The similarity system

OK, I think it is easier to ignore people who do not know when to stop talking about trivia.

Over the last several pages of the thread, there has been a fair share of criticism of the thread, some focused on the mathematical definition of the similarity principle as placed recently on the first page of the thread by the thread starter, claiming that it is complete gibberish. So I think it might be nice to have a discussion of the meaning of the definition of similarity.

In the next post I will attempt to give my understanding of this definition, with examples from the thread. I invite people to join the discussion.

Cheers,
k

A good read...

Very interesting, thanks!

I was considering doing something similar for the news trading on forex. The idea being (similar as in the Vigovski's thread) that minutes before the news the informed participants pump the money into the currency and this can be detected. Since the volumes they place are huge, but, of course, they don't use my poor man's broker giving me the information about volumes, a good measure to see what is going could be to look at the large price movement with small volume i.e. changes in price divided by volume before the news.

k

Anyway, back on topic, what does this mean?



First some background information: I am not a mathematician, I hope that I don't offend any matematicians by my loose language.

The main idea of the definition is to establish a link (mathematically, this is done by a function or a mapping) between two stochastic processes.

The processes live inside measurable metric spaces (X,mu,d), here X is a vector space, which means an abstract set of elements with various properties (http://en.wikipedia.org/wiki/Vector_space) satisfied. On top of this structure there is a metric (http://en.wikipedia.org/wiki/Metric_space) defined by a measure of distance d. This means that for every two elements of the metric space, say x and y, there is a function d(x,y) that satisfies all the properties postulated for a distance function (http://en.wikipedia.org/wiki/Metric_space#Definition). And on top of this in these structures there is a concept of measure (http://en.wikipedia.org/wiki/Measure_%28mathematics%29) mu.

Note that here it can be that both the metric spaces are the same, or that one is the subspace of another one etc.

For the case of price signal, we have a one-dimensional vector space of real numbers, where the "distance" between two elements is given by the absolute value of their difference.

On these structures the stochastic processes are defined. For this definition one also needs a concept of time T, and a specific time t. One would typically have a stochastic process of a price signal, or a stochastic process of an indicator readout. One example: While there is nothing really "stochastic" about the Stochastic oscillator (it is a well defined function of the price and the period), it can still be seen as a stochastic process in itself. So, indicator values can also be thought of as stochastic processes in this abstract manner.

Then, most importantly, one says that the two stochastic processes are "similar" or "lambda"-isomorphic (isomorphism (http://en.wikipedia.org/wiki/Isomorphism) I loosely perceive as formal equivalence, since there is a function mapping every element from the first set to the second set and vice versa) by introducing a function g. g acts on the vector space X1 (for example the price signal in our case, more precisely the domain is a partition of the first set) and assumes values in the space X2, for example the values of the indicator.

Now the properties of this function/mapping g are interesting:
1) The measure of the co-domain X2 not mapped into is zero. So it means that almost surely every element of the set X2 is a result of the function g acting on an element from X1.
2) Probabilistic property of the function g: For every element from the co-domain g(a) it is possible to associate the value of the stochastic process X2, and that implies that the value of the stochastic process X1 at this moment t is within a surrounding (open-ball) of the element a with a probability equal to 1. Here one establishes the link between the isomorphism and the two stochastic processes. Essentially, this means that the two stochastic processes are equivalent in a probabilistic way, such that when one knows the value on the second process X2(t), one can approximately deduce where the first process value X1(t) is.

So in my opinion there is nothing really strange or superfluous or wrong in this... It looks to me, having read a bit of mathematics books in my time, a mathematically sound definition of a link/isomorphism between two stochastic processes. We need the metric spaces to define the distance (surrounding/proximity of element), we need the measure to define the probability, etc. etc. The main problem of this is to establish the isomorphism and to check the properties 1) and 2).

I am happy to hear other constructive or critical opinions,
Cheers,
k

Man looks like a missed a lot this weekend... But after reading it all its the same stuff thats been going on since page 598. A bunch of balogne.

Glad to see some good posts are coming through all the mud, by Kprsa and others, keep up the good work brothers! thanks for trying to steer this ship back on course!!



My advice to all the haters, need to go read what NT said and if they do not understand that 2+2=4 please take your business else where.

Any my advice to all the good men and women(if any) on this thread, stop feeding the trolls. They will find food somewhere else!

Pip

Good work, thank you mate.

IMHO this whole probability trading method can only be use profitably if you have another definitive intraday trend indicator in order to be able to quantify the pip value of a particular pairs in that particular trend. Different pairs have different dynamics and values which add to the complexity of this method thereby generating a lot of controversial ideas and trading innuendoes and terminology which I believe the OP himself is unable to really profit from it. Things can be easily be done with just an appropriate and suitable momentum tool for each individual pairs. No unnecessary hassles to further complicate this already very tough and challenging trading business. Simply have a method which can tell you exactly whether you are right or you are wrong. But I suspect that 99% of the traders is somewhere in between not knowing if one is exactly right or really wrong. Just sharing my thought and idea with no malaise intended. Agreeable and unagreeable ideas will be appreciated. Scalpers and their scalping methods which are currently guided and benefited by this method got to stay alert and nimble to avoid imminent drawdowns coupled with a punctured confidence. Traders who is highly leveraged and trading with limited resources and funds are advised to stay clear of this method before doing a irreversible damage to their and their family life saving accounts.

Happy for any constructive comments.

Good luck and happy trading.

GS.

I've just read from somewhere, but, does the cycleidentifier repaints?

While I would agree with most of this (apart about the existence of the indicator which tells you exactly what to do), I really don't see anything specific about the similarity system here. You can make exactly the same comment on any other trading system thread. (Almost?) everybody has to deal with probabilistic trading. The difference here is that we typically know what the probabilities are.

k

Yes, very much so. Read this comment (http://www.forexfactory.com/showthread.php?p=7723681#post7723681) if you want to verify for yourself.
It is possible to combine the CIs (which more-less denote specific fractals on the price chart) with confirmed end-of-bar transient (fractal) zones to increase the probability of non-repainting, and it is possible to quantify the probability of this occurring.

k

Thank you K for steering this thread back to whats important. For a non-mathematician you sure can fake it pretty good! I have not really bought into the dissimilarity concept so much, not saying it doesnt have merit.
If anything it seems to me that a 'wormhole' opening will only be for a short period, so the time to profit is when the wormhole first opens. As for wormholes being a future target (price returning) I merely think that is the phenomenon of price being recurrent majority of the time.

I would like to ask a bit more about TZ's stats tomorrow when I get the chance.

Again thank you K.

(edit, to make the statement more precise): I agree (http://www.forexfactory.com/showthread.php?p=7874160#post7874160) with (http://www.forexfactory.com/showthread.php?p=7853052#post7853052) your sentiment about the wormholes as a price to return to, but I think they are useless even at their time of inception.

k

Of course I understand this is the best method you have seen and tried and I fully appreciate it. And I know very few can have a method better than this as i have anticipated and mentioned in my earlier post.

Anyway and anyhow Good luck and wish you the wildest success, it is a dream worth chasing for.......

This will be my last post in this thread.

GS.

I meditated on what you said here for about a day. Finally i came up with something:

People who are very insecure and dont know themselves cant even admit to themselves much less other people when they dont understand something. They try to hide the fact that they dont understand by spewing garbage. People like that are not ready to learn.

People who sort of have a grip on themselves can admit to themselves and even other people that they DONT understand. To me this is where learning begins. I think there is little point in reading a book you already understand, what would be the point? Same thing here.

I dont have to tell you where you fit into this picture
And there is little point in telling someone else where they fit into this, what would be the point? They are not ready to learn anyway.

Happy pirateing!

Haha, you did make the same comment on all the other threads!
OK, I am intrigued, so I will study your previous comments and see what I come up with.

k

And that is the truth, the whole truth and nothing but the truth.
k

I call it as i see it!

Does anyone have the .ex4 of CycleIdentifierwHistory?

Could anyone tell me should I buy or sell at these green arrows?

Regards,

I suggest you find and follow a member here by the name of "Ullen" (I'm not suggesting you bother him with PM's mind you, he may or may not have his handsfull for all I know). But I do suggest you read his previous posts he has made.... numerous posts on Similarity trading and I believe he has his own thread too, goodluck!