Indicators/Alerts coder
  • #271
  • Jan 19, 2014 8:30am Jan 19, 2014 8:30am
@ gg53,
I was running tests and I think because of the calculation of probability "next bar" still some bad signs. Can you explain how to perform the calculation of probability of "next bar"?
I conclude that the calculation is wrong "next bar" because the other steps are done properly and correctly predicted.
thank you very much
  • #272
  • Jan 19, 2014 9:57am Jan 19, 2014 9:57am
Quoting gg53
{quote} beachloup, Burnsss example and my response are the way I'm currently doing it. I understand your logic - but please try to understand mine. In that example: We currently have actual length of 10. When checking length 11, 12, 13 we already "looking" into the "future" (within +3 tolerance). So we should check our current length (+tolerance) against 1 step into the further "future" (#14), otherwise there is no end to it... The whole issue of tolerance is to compensate for very small bars (causing longer ZZ bar count) or very large bars (causing...
Thanks, I just wanted to be sure my code is as close to yours as possible so I can try to replicate your results

I will carry on, at the moment I'm not seeing anything like your results so I will keep looking for the details I have missed or poorly implemented. Cheers
  • #273
  • Jan 20, 2014 1:14pm Jan 20, 2014 1:14pm
The SHORT signal on Friday close (post #263 (http://www.forexfactory.com/showthread.php?p=7221732#post7221732)) ended up in ~30 green pips on this week market open...
FerruFx, is everything ready for our arrival? ;-}

G.
Attached Image (click to enlarge)
Click to Enlarge Name: eurusd.m5_2.png Size: 33 KB
  • #274
  • Jan 20, 2014 4:36pm Jan 20, 2014 4:36pm
I finally finished the old thread and this one.
I have one question.
At the start the bar sequences were done for every bar, then every other bar, etc.
A later post only refered to doing every bar.
Which is correct?
I do have an issue with using every other bar or 3rd bar etc.
Unless these are handled differently than the every bar then how can the patterns be considered the same?

At some point in the reading Data Structure came to mind.
This is a lot like a chess opening book to determine the next best move based on past games.
This is much easier because the data is basically a binary tree or trinary tree as suggested by One Day.
An earlier post mentioed keeping the counts in an array that is really an array form of a binary tree.
When a new candle appears it is only necessary to update a very small portion of the resulting tree.
I hope the new build includes structures.
That should save on time.

I plan to code the problem using trees to hold the data for zigzag legs, fractals within the zigzag and bar sequences within the zigzag.
The next release of MT4 included pointers and I hope structures.
I can do this in MT4 now but it is much easier having pointers and structures.
Robert
  • #275
  • Jan 20, 2014 4:54pm Jan 20, 2014 4:54pm
@MrPip,
In the old thread EURUSDD mentions using NON-LINEAR segments... every other bar, every third bar, etc.

We don't do that.
All data is LINEAR segments.

Carl
  • #276
  • Jan 20, 2014 4:59pm Jan 20, 2014 4:59pm
Quoting gg53
{quote} killerno, Here you go: G. {image}
Hi GG, I've got my signals similar place then yours, but as soon as a long leg comes (after the 3rd down signal in your picure) I get my UP signal too early and after it keep failing. I dont know how to filter these early signals. MAybe the Handling Prediction errors what I am missing.

regards
  • #277
  • Jan 20, 2014 5:53pm Jan 20, 2014 5:53pm
Quoting gg53
FerruFx, is everything ready for our arrival? ;-} G. {image}
MT4/MT5 EAs/Indicators/Alerts coder
  • #278
  • Jan 20, 2014 9:00pm Jan 20, 2014 9:00pm
Quoting beachloup
{quote} I'm glad I asked the question now. I think my SET1 should be accurate because as emonts shows there is only a fixed number of combinations available when building substrings 4-13. If the numbers are wrong then I have made a mistake in my code (possible) or not understood the concept to create SET1 (likely). To create SET1 I followed the method in post 22, I have put a screenshot of my C# code that is fully commented so if anybody can spot the error then please let me know. I can explain the differences in SET2 and SET3 but first I want to...
You are sequencing strings of length 4-13, but are not sequencing multiple length 4 strings, multiple length 5s, etc. Instead of 6 (below) think of 13 as the total string length you are trying to break up into pieces.

Quote
[We have a string of 6 items, {011001} - and we want to create substrings length 4-6.
0110 - 4 length (start from first item, count 4)
1100 - 4 length (start from second item, count 4)
1001 - 4 length (start from third item, count 4)
01100 - 5 length (start from first item, count 5)
11001 - 5 length (start from second item, count 5)
011001 - 6 length (start from first item, count 6)
  • #279
  • Jan 20, 2014 9:25pm Jan 20, 2014 9:25pm
Quoting gg53
The Logic: We start with a new ZigZag leg.Watching the incoming fractals in real time, we check against SET2 what are the probabilities of more incoming fractals for what we currently have?
I'm sure it has been asked before but I couldn't find the answer. What is the procedure when the fractal pattern is not found in SET2? Move on to the next step? Reduce fractal length and try again for a match? Other procedure?

Adding the fractal pattern to the database makes sense, but don't you have to wait for the end of the zig-zag first to get the complete fractal pattern? So the current prediction will generally be wrong when the fractal pattern is not found in SET2?

Also, is zig zag direction being stored in the database anywhere? Is zig zag direction linked in any way to the fractal patterns in SET2 or elsewhere?
  • #280
  • Edited 3:05am Jan 21, 2014 2:45am | Edited 3:05am
Hey FXEZ,

I am just finishing reading the old thread and this one. May be i am wrong, but i will try to answer :

Quoting FXEZ
What is the procedure when the fractal pattern is not found in SET2? Move on to the next step? Reduce fractal length and try again for a match? Other procedure?
I do calculation at every candle Closed. If there is fractal pattern that is not found in SET2, just wait for the next candle coming, then do the calculation again from the beginning .... (But if there is NO fractal at all at current zigzag leg, then we just look for a zigzag leg with just one "0" or "1" fractal at the end).
See also post #132 (http://www.forexfactory.com/showthread.php?p=7203554#post7203554)

Quoting FXEZ
Adding the fractal pattern to the database makes sense, but don't you have to wait for the end of the zig-zag first to get the complete fractal pattern?
I think we are predicting the end of zigzag, before it is really end. So, we are not waiting for the end of zigzag to do the calculation.

Quoting FXEZ
So the current prediction will generally be wrong when the fractal pattern is not found in SET2?
I think "yes". But in case our prediction is wrong, then we need the "adaptability" process .. See also post #237 (http://www.forexfactory.com/showthread.php?p=7218910#post7218910)

Quoting FXEZ
Also, is zig zag direction being stored in the database anywhere? Is zig zag direction linked in any way to the fractal patterns in SET2 or elsewhere?
After I downloaded the attachment at post #1, and put it to MT4, it created a csv file, which has zigzag direction in it. So, i assume that we need to put it in our database. See also post #130 (http://www.forexfactory.com/showthread.php?p=7202744#post7202744) & post #133 (http://www.forexfactory.com/showthread.php?p=7203588#post7203588)
1 trader viewing now