Fractals, ZZ & the Pissing Dog | Page 28 | 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

  • #541
  • Mar 26, 2014 5:45am Mar 26, 2014 5:45am
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, after all the flowers you got I have an other question.
Technically we see you vague at the horizon, as person of course you are very close.
Are you still willing to narrow the technical gap. You can get even more close.

Best regards,

Huub
It is not important how often you fall down, but how often you get up again
  • #542
  • Edited 7:12am Mar 26, 2014 5:48am | Edited 7:12am
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)...
Yes crodzila is it doing this way but it was never ever stated by gg that this is the way he is doing it. See here. http://www.forexfactory.com/showthread.php?p=7206574#post7206574

Quote
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,

My history is build from 2010 - 1.1.2012 and the generated signal is at January 10th 2012 so it is out of sample. I want to use 2012 to fill my database with 'adaptability signals' and 2013 to look at the signals generated after that.
Regards the filtering. I created a filterdataset and used it to filter my data. but now i look it up and '001101111' doesn't appear in it so there might be a bug in my code. I will hunt this one down now Thanks.
Attached Image (click to enlarge)
Click to Enlarge Name: fractals 3.png Size: 39 KB


EDIT: I filtered my set created my set 3 but in the logic checked it still against set 1 .



Quote
(Finished...) On our way to "double within a month

congratulations.
  • #543
  • Mar 26, 2014 9:37am Mar 26, 2014 9:37am
Quoting Kilian19
{quote} Yes crodzila is it doing this way but it was never ever stated by gg that this is the way he is doing it. See here. http://www.forexfactory.com/showthread.php?p=7206574#post7206574
@ Killian and @JimSterk, I guess I shouldn't post at 3AM. I mixed up the two methods and Killian and JimSterk are of course correct according to G's own statement. (As an aside, in thinking about "calculating the probabilities" a bit more I've come to the conclusion that calculating length probabilities is ripe for a hypothesis test with a confidence interval. Of course this is not what G has stated but his method isn't a proper probabilities calculation but more of a probabilities heuristic.)

Regarding the total length calculation, (using Killian's example) if you find that max(0,1,2,3,4,5,6) > 7, then doesn't this mean that you are estimating that the current length is correct with respect to a length 4 bars into the future? It seems this could potentially lead to indications of a predicted fractal turn being as many as 4 bars early according to the "probabilities" calculated, or am I missing something?

@G, congrats on reaching your goal. The end is much smoother than the beginning!
  • #544
  • Mar 26, 2014 10:15am Mar 26, 2014 10:15am
Congratulations gg53. I knew the results would be good.
thank you very much
  • #545
  • Mar 26, 2014 1:59pm Mar 26, 2014 1:59pm
Quoting FXEZ
{quote} Regarding the total length calculation, (using Killian's example) if you find that max(0,1,2,3,4,5,6) > 7, then doesn't this mean that you are estimating that the current length is correct with respect to a length 4 bars into the future? It seems this could potentially lead to indications of a predicted fractal turn being as many as 4 bars early according to the "probabilities" calculated, or am I missing something? @G, congrats on reaching your goal. The end is much smoother than the beginning!
That is why I liked my Gaussian method of finding length better, because it would identify the "cluster" of lengths among the set of lengths found.
In theory I think it should work better, but I haven't found that it is any more accurate than what G is doing, his algo against mine.

The max(0,1,2,3,4,5,6) > 7 does not make sense to me. As you pointed out, this method could be true for 13 bars.

Carl
  • #546
  • Mar 26, 2014 1:59pm Mar 26, 2014 1:59pm
Bravo GG,

Your results are awesome. You did it
  • #547
  • Mar 26, 2014 2:08pm Mar 26, 2014 2:08pm
I fixed the next bar calculation and played around with the parameters as to how many of the lengths have to agree upon getting a signal but the accuracy just doesn't want to improve by much. I really don't know what to do now as I have tried out so many adjustments without improving much. gg any remarks to my post #529?


GG's algo makes sense to me as legs tend to be longer or shorter than usual due to very large candles or very short candles. (he called it overshooting).
  • #548
  • Mar 26, 2014 4:21pm Mar 26, 2014 4:21pm
Congratulations GG53 !!!...Truely happy for you...you have inspired us all!!! thanks for all your hard work and patients...Best Regards, X
  • #549
  • Edited 7:28pm Mar 26, 2014 7:15pm | Edited 7:28pm
Guys,

To answer, directly and indirectly, many related questions & PM's:
At the end of each ZZ leg there is a fractal.
We start counting the first fractal AFTER that one (not including the fractal at the end of previous leg).
Our historical database INCLUDES the last fractal in its next ZZ fractal pattern.
So, our search for "next bar in opposite direction" should include one more fractal that didn't reveal itself in real-time.

Does that makes sense? Ahhha factor? ;-}

G.

EDIT:
To elaborate - if probabilities shows pattern of 4 fractals and you already got 4 fractals - you are too late by at least 3 bars...
  • #550
  • Mar 26, 2014 7:45pm Mar 26, 2014 7:45pm
Quoting gg53
At the end of each ZZ leg there is a fractal. We start counting the first fractal AFTER that one (not including the fractal at the end of previous leg). Our historical database INCLUDES the last fractal in its next ZZ fractal pattern. So, our search for "next bar in opposite direction" should include one more fractal that didn't reveal itself in real-time. Does that makes sense? Ahhha factor? ;-} G. EDIT: To elaborate - if probabilities shows pattern of 4 fractals and you already got 4 fractals - you are too late by at least 3 bars...
Yes that is the way I am doing this currently isn't it? At the beginning we do not have any fractals. but in that case what do we do?

What does the next bar in opposite direction has to do with fractals?
  • #551
  • Mar 26, 2014 7:59pm Mar 26, 2014 7:59pm
Quoting Kilian19
{quote} Yes that is the way I am doing this currently isn't it? At the beginning we do not have any fractals. but in that case what do we do? What does the next bar in opposite direction has to do with fractals?
Kilian19,

I don't know what you are doing - I didn't checked your code...

You start looking for "next bar..." AFTER you have satified probability of fractal pattern.
And as I said in previous post: if you got 4 fractals that match 4 fractals probability - you missed by at least 3 bars.

G.
  • #552
  • Mar 26, 2014 8:02pm Mar 26, 2014 8:02pm
Quoting gg53
{quote} Kilian19, I don't know what you are doing - I didn't checked your code... You start looking for "next bar..." AFTER you have satified probability of fractal pattern. And as I said in previous post: if you got 4 fractals that match 4 fractals probability - you missed by at least 3 bars. G.
Ah okay thanks
  • #553
  • Mar 27, 2014 3:53am Mar 27, 2014 3:53am
Hello GG,

Congrats for your results, impressive !

I want to point something out, i analyse your results, and i see that 95 % of your trades starts at the exactly same hour, so @ 7:00.

I can't beleive the algo give you signals at always the same hour, day after day.

Can you enlight us, there is something in the pudding like grandma say

Thanks !
  • #554
  • Mar 27, 2014 11:19am Mar 27, 2014 11:19am
Quoting LittleCaro
Hello GG, Congrats for your results, impressive ! I want to point something out, i analyse your results, and i see that 95 % of your trades starts at the exactly same hour, so @ 7:00. I can't beleive the algo give you signals at always the same hour, day after day. Can you enlight us, there is something in the pudding like grandma say Thanks !

LittleCaro,

This is because of the CurrencyStrength - trade signal must be on swing point M5/M15 (or few bars after it) AND confirmed by H1/H4 CurrencyStrength.
~07:00 is the last hour of Asian session before Europe Session (Franfurt open):
Trying to predict where the market will go when the Elephants finish their morning coffee...

G.
  • #555
  • Mar 27, 2014 11:31am Mar 27, 2014 11:31am
@ gg53
very good results.
I think we're all confused with the logic of the system, because we are doing the same calculations, and the results are not good.
1 - Pattern match fractal.
We have to look for the synchronization between our database and real time.
When you start counting, we ignore the last fractal previous zleg.
We compare occurrences (adding or removing a fractal of our database)
* Your perform probability calculations, or just compare occurrences?
2 - Length
Not the most important step, but it gives us an approximation of the end of the zleg.
Use Gauss.
3 - 'Next bar "
We take the last 12 bars and add the last bar, which we know the probability.
Earlier you said that we should add a bar that corresponds to the end of zzleg fractal.
We compare occurrences.
* You do probability calculations or just compare occurrences?.
Many thanks
  • #556
  • Mar 27, 2014 6:15pm Mar 27, 2014 6:15pm
Quoting burnssss
@ gg53 very good results. I think we're all confused with the logic of the system, because we are doing the same calculations, and the results are not good. 1 - Pattern match fractal. We have to look for the synchronization between our database and real time. When you start counting, we ignore the last fractal previous zleg. We compare occurrences (adding or removing a fractal of our database) * Your perform probability calculations, or just compare occurrences? 2 - Length Not the most important step, but it gives us an approximation of the end...


burnssss,

I just compare occurrences.

G.
  • #557
  • Mar 28, 2014 12:31pm Mar 28, 2014 12:31pm
Quoting gg53
{quote} LittleCaro, This is because of the CurrencyStrength - trade signal must be on swing point M5/M15 (or few bars after it) AND confirmed by H1/H4 CurrencyStrength. ~07:00 is the last hour of Asian session before Europe Session (Franfurt open): Trying to predict where the market will go when the Elephants finish their morning coffee... G.

Thank you for your explanation GG

Me too, i make my day in the morning 15 mins before London opens.

Have a great week-end !
  • #558
  • Mar 28, 2014 3:49pm Mar 28, 2014 3:49pm
Good article about probabilities with source code.
Pitty it's only in Russian and not translated. I used Google translate.
http://www.mql5.com/ru/articles/271

G.
  • #559
  • Mar 28, 2014 7:44pm Mar 28, 2014 7:44pm
Quoting gg53
Good article about probabilities with source code. Pitty it's only in Russian and not translated. I used Google translate. http://www.mql5.com/ru/articles/271 G.
Thanks for the link, G! Math boffins, I wonder do fractals and ZZs follow bell curve? Mandelbrot certainly made it clear that price doesn't.
  • #560
  • Mar 29, 2014 2:42am Mar 29, 2014 2:42am
Quoting gg53
Good article about probabilities with source code. Pitty it's only in Russian and not translated.
http://www.mql5.com/ru/articles/271 G.
english http://www.mql5.com/en/articles/271
1 trader viewing now