Fractals, ZZ & the Pissing Dog | Page 24 | 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

  • #461
  • Feb 20, 2014 4:43pm Feb 20, 2014 4:43pm
Quoting burnssss
Hi gg53 I'm still scratching their heads and do not know how to use the information "actuall bar" and "next bar". Above all do not know the role that "actuall bar". I was confused by the post of page 19. "Relating to our "puzzle" - We also have 3 options: use the current pattern, use longer pattern ending with "0", use longer pattern ending with "1". We know our current pattern status and we know the other optional pattern probabilities. Try to apply the above logic to "next bar" and "fractal pattern longer". Use the math that is clearly explained."...

burnssss,

I didn't understand the first part.
The second part of comparing M5 & M15 - I didn't do the matching, comparison and/or linking the two for confirmation.

G.
  • #462
  • Feb 21, 2014 2:24pm Feb 21, 2014 2:24pm
[Quote = gg53; 7295414] {quote} burnssss, yo no entendía la primera parte. La segunda parte de la comparación de M5 y M15 - yo no hice el juego, la comparación y / o vinculación de los dos para la confirmación. G. [/ quote]

Very good gg53 ,
The first part of the previous post refers to using the information as " actuall bar" and "next bar".
I and most people who follow you for months , we calculated the odds of "actual bar" and "next bar" as follows :
ex :
we have the following string:
[1 0 0 1 0]
and want to know what type of candle is as follows:
[ 1 0 0 1 0 X ]
then we create substrings of various lengths ( one example is , the numbers are not real ) :
[ 0 1 0 0] ---------- 12000
[0 1 0 1 ] ----------- 13000
[0 0 1 0 0] -------- 5000
[0 0 1 0 1 ] -------- 6000
[1 0 0 1 0 0] ----- 1000
[ 1 0 0 1 0 1 ] ----- 2000
We note that we have increased likelihood of [1]
We do this to " actuall bar" and "next bar" , but when you get signal?
You could explain this process ?
Maybe you can get more precision and very little loss if you contrast the information from 2 TF , and also assume a stochastic process to verify points of inflexion .
I think it would be perfect.
Many thanks
  • #463
  • Feb 23, 2014 4:02pm Feb 23, 2014 4:02pm
Quoting Etecy
#property copyright "Copyright 2012, Deltabron - Paul Geirnaerdt" #property link "http://www.deltabron.nl" from Nanningbob thread {image} {file} {file} {file}
I am again not getting emails.
TMA repaints so I do not know if your currency strength indicator is accurate, especially in the past.
There are EAs that do not repaint and I will add code for them to see if that helps.
Robert
  • #464
  • Feb 23, 2014 4:33pm Feb 23, 2014 4:33pm
No comments about the Chess Opening idea. Maybe no chess players in the group.
Here is a better explanation of my approach.
Start at the oldest bar and move forward until a confirmed zigzag is found.
The zigzag bar will be the start of the tree. If a candle is up go right, if down go left. At each node of the tree data is stored as is done in the sql database but also counts pf how often a node is visited. This will give the counts of sequences. When a fractal appears that node will hold that information as well as various calculations and totals. When a new ZigZag is confirmed there is a return to the base of the tree to start again. When the structure is complete the tree is used to follow the branches of the new bars since the zigzag. At each node there will already be information useful for predictions.

BTW, the new build does not appear to have normal pointer types so must use a tree based on classes instead of a simple structure.
I also saw where arrays are passed by value. Have not checked about pass by reference. By value will not allow changes to the array and I know of at least 1 zigzag indicator that passes arrays to be modified in a function.

While I am at it has anyone tried this version of ZigZag? It uses 3 different sets of input settings.
Might give a variation of zigzag fractals within larger zigzag patterns. I am using fractals from the Mandlebrot concept, not the indicator.

Robert
Attached File(s)
File Type: mq4 3_Level_ZZ_Semafor.mq4   8 KB | 379 downloads
  • #465
  • Feb 24, 2014 8:53am Feb 24, 2014 8:53am
[Quote = MrPip; 7299172] No hay comentarios sobre la idea de apertura de ajedrez. Tal vez no hay jugadores de ajedrez en el grupo. He aquí una explicación mejor de mi enfoque. Comience en el bar más antiguo y avanzar hasta que se encuentre un zigzag confirmado. La barra de zigzag será el comienzo del árbol. Si una vela está a ir a la derecha, si ir por la izquierda. En cada nodo de la de datos de árbol se almacena como se hace en la base de datos de SQL pero también cuenta PF con qué frecuencia se visita un nodo. Esto le dará a los condes de secuencias. Cuando aparezca un fractal ese nodo tendrá esa información, así como diversos cálculos y totales. Cuando se confirma un nuevo ZigZag hay un retorno a la base del árbol que empezar de nuevo. Cuando la estructura es completa el árbol se utiliza para seguir las ramas de las nuevas barras desde el zigzag. En cada nodo ya habrá información útil para las predicciones. Por cierto, la nueva generación no parecen tener los tipos de puntero normales así deben utilizar un árbol basado en clases en lugar de una estructura simple. También vi donde arrays se pasan por valor. No he comprobado sobre pase por referencia. Por valor no permitirá cambios en la matriz y sé de al menos 1 indicador de zigzag que pasa arrays ser modificados en una función. Mientras estoy en ello ¿alguien ha probado esta versión de ZigZag? Utiliza 3 juegos diferentes de ajustes de entrada. Podría dar una variación de los fractales en zigzag dentro de patrones en zigzag más grandes. Estoy utilizando fractales del concepto Mandlebrot, no el indicador. Robert {archivo} [/ quote]

MrPip
options and combinations to get a complete system needed.
You could put a visual example?
It helps me understand.
thank you very much
  • #466
  • Feb 24, 2014 1:52pm Feb 24, 2014 1:52pm
Hello GG,

Can you please post some recents screenshots of the algo on the EURUSD in 5 and 15 minutes ?

In order to compare with our work.

If you can, it will help us a lot.

Thanks !
  • #467
  • Edited 11:26am Feb 25, 2014 10:06am | Edited 11:26am
Hi,

First time posting to Forex Factory, and indeed I feel privileged to do so in such an illustrious thread!

Thank you gg53 and others for your openness and willingness to assist others in this venture …

I have coded UltimateTruth to run in NinjaTrader (see screen shot below for today’s signals… times are EST).

I may have misunderstood some points and have a question about the “next bar” method (I really have been through this thread a number of times, so if I have missed the answer somewhere and it is actually staring me in the face, I apologize … I put it down to temporary episodes of coder’s tunnel vision… ).

My question:

 

  1. Do the “Next Bar Opposite” rules require the next bar actually to be opposite before a signal is accepted?
  2. Or is a signal accepted simply on the basis of the next bar probabilities (as outlined in the probabilities document) aligning as described?


Many thanks.

Attached Image(s) (click to enlarge)
Click to Enlarge Name: $EURUSD (5 Min) 25_02_2014 FF.jpg Size: 181 KB
Click to Enlarge Name: $EURUSD (5 Min) 25_02_2014 2.jpg Size: 171 KB
  • #468
  • Feb 25, 2014 11:55am Feb 25, 2014 11:55am
Quoting OtroTrader
Hi, First time posting to Forex Factory, and indeed I feel privileged to do so in such an illustrious thread! Thank you gg53 and others for your openness and willingness to assist others in this venture … I have coded UltimateTruth to run in NinjaTrader (see screen shot below for today’s signals… times are EST). I may have misunderstood some points and have a question about the “next bar” method (I really have been through this thread a number of times, so if I have missed the answer somewhere and it is actually staring me in the face, I apologize...
Nice progress OtroTrader. Are you handling the error predictions similar way than GG? (only in fractals)
thanks
  • #469
  • Feb 25, 2014 12:45pm Feb 25, 2014 12:45pm
Hi, killerno

Thanks for your reply.

My UltimateTruth indicator is still work-in-progress; there are still points I need to iron out (but I am getting close!).

And your question (how am I handling error predictions?) helpfully got me re-reading the adaptability and error prediction materials.

In terms of adaptability, I believe the proposed "adaptability methodology" is equivalent to ensuring that the database is up to date and includes the most recent zig zags, fractals and DNA. So I update the database every so often with all the data I have for the pair in question.

In terms of error prediction, after re-reading the materials I think the answer to the question I posed in post #469 is that yes, the signal must wait until there is a confirmed bar in the opposite direction, as that is the whole basis of the three door problem (i.e. being handed an extra piece of info about which door it isn't).

So, I think you have helped me answer my question. Yes, signal confirmation must wait for a door (no... sorry, a bar) in the right direction. Is that right?
  • #470
  • Feb 25, 2014 2:33pm Feb 25, 2014 2:33pm
Quoting OtroTrader
In terms of error prediction, after re-reading the materials I think the answer to the question I posed in post #469 is that yes, the signal must wait until there is a confirmed bar in the opposite direction, as that is the whole basis of the three door problem (i.e. being handed an extra piece of info about which door it isn't).
Good job with the indicator!

So, when you have a pattern and length match, you predict the next bar, if it is in the right direction, you wait for it to finish. If it turned out correct, do you predict the next for a signal or you paint the signal? I must say I've let it fire right after pattern and length matches, I don't know if it's the right thing to do. I remember at on point I did wait one bar between pattern/length match and next bar prediction, but it made it worse. But I didn't check the direction correctness either.

Can I ask you do you get those signals in the middle of ZZ's, which couldn't be more wrong?

Thanks!
C.
  • #471
  • Feb 25, 2014 5:39pm Feb 25, 2014 5:39pm
Quoting candero
Can I ask you do you get those signals in the middle of ZZ's, which couldn't be more wrong?
Thanks for the response, Candero.

They are points where there had been a pattern match, and then the length was correct, and next bar probabilities said next bar ... but the signal was not checking that a bar did actually form in that direction. That I have to change and sort out for the next iteration! ... Does this explanation make sense and answer your question?
  • #472
  • Feb 26, 2014 12:06am Feb 26, 2014 12:06am
Would anyone who gets good signals and bad ones as well, mind posting some screenshots of unprofitable signals? Everyone who's been working on this seems to get a mixture of good and bad, and it's looking likely that's as good as it gets. I would like to see if it's possible to avoid entering on the bad signals and enter only on the good ones.
  • #473
  • Feb 26, 2014 7:12pm Feb 26, 2014 7:12pm
gg53, Carl,

Guys I have been following this thread silently for a while now. I have been coding using Oracle DB. Your resources have been great.

I only have one issue, with everything that has been presented. It is the method for determining the higher probability ZigZag. I refer to the first section presented by Carl in his document. There is a specific method for length and next bar. ZigZag optimal choice is very unclear.

Carl uses "101" fractal as an example, and then goes to look at the next option "1010" and "1011". The fact that "101" is a subset of "1010" and "1011" will mean that it will ALWAYS have higher hits in the DB compared to the two parent fractals.
Could you please explain in some other detail why "101" would be a better choice than "1010" or "1011" and why? What exactly triggers this decision? I ask this because I am trying to code this. I have to make a list of decisions to code to make the choice.

I look forward to some descriptions and advice. I thank you in advance. By the way I have all this working with the new build 610. It has been a very long 6 weeks.
  • #474
  • Feb 27, 2014 10:14am Feb 27, 2014 10:14am
Quoting gg53
(Continued...) On our way to "double within a month" (or not...) Starting over the Trade Explorer. Too many significant bugs and lost days over the 600 build issues. G.
(Continued...) On our way to "double within a month" (or not...)
Most of the bugs fixed (I think...) and issues with build 600+ also fixed.
For a week now, the algo is working, continuing from last position and trying to recover.
Let's see... ;-}

G.
  • #475
  • Feb 27, 2014 2:31pm Feb 27, 2014 2:31pm
Hello GG53,

I have been following this thread (and the previous) for awhile now....I havent posted on account that most of this is over my head and I would just be cluttering the thread...but I must say that it some of the most brilliant work Ive seen come across FF in the 7 years Ive been hanging around here....truely remarkable !! A huge thank you to everyone whose been involved...GG53 of course I wont ask you for details on your algo...but like some of the other request it would be greatly appreciated if you could post a few screenshots of some charts ...just for comparison reasons....If you cant , thats ok too...Still really appreciate all youve given here...Best Regards, X
  • #476
  • Feb 27, 2014 2:35pm Feb 27, 2014 2:35pm
@ gg53
I'm happy to see you again.
Regarding the calculation of probabilities "Current bar" and "next bar", a binomial using?.
I'm a little lost in this step, I usually do not have very good signals due to failure in this step.
Long ago, you mentioned that the calculation of probabilities was the same fractal pattern and "next bar".
But I'm not getting good results.
You could guide us in this passage?. How to calculate the odds and how you use "actuall bar"?
Many thanks and I hope that the algorithm works perfectly.
  • #477
  • Feb 27, 2014 9:21pm Feb 27, 2014 9:21pm
burnssss
I think I can help you with that part. Are you referring to Carl's document section "Next Bar" probability? If so, it works like this:
Take the current string in a zigzag to be: 011001101
According to Carl, we need to look at he next bar possibility to be either: 0110011010 or 0110011011, the above with either a 1 or 0. If we break that string up into sub-strings of length 4 to 13 starting from the end. So the string may be much longer but we only look at what we have in the DB, so we use string from 4 to 13. So in a set the results may look like:
SYMBOL, BPATTERN, BPLENGTH, HITS, PERCENT_HITS
GBPJPY "1010" 4 19692 50.39 *
GBPJPY "1011" 4 19386 49.61
GBPJPY "11010" 5 9843 51 *
GBPJPY "11011" 5 9458 49
GBPJPY "011010" 6 5060 49.65
GBPJPY "011011" 6 5131 50.35 *
GBPJPY "0011011" 7 2544 50.48 *
GBPJPY "0011010" 7 2496 49.52
GBPJPY "10011011" 8 1333 50.55 *
GBPJPY "10011010" 8 1304 49.45
GBPJPY "110011010" 9 650 48.44
GBPJPY "110011011" 9 692 51.56 *
GBPJPY "0110011010" 10 333 48.26
GBPJPY "0110011011" 10 357 51.74 *

Sorry if the table does not line up.

Each length has an option of a zero or one. The one with the greater hits wins in each case. So add up the winning "1"s and the winning "0"s and compare. The most wins out as the most likely to occur.

In the above case: lengths 4 and 5 the 0s win - 2 wins; all others the 1s win - 5 wins. Therefor the most likely next bar is a "1". It wins 5 to 2.

I hope I have explained the right section?
  • #478
  • Edited 10:39am Feb 28, 2014 10:26am | Edited 10:39am
Quoting ricduc
Carl uses "101" fractal as an example, and then goes to look at the next option "1010" and "1011". The fact that "101" is a subset of "1010" and "1011" will mean that it will ALWAYS have higher hits in the DB compared to the two parent fractals. Could you please explain in some other detail why "101" would be a better choice than "1010" or "1011" and why? What exactly triggers this decision?
ricduc, you did not look at the rest of the examples in my document... please read it again.

According to my database at the time, and your logic of subsets of the larger fractal patterns will ALWAYS have more hits, is just not true.
We are using exact patterns when inserting in the database and not chopping them up into smaller patterns. That is why your logic is incorrect.

10 has 340 hits.
101 has 599 hits.
100 has 53 hits.

0101 has 77 hits.
01010 has 138 hits.
01011 has 8 hits.

That is how you will find out the the current (real-time) fractal pattern is probably longer.
So if the length is longer for the longer pattern, then use the longer pattern length to wait before going to "next bar."

Carl
  • #479
  • Feb 28, 2014 11:18am Feb 28, 2014 11:18am
Crodzilla & riduc
Crodzilla sorry, I was the cause of all this confusion.
I asked only for the step of "next bar". I do like riduc, but I get enough false entries. Keep the algorithm from the beginning of "the Ultimate Truth". gg53 obtained very good results without applying step 2 (probability of length). Applied only step1 (fractal pattern) and step3 (next bar) and furthermore, both steps in the same manner.
I think I'm still doing something wrong.
Many thanks
  • #480
  • Mar 2, 2014 5:46pm Mar 2, 2014 5:46pm
Carl,

I am not breaking set2 up, just looking at the fractal patterns. What I don't quite get, and I have read your document many times, is at what point do you start looking at a higher fractal pattern, when the lower fractal pattern has occurred many more times?
Let me expalin further. In your examples you use 101 occurring 509 times, and the current length is 18 bars. It has much highre probability than the next higher pattern 1011 or 1010. Let's say the zigzag leg is an up leg. So the 101 pattern may already be at the turning point, meaning that it has started to run in a down leg. The possibility is that it may contine to reach another high, pattern 1011 before moving down, or possibly even 10101 before moving down. Therefore at 101 you are either searching for a down bar, or looking to achieve a new high. This is what is unclear in the probility document, this is not well explained. What is the logic to tie together the sections zigzag, length and nextbar? What is the logic to say that because I am at 101, length 18, I have reached a zigzag high and that the elliot wave will move down?
I an not looking at being lectured too, I am asking because this information is lacking in the document. You just give examples, which I can quote as well with the data I have. Your examples leads me to more questions. They do not give a clear logic path. The last two sections are clear to me. It is easy to determine the optimal length and calculate nextbar. What I am asking is what exactly leads me to believe that I have reached the end of this fractal pattern? Obviously if 1011 appears then it is obvious that this may or may not be the end of the up zigzag.
Remember that this is a discussion and I am trying to understand your logic which is not clear to me, which is not a criticism.
1 trader viewing now