Fractals, ZZ & the Pissing Dog | Page 27 | 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

  • #521
  • Mar 23, 2014 3:28pm Mar 23, 2014 3:28pm
Hello GG53....would it be correct to say that the entry shown on the explorer chart is where your algo is printing its arrow? Thanks in advance....X
  • #522
  • Mar 24, 2014 4:20pm Mar 24, 2014 4:20pm
@ gg53
Lately it is getting very good results. I agree with you on money management.
I continue to study the system and still have doubts, always the same step.
We always understand the calculation of probabilities, as comparison of occurrences. But this step is hiding something more.
For the calculation of probabilities, use your "bayes"? "Markov"? "Hipergeometric load"?.
Be the calculation of probabilities is the same in step 1 and 3.
So, I'd appreciate a lot a brief explanation on this particular question.
Thank you very much and very glad you get good results.
  • #523
  • Mar 25, 2014 2:30am Mar 25, 2014 2:30am
Quoting emonts
Hi GG, First of all I see the system starts running really well. My respect. I have a question about step 3 of the algo (next bar prediction). Is the resident time in this step 3 only a few bars, or can it also be much longer if we have a strong up or down trend. Your answer is appreciated very much. BR Huub
I think as long as not a new fractal appears.
  • #524
  • Mar 25, 2014 4:19am Mar 25, 2014 4:19am
Quoting Trader_X
Hello GG53....would it be correct to say that the entry shown on the explorer chart is where your algo is printing its arrow? Thanks in advance....X

Trader_X,

ForexGT_Arrow is part of the algo.
It should also match individual currency strength/weakness, volatility and pair activity - to signal accurate entry.

G.
  • #525
  • Mar 25, 2014 10:19am Mar 25, 2014 10:19am
@ gg53
regard to post 522, could you explain?
thank you very much
  • #526
  • Mar 25, 2014 2:25pm Mar 25, 2014 2:25pm
Quoting burnssss
@ gg53 regard to post 522, could you explain? thank you very much
burnssss,

Really sorry, but I don't understand your issue/question.
Probabilities are always calculated the same way in all steps.
There is no "bayes","Markov" or "Hipergeometric load" in my probabilitis calculations.

G.
  • #527
  • Mar 25, 2014 2:46pm Mar 25, 2014 2:46pm
@gg53
Can you please give an en détail example of your fractal prediction routine?
Something like

We have Fractalpattern 00 with length 16

00 has been counted 310 times from set2
00 with zz length 16 has been counted 2 times from set2
calculate -> 2 / 310 = 0,006

001
001 has been counted 42 times
001 with zz length 16 has been counted 1 time
calculate -> 1 / 42 = 0,02

000
000 has been counted 60 times
000 with zz length 16 has been counted 5 times
calculate -> 5 / 60 = 0,08

Whichever is highest wins, in our case 000 is our prediction
When divide by zero then break and wait for next fractal ?
  • #528
  • Mar 25, 2014 4:04pm Mar 25, 2014 4:04pm
@ gg53
sorry for bad explain my problem.
I have doubt in calculating probabilities. I know you make the same calculations in each step.
But I have doubt in calculating probabilities.
Comparison of occurrences of substring in different lengths not seem to give good results.
So I think there is an associated calculation and want to know how to perform this calculation.
Thank you very much and I hope that your system has excellent results.
  • #529
  • Mar 25, 2014 7:27pm Mar 25, 2014 7:27pm
After recreating my indicator I am getting worse signals than I got months ago but I followed the instructions as close as possible and it made sense.

I used the exact 2 years of history data as gg to create my database and randomly choose January 10th to be my starting point. The first signal I generated is completely wrong so please can we go down the logic on 1 example please.


Attached Image (click to enlarge)
Click to Enlarge Name: i9alznyh.png Size: 50 KB


The blue arrow points up at the beginning of the leg. This was defined by me.

The algo is divided into 3 different parts 1. Fractals prediction (is the current pattern the correct pattern), 2.Total ZZ length Prediction (is the leg as long as it is supposed to be?) 3. “Next bar” in opposite direction prediction.

--------------------------------------------
Start/Assumptions

I talk about the point where I got the red down signal.

 

  1. We are in an Upwards move so the leg ends on a 1 fractal and a 0 bar.
  2. We do not have any fractals yet so we just assume a {1} fractal to be there. (Q1: Is that the correct way of doing it?)

​




------------

1. Fractals prediction

My current fractal sequence is {1} and I check this one against the next longer ones which are {10} and {11}

I just lockup how often these fractal pattern appear in my database.

Attached Image


{1} appears the most and we look for a pattern with a 1 at the end so we move on to the next step.

Q2 If {11} would have appeared more often we would just go to the next step with that pattern? Only in case of {10} winning we would terminate the logic right there.



------------

2.Total ZZ length Prediction

Current Sequence: {1}
Current # of Bars in leg: 3

Now I look up how often the current Sequence (1) appears within our current length with a +-3 margin of error. If length +4 appears more often the current pattern doesn't find and we wait.

Attached Image


The condition max(length 0-6) > length 7 is met as 52 > 32

Q3 any mistakes here? If 11 would have won in step 1 we would just look up the length of the 11 string?




------------

3. “Next bar” in opposite direction prediction

Take the past 12 bars and see if the next bar is going to be in the opposite direction (a 0 bar).

Attached Image


This is successful as well so I got my signal. Where are my mistakes?
  • #530
  • Edited 4:58pm Mar 26, 2014 1:09am | Edited 4:58pm
Hello everyone!

First off, thanks again to gg53, crodzilla, and flx23! I'm currently implementing the doc provided by flx23. Finally I'm beginning to understand how this works, but I still need to source or port the MetaTrader depth/deviation/span ZigZag indicator. So for now I'm relying on NinjaTraders stock ZigZag . Not ideal I know, but just testing for now.

In C#, using Lists and Linq for queries, I'm able to build (rebuild weekly or daily) the data sets in ~10 seconds with an old single-core 1.8Ghz AMD Sempron laptop CPU (yes I need an upgrade).

Quick question...

Considering the repainting nature of the ZigZag indicator, how is everyone determining their current ZigZag direction?

Right now I'm looking at the data in reverse until I encounter a turning point, but my latest turning point isn't confirmed until I encounter an opposite ("first repaint") turning point. This method will always make my current ZigZag direction either late or inaccurate.

Any advice is appreciated! Thanks!
  • #531
  • Edited 2:16am Mar 26, 2014 1:54am | Edited 2:16am
Quoting Kilian19
------------ 2.Total ZZ length Prediction Current Sequence: {1} Current # of Bars in leg: 3 Now I look up how often the current Sequence (1) appears within our current length with a +-3 margin of error. If length +4 appears more often the current pattern doesn't find and we wait. {image} The condition max(length 0-6) > length 7 is met as 52 > 32 Q3 any mistakes here? If 11 would have won in step 1 we would just look up the length of the 11 string? -
Killian, I read this as relating to info from page 13 of FractalDNA.pdf (Algorithm 14 Crozilla's method) that since you are on bar # 3 that you should sum up (0 + 11 + 24 + 27) and compare to (52 + 55 + 40+ 32) where the first group is smaller than the second group (so no new bar prediction).

If you use Algorithm 15 (page 14 G's method) then it would be (0+11+24+27+52+55+40)/7 vs 32 or 29.86 < 32 so no new bar prediction on bar #3 with either method.
www.forexfactory.com/attachment.php?attachmentid=1393461&stc=1&d=1395793185 (http://www.forexfactory.com/attachment.php?attachmentid=1393461&stc=1&d=1395793185)
  • #532
  • Mar 26, 2014 2:54am Mar 26, 2014 2:54am
Quoting Mr_You
Hello everyone! First off, thanks again to gg53 and flx23! I'm currently implementing the doc provided by flx23. Finally I'm beginning to understand how this works, but I still need to source or port the MetaTrader depth/deviation/span ZigZag indicator. So for now I'm relying on NinjaTraders stock ZigZag . Not ideal I know, but just testing for now. In C#, using Lists and Linq for queries, I'm able to build (rebuild weekly or daily) the data sets in ~10 seconds with an old single-core 1.8Ghz AMD Sempron laptop CPU (yes I need an upgrade). Quick...
mql4 code:

Inserted Code
while(n<2)
{
   if(zig>0) zag=zig;
   zig=iCustom(NULL, 0, "ZigZag", 0, zzlength);
   if(zig>0) n+=1;
   zzlength++;
}
truezzlength = zzlength-2;
if (zig<zag) { zzdirection = "up"; }
if (zig>zag) { zzdirection = "down"; }
  • #533
  • Edited 3:45am Mar 26, 2014 3:05am | Edited 3:45am
@Kilian19 & FXEZ

my algo is exactly same procedure as Kilian19.

gg53 said:
max(-3, -2, -1, 0, +1, +2, +3) > +4

so 55 > 32, and length calculation is correct.
no dividing by 7. or i am missing something?

please see post#203


edit:
thank you for sharing your method and database, Kilian19.
i am exactly the same as your algo. mine is:

1) fractals prediction
1 ----- 980
10 ----- 198
11 ----- 298

2) total length
1 ----- 39
2 ----- 81
3 ----- 86
4 ----- 170
5 ----- 162
6 ----- 133
7 ----- 97

3) bar prediction
1110 ----- 12578
1111 ----- 10714
01110 ----- 6730
01111 ----- 5848
101110 ----- 3295
101111 ----- 2839
1101110 ----- 1619
1101111 ----- 1320
01101110 ----- 873
01101111 ----- 705
001101110 ----- 476
001101111 ----- 0
0001101110 ----- 0
0001101111 ----- 0
00001101110 ----- 0
00001101111 ----- 0
000001101110 ----- 0
000001101111 ----- 0
0000001101110 ----- 0
0000001101111 ----- 0

therefore, my algo satisfy 1) and 2), but not 3). (so no down arrow would appear)

note that my data is 2011-2013(3 years) from dukascopy and i currently test 2014 for out of sample.
so the result may be different.

i noticed your test date is 2012/01/10, so maybe you should test out of sample date(if your database is 2011-2012).
also, have you filtered out the bar sequence by fractal pattern?
because my set3 suggest all "zero" for 10-13 strings.

thanks,
  • #534
  • Mar 26, 2014 3:14am Mar 26, 2014 3:14am
Quoting FXEZ
{quote} Killian, I read this as relating to info from page 13 of FractalDNA.pdf (Algorithm 14 Crozilla's method) that since you are on bar # 3 that you should sum up (0 + 11 + 24 + 27) and compare to (52 + 55 + 40+ 32) where the first group is smaller than the second group (so no new bar prediction). If you use Algorithm 15 (page 14 G's method) then it would be (0+11+24+27+52+55+40)/7 vs 32 or 29.86 < 32 so no new bar prediction on bar #3 with either method. www.forexfactory.com/attachment.php?attachmentid=1393461&stc=1&d=1395793185 (http://www.forexfactory.com/attachment.php?attachmentid=1393461&stc=1&d=1395793185)...

Hi just for my own understanding.
All countings are from the blue arrow onwards.
At the blue arrow we can have a leg endpoint and if, it must be a fractal.
So the fractal confirmation we have 2 bars later, the second bar.
In the past we discussed that for confirmation of a new leg we need a fractal.
This would be bar 8.
But atbar 4, the zigzag indicator will draw an up leg.
So from this point onwards we are sure we are in a new leg.

If wrong comments are appreciated.

BR.
Huub
It is not important how often you fall down, but how often you get up again
  • #535
  • Mar 26, 2014 3:39am Mar 26, 2014 3:39am
(Finished...) On our way to "double within a month" (or not...),
[see FF Trade-Explorer]

The algo uses the ForexGT_Arrow algo combined with CurrencyStrength scanner and Activity indies.
Loosing positions are hedged at reasonable levels with unique method to get out of that hedge.

First time running of this monster EA started by revealing a lot of bugs and the MT4 600 build issues.
After fixing bugs and adapting indies to the 600 build - things started to look Ok.

The parameter settings are of very high risk - but can easily changed to moderate or low risk trading.
The reason for that was to reveal its weak point and try to overcome them by algo improvements.

Target achieved. we doubled the account within a month (100.8% on March 2014 from previous month balance)
Now it's time for data analysis of the results and the improvements stage.

G.
  • #536
  • Mar 26, 2014 3:51am Mar 26, 2014 3:51am
Quoting gg53
(Finished...) Target achieved. we doubled the account within a month (100.8% on March 2014 from previous month balance)
Congrats ! All that work you've done is in this result ...
MT4/MT5 EAs/Indicators/Alerts coder
  • #537
  • Mar 26, 2014 3:58am Mar 26, 2014 3:58am
Quoting gg53
(Finished...) On our way to "double within a month" (or not...), [see FF Trade-Explorer] The algo uses the ForexGT_Arrow algo combined with CurrencyStrength scanner and Activity indies. Loosing positions are hedged at reasonable levels with unique method to get out of that hedge. First time running of this monster EA started by revealing a lot of bugs and the MT4 600 build issues. After fixing bugs and adapting indies to the 600 build - things started to look Ok. The parameter settings are of very high risk - but can easily changed to moderate or...
very hard-work behind this.
  • #538
  • Mar 26, 2014 4:02am Mar 26, 2014 4:02am
Quoting gg53
(Finished...) On our way to "double within a month" (or not...), [see FF Trade-Explorer] The algo uses the ForexGT_Arrow algo combined with CurrencyStrength scanner and Activity indies. Loosing positions are hedged at reasonable levels with unique method to get out of that hedge. First time running of this monster EA started by revealing a lot of bugs and the MT4 600 build issues. After fixing bugs and adapting indies to the 600 build - things started to look Ok. The parameter settings are of very high risk - but can easily changed to moderate or...

Hi GG,

I can only say Chapeau

BR.
It is not important how often you fall down, but how often you get up again
  • #539
  • Mar 26, 2014 4:28am Mar 26, 2014 4:28am
Quoting emonts
{quote} Hi just for my own understanding. All countings are from the blue arrow onwards. At the blue arrow we can have a leg endpoint and if, it must be a fractal. So the fractal confirmation we have 2 bars later, the second bar. In the past we discussed that for confirmation of a new leg we need a fractal. This would be bar 8. But atbar 4, the zigzag indicator will draw an up leg. So from this point onwards we are sure we are in a new leg. If wrong comments are appreciated. BR. Huub
Again just for my own understanding.

Suppose we have a fractal pattern of length 4.
The fourth fractal is confirmed 2 bars later.
Suppose we are in step 2 of the algo and have to check for length.
If we take the length distribution for a fractal pattern of 4, we are already 2 bars behind the mean length, for simplicity take gaussian distribution.
So always late, right side of the distribution.
We know by definition swing point should be a fractal.
So do we not need to take when we have a fractal pattern of 4, the length distribution of a fractal pattern of 5.

BR.
It is not important how often you fall down, but how often you get up again
  • #540
  • Mar 26, 2014 4:47am Mar 26, 2014 4:47am
Quoting gg53
(Finished...) On our way to "double within a month" (or not...), [see FF Trade-Explorer] ........ Target achieved. we doubled the account within a month (100.8% on March 2014 from previous month balance) Now it's time for data analysis of the results and the improvements stage. G.
G,

Congratulations!!!

Your labour of love has paid off. Well done.
0 traders viewing now