This is the development community for the indicator that was discussed in great detail at:
http://www.forexfactory.com/showthread.php?t=464948
Please upload documents such as notes from this thread that you think are important and code samples that you have.
GitHub: https://github.com/PenguinTraders/FractalArrowPrediction
Instructions: http://penguintraders.com/groups/fractal-arrow-prediction/forum/topic/installation-getting-started-guide/
Group Admins
How to: Creation of Fractal Arrow Indi
This topic has been reported for inappropriate content
- This topic has 86 replies, 14 voices, and was last updated 9 years, 9 months ago by
.
- Posts
@smallcat @muuh
What I would do is when I record the data (fractals & zigzag pattern), record only the fractal and bar patterns where there were no gaps were in between close and open prices. Keep in mind that there is usually a very small gap of less than a pip during low/high volatile times. So I would first define what a gap is, maybe greater than 2pips in size. And I would look at the bars contained within the zigzag leg and if the bars contained within the zigzag have any gaps greater than 2pips, then I wouldn’t include it in my set.@muuh
Sorry to hear about the depression and the difficulties you are going through. It’s always more difficult to look at the brighter side and to be optimistic.. But know that no matter how good or happy or bad or sad it may be, they all pass and fade away. So your depression will too fade away and it will be of the past
Wish you a happier and successful new year!Focus, Patience, Determination & Order in chaos
Thanks for your kind words Savor0.
Mh playing with the idea of using some kind of average bar size to determine if the gap is to big. But it is probaly overkill
A ZigZag is allowed to have gaps.
@muuh, thank you very much for this good information bro …
@saver0, thank you very much too my friend …@muuh,
Do you know this 0.00005 gaps allowed for which TF ? Is it for M5 or also can be applied to M15, M30, H1 etc ? How to get that number (may be using ATR or other tools) ?
Thanks
So I have reprogrammed the logic for filling my database.
Now I have to start rewriting the logic for the prediction.
And I have to start with the hardest part. Doing the first step, the Pattern Match.
Is here someone who is relativly sure how the pattern match, the first step works?Maybe even GG can explain it again in a looooooong post.
According to carl’s pdf its only comparing occurences.
If ‘current fpattern’ has more occurences than ‘current + 1’ or ‘current +0’ take current.
Thats what i always thought.
Yep only comparing.
But comparing what against what? There are so many ways of doing that http://penguintraders.com/wp-content/plugins/wp-monalisa/icons/wpml_sad.gifYes I know that too ;)
I got a message from VlanFx with new information on how to do the first step. Will try it out over the next days because it needs some change in the database.
You can expect an answer from me before the end of next Monday. If you don’t hear from me some aliens kidnapped me or I’m dead :p
So still having a few issues with this new way of doing the first step.
But what’s promising is that even my not really working algo profits massivly from using fractals based on close.
I’m starting to have a little success
It still needs a bit of tweaking due to way too early signals.
There is some kind of filter missing for getting longer signals…..mmmmmh
I’m starting to have a little success
It still needs a bit of tweaking due to way too early signals. There is some kind of filter missing for getting longer signals…..mmmmmhPics or it didn’t happen
Yeah pics are coming. I have to program an indi in MT4 to get my signals on chart. Currently it is all by hand
Hi all
Posting two charts of my working/not working prediction.
In the first Picture you can see two working signals as an example. In the second Picture you can see where my algo has problems.
As soon as the “swing” or better the amount of bars between to fractals are longer as usually i get to early signals. The combination of pattern match and length match seems to be my problem.
I think my pattern match is wrong or maybe even what i compare in my pattern match.
Suggestions how to solve this?
My own. But I haven’t done anything different for Length Match and Next Bar Prediction.
The “Delay between Fractals” is giving me a headache. Maybe when checking against probability of pattern beeing longer (the amount of bars in the pattern), i don’t have to count together but look at each possibilitie individually and decide which is more likely.
Or when checking against Set2 I need more infomation about the bars that make up that pattern.
Another way i can think of is predicting many swings at the same time and when I get enough swing points at the same time it is a valid swing
My own. But I haven’t done anything different for Length Match and Next Bar Prediction. The “Delay between Fractals” is giving me a headache. Maybe when checking against probability of pattern beeing longer (the amount of bars in the pattern), i don’t have to count together but look at each possibilitie individually and decide which is more likely. Or when checking against Set2 I need more infomation about the bars that make up that pattern.
I’m not sure, have you looked through the document to see if the implementation is exact?
Also are you storing proper fractals and not the ones with gaps?Focus, Patience, Determination & Order in chaos
My own. But I haven’t done anything different for Length Match and Next Bar Prediction. The “Delay between Fractals” is giving me a headache. Maybe when checking against probability of pattern beeing longer (the amount of bars in the pattern), i don’t have to count together but look at each possibilitie individually and decide which is more likely. Or when checking against Set2 I need more infomation about the bars that make up that pattern.
I’m not sure, have you looked through the document to see if the implementation is exact?
Also are you storing proper fractals and not the ones with gaps?Focus, Patience, Determination & Order in chaos
If the document was right, everyone would have a working version. I tried many different ways of next bar prediction and they all give you the same signal. sometimes one bar earlier, sometimes one bar later. And Length prediction is the way gg told us but if we implemented it right….. how should i know?
Yes I’m storing only valid fractals
Guys, if you want i can post the little bit of skype chat I had with GG. In fact I’m running out of options and maybe someone else can use the hard to understand information GG gave. Sadly it is nearly impossibly to get GG to answer the few questions I still have.
Guys, if you want i can post the little bit of skype chat I had with GG. In fact I’m running out of options and maybe someone else can use the hard to understand information GG gave. Sadly it is nearly impossibly to get GG to answer the few questions I still have.
I think it doesn’t hurt if you do so, it will be more than welcome! What I’m particularly interested in are the questions you can’t get an answer to, sometimes these spark great ideas. Thank you for your efforts!
This is I think an important question:
What does this sentence mean from a post you made in FF: “Our historical database INCLUDES the last fractal in its next ZZ fractal pattern” Does it mean that the first fractal i store in my database for my current leg is the same fractal i stored as my last fractal in the previous leg?
or this:
Are you using the -1 bars in your fractal pattern match to get the delay between two fractals?
@GG53 I think what would really help if you could show us a chart and mark the bar where your first step gave you a go, then mark the bar where you got the length match. Or the other way around mark the bars where your pattern match gave you no signal and so on…..Then you don’t need to answer any questions anymore because everyone can see in the chart where he must have a signal or not
What’s the “delay” between fractals? I don’t remember seeing it anywhere except for the last couple of posts here.
The delay is the amount of bars between two Fractals. If the amount is higher then usual, for example 7 or 8 consecutive up bars, there are to early signals.
A quick and dirty fix would be to stop predicting when you have 3-4 consecutive up or down bars. But it wouldn’t solve the issue and you still get to early signals when the bars between to fractals is higher then usually
It still needs a bit of tweaking due to way too early signals. There is some kind of filter missing for getting longer signals…..mmmmmh
- You must be logged in to reply to this topic.