Secret footprint in the markets | Page 3 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Dax MT4 delta and imbalance footprint chart 26 replies

can anyone guide me how to profit from footprint chart 8 replies

Futures Footprint Charts 3 replies

Help request for footprint charts 2 replies

Tape, footprint and order book trading 4 replies

Attachments: Secret footprint in the markets
Exit Attachments

Secret footprint in the markets

  • #41
  • Oct 11, 2014 8:39pm Oct 11, 2014 8:39pm
give a try with 5bits & Daily Chart
this may give some light

Quoting Proximus
{quote} I could just remove the 2 values altogether because they are representing the same CLOSE price as if the market didnt moved at all. But why does the binary files help, how do I know that the 8 size cluster is the optimal one, if we are talking about clustering the data, then any number of cluser size can be good. For example for a morning star pattern we should have a 3 size cluster. While the Bill Williams fractal method suggests 5 size clusters. So we cant know which cluster size is the optimal until we didnt tested it, if we want to do...
My Best Trade is Next Trade!
  • #42
  • Oct 12, 2014 1:59am Oct 12, 2014 1:59am
Quoting Proximus
I also made this method more visual, by coding each option into a color in excel, so: black background = "X" --> previous bar close >current bar close blue background = "O" --> previous bar close <current bar close yellow background = "-" --> previous bar close =current bar close And this is what it looks like, I've deleted the text to be more visual, so only the background is visible {image} Here is the macro: Sub JONNY() Dim i As Long Dim j As Long For i = 1 To 1125 For j = 1 To 200 If Cells(i, j).Value = "X" Then Cells(i, j).Interior.Color =...
Nice and RANDOM (:
  • #43
  • Oct 12, 2014 6:09am Oct 12, 2014 6:09am
Quoting Mingary
{quote} Nice and RANDOM (:
Yes it may seem random, but there can be still a subtle pattern under the noise, which can be detected only by math tools and not by the naked eye.

Quoting MAM.Trader
give a try with 5bits & Daily Chart this may give some light {quote}
Yes but why 5 and not 8, why 7 and not 3, how do you select the most optimal cluster size if you dont know nothing about the properties of the historical data?
"There's a sucker born every minute" - P.T. Barnum
  • #44
  • Oct 12, 2014 7:05am Oct 12, 2014 7:05am
Quoting Proximus
Yes but why 5 and not 8, why 7 and not 3, how do you select the most optimal cluster size if you dont know nothing about the properties of the historical data?
Google for Automatic Relevance Determination

I never found any clustering of the candles where a pattern gives more than 50% probability for the direction. I expected the algorithms to re-discover the usual patterns but all the approaches rejected them all. Is it because I cannot label the group of bars correctly (assigning them a BULLISH or BEARISH tag) or because there is nothing to be found? That's the question.
No greed. No fear. Just maths.
  • #45
  • Oct 12, 2014 10:51am Oct 12, 2014 10:51am
Quoting PipMeUp
{quote} Google for Automatic Relevance Determination I never found any clustering of the candles where a pattern gives more than 50% probability for the direction. I expected the algorithms to re-discover the usual patterns but all the approaches rejected them all. Is it because I cannot label the group of bars correctly (assigning them a BULLISH or BEARISH tag) or because there is nothing to be found? That's the question.
There could be patterns hidden, but i agree that CLOSE[i]>CLOSE[I+1] type of coding is pretty simplistic and naive. If there is any pattern to be found, then obviously it has to be a very complicated pattern so that nobody with average knowledge or limited CPU can find it. I`m sure that quant hedgefunds do research this phenomena and they have good patterns.

I also talked about that quant firm which provides research services to institutions and only share the research after a ND agreement.And they are specialized in data mining and clustering.

So I have no clue how they do it, but i think this is anyway a dead end, simply because either the pattern is so complex that average PC's cant handle the calculations, or there is none to be found and this is a waste of time.

So either one of the cases is true, this is a dead end approach to find a holy grail.
"There's a sucker born every minute" - P.T. Barnum
  • #46
  • Oct 12, 2014 11:15am Oct 12, 2014 11:15am
5 isnt random. 3 bar fractal needs 5 bars. 2 left + 2 rigt +1 High/Low bar

Quoting Proximus
{quote} Yes it may seem random, but there can be still a subtle pattern under the noise, which can be detected only by math tools and not by the naked eye. {quote} Yes but why 5 and not 8, why 7 and not 3, how do you select the most optimal cluster size if you dont know nothing about the properties of the historical data?
My Best Trade is Next Trade!
  • #47
  • Oct 12, 2014 11:18am Oct 12, 2014 11:18am
Quoting MAM.Trader
5 isnt random. 3 bar fractal needs 5 bars. 2 left + 2 rigt +1 High/Low bar {quote}
Yes but who said that 5 is the magical number, you can say it's 3, or 7 or any odd number except 1. It is just an unverified claim. I`m sure mr. Bill Williams didn't tested quantitatively the cluster sizes and came up with the conclusion of the "5" is the absolutely best number.I`m more likely to believe that he pulled that number out of his *ss.

Otherwise show me the scientific documentation about this cluster analysis where it's scientifically proven that 5 is the magic number.
"There's a sucker born every minute" - P.T. Barnum
  • #48
  • Oct 12, 2014 11:28am Oct 12, 2014 11:28am
I assume that you can read & you can understand.
I did NOT say it is MAGICAL NUMBER.
I siad it is the number of bars needed to create a fractal.

Never mind ..........

Quoting Proximus
{quote} Yes but who said that 5 is the magical number, you can say it's 3, or 7 or any odd number except 1. It is just an unverified claim. I`m sure mr. Bill Williams didn't tested quantitatively the cluster sizes and came up with the conclusion of the "5" is the absolutely best number.I`m more likely to believe that he pulled that number out of his *ss. Otherwise show me the scientific documentation about this cluster analysis where it's scientifically proven that 5 is the magic number.
My Best Trade is Next Trade!
  • #49
  • Oct 12, 2014 11:32am Oct 12, 2014 11:32am
Quoting Proximus
{quote} There could be patterns hidden, but i agree that CLOSE>CLOSE[I+1] type of coding is pretty simplistic and naive. If there is any pattern to be found, then obviously it has to be a very complicated pattern so that nobody with average knowledge or limited CPU can find it. I`m sure that quant hedgefunds do research this phenomena and they have good patterns. I also talked about that quant firm which provides research services to institutions and only share the research after a ND agreement.And they are specialized in data mining and clustering....
CPU power is no longer an argument nowadays. Should you need 1000 big CPUs you can just rent a grid at amazon for a few hours. Also I don't think the patterns must be "very complicated" because the more complex they are the less often you'll find them. If you have only a few samples you can't decide their statistical value.

I could reproduce some of Relativity's finding (see Rel's lab). People in the thread Fractals, ZZ & the Pissing Dog seem to have found something. Ronald Raygun's Pattern-Finding EA looks like to have some value too. For this reason I think that "only big institutions can find something" doesn't hold either.

The usual candlesticks pattern only make sense near key levels like S/R, trend lines, long term MAs... I think the pattern shall consider the context. You have to data mine both the context and recent candles formation.
No greed. No fear. Just maths.
  • #50
  • Oct 12, 2014 11:35am Oct 12, 2014 11:35am
discussing with him is waste of time. let him create his own periods & numbers from his own part of body

Quoting PipMeUp
{quote} CPU power is no longer an argument nowadays. Should you need 1000 big CPUs you can just rent a grid at amazon for a few hours. Also I don't think the patterns must be "very complicated" because the more complex they are the less often you'll find them. If you have only a few samples you can't decide their statistical value. I could reproduce some of Relativity's finding (see Rel's lab). People in the thread Fractals, ZZ & the Pissing Dog seem to have found something. Ronald Raygun's Pattern-Finding EA looks like to have some value too. For...
My Best Trade is Next Trade!
  • #51
  • Oct 12, 2014 11:36am Oct 12, 2014 11:36am
Quoting MAM.Trader
I assume that you can read & you can understand. I did NOT say it is MAGICAL NUMBER. I siad it is the number of bars needed to create a fractal. Never mind .......... {quote}
That is not what I am saying, I say that how do you know that fractals of 5 size are the holy grail? You can call it fractal or whatever i dont care, the point is that:
You have no scientific proof behind that number just bunch of 70's technical analysis crap.

Ignoring me just proves your own inability to distinguish the difference between retail trader myths and scientific proof.
"There's a sucker born every minute" - P.T. Barnum
  • #52
  • Oct 12, 2014 11:37am Oct 12, 2014 11:37am
Quoting MAM.Trader
I said it is the number of bars needed to create a fractal. Never mind .......... {quote}
What Bill Williams called a fractal is what Thomas Demark called a TD point. More precisely a 2-TD point because there are 2 bars both side of the middle bar. But Demark also defined a N-TD point as a fractal with N bar each side. Proximus' argument about the correct number of bars is therefore valid.
No greed. No fear. Just maths.
  • #53
  • Oct 12, 2014 11:43am Oct 12, 2014 11:43am
I already tested 5 bars months ago with different methods of calculations. daily 5 bars analyses has a potential. but actually I am on other project regarding the waves, so wanted to tell him if he can give try while he is on that path.
"the person who thinks knows everything, knows nothing" this gentelman has an answer for every post/word.

Quoting PipMeUp
{quote} What Bill Williams called a fractal is what Thomas Demark called a TD point. More precisely a 2-TD point because there are 2 bars both side of the middle bar. But Demark also defined a N-TD point as a fractal with N bar each side. Proximus' argument about the correct number of bars is therefore valid.
My Best Trade is Next Trade!
  • #54
  • Oct 12, 2014 11:46am Oct 12, 2014 11:46am
n bar fractal is not new. you can set anybar fractal with dukas platform. but in general 3 bars fracal used to pin the high & low. this is most common. when trading using most common tools gives best results. my2cnts

Quoting PipMeUp
{quote} What Bill Williams called a fractal is what Thomas Demark called a TD point. More precisely a 2-TD point because there are 2 bars both side of the middle bar. But Demark also defined a N-TD point as a fractal with N bar each side. Proximus' argument about the correct number of bars is therefore valid.
My Best Trade is Next Trade!
  • #55
  • Oct 12, 2014 11:47am Oct 12, 2014 11:47am
The Ultra experienced traders use a very simple rule. they don't use technical analysis they even don't trust it .they just stare at bare charts with no indicator. but they can read it better than us. they look the bare bars and easily decide if the price is cheap!!!(if yes they go long), if it is still too expensive the go short. if none, they wait .
I have no idea of how they understand if a pair is cheap or expensive !!! maybe once they've been great wholesalers.
  • #56
  • Oct 12, 2014 11:48am Oct 12, 2014 11:48am
I am done. I dont have time to discuss nosense
GL
My Best Trade is Next Trade!
  • #57
  • Oct 12, 2014 11:53am Oct 12, 2014 11:53am
Quoting PipMeUp
{quote} CPU power is no longer an argument nowadays. Should you need 1000 big CPUs you can just rent a grid at amazon for a few hours. Also I don't think the patterns must be "very complicated" because the more complex they are the less often you'll find them. If you have only a few samples you can't decide their statistical value. I could reproduce some of Relativity's finding (see Rel's lab). People in the thread Fractals, ZZ & the Pissing Dog seem to have found something. Ronald Raygun's Pattern-Finding EA looks like to have some value too. For...
I disagree, it's not just about the current need for computing power but the constant need for it. Because you also have to rethink the course of the research after analysing every result. So you need to rent it out like for 6 months or so. How much would that cost you?

I bet that the cost of this rent would overpass the profit that the research will give to you (unless you have billions to trade with like HF have). Average retail traders simply dont have the funds to allocate for this kind of research.

Also what if the pattern is so abstract that you don't find it after 6 months, i think luck also plays a big role in this kind of research because you dont even know what to look for.

Rel's lab and the others might have been lucky that they found those patterns, who knows after how many hours of watching price. Is it just worth the effort?
"There's a sucker born every minute" - P.T. Barnum
  • #58
  • Oct 12, 2014 1:58pm Oct 12, 2014 1:58pm
Quoting tablet8
The Ultra experienced traders use a very simple rule. they don't use technical analysis they even don't trust it .they just stare at bare charts with no indicator. but they can read it better than us. they look the bare bars and easily decide if the price is cheap!!!(if yes they go long), if it is still too expensive the go short. if none, they wait . I have no idea of how they understand if a pair is cheap or expensive !!! maybe once they've been great wholesalers.
Agree. Draw more lines to accumlate experience by hand and then compare and then make conclusion. Hand and pencil are traders' good friend !
  • #59
  • Oct 12, 2014 2:20pm Oct 12, 2014 2:20pm
Quoting MAM.Trader
5 isnt random. 3 bar fractal needs 5 bars. 2 left + 2 rigt +1 High/Low bar {quote}
You're correct, fractals are useful. On D1, EURUSD, after a 5-candle fractal pattern, the price has less than 30% probability to go beyond the fractal (above the top fractal or below the bottom fractal) within the following two weeks (=10 trading days). It's quite similar for other timeframes as well.

k
  • #60
  • Oct 12, 2014 6:06pm Oct 12, 2014 6:06pm
Quoting PipMeUp
{quote} CPU power is no longer an argument nowadays. Should you need 1000 big CPUs you can just rent a grid at amazon for a few hours. Also I don't think the patterns must be "very complicated" because the more complex they are the less often you'll find them. If you have only a few samples you can't decide their statistical value. I could reproduce some of Relativity's finding (see Rel's lab). People in the thread Fractals, ZZ & the Pissing Dog seem to have found something. Ronald Raygun's Pattern-Finding EA looks like to have some value too. For...
Thank you PipMeUp =)
http://www.forexfactory.com/showthre...07#post7795907
0 traders viewing now