Thoughts on some of the absurdity of multi-currency analysis. - page 17

 
getch писал(а) >>

P.S.

The developers of MT5 have announced that they have managed to create a highly specialized algorithm of compression of quotes history, which excels all universal archivers in compression. The secret of it is an effective data conversion before compression. This is not a conversion of basis majors, but transformation of single trading instruments. If this transformation is disclosed by the developers, it can be used (without reinventing your own wheel) as a starting point to the goal of baseline transformation.

The basis without loss of information in compression is a single candlestick.

You can compress it simply: do not store for each candlestick information about rate, but only increments in punctuations relative to the previous candlestick and considering that H>O,C and L<O,C. For storing one candlestick HLOC 4 bytes is enough, but you can't compress it. At the beginning of a chart or when a gap is more than 2^8 pips, we should add a special symbol which sets a new reference point. This can also be compressed if volatility changes are taken into account and periodically a symbol is inserted indicating the dimension in bits to be used for storing subsequent HLOC increments. Time is also compressed - if there is a gap, a special symbol with a new readout point is inserted, otherwise time increments by candlesticks are considered according to TF

Anyway, lossless compression uses regularities, which cannot be directly used in forecasting. Otherwise MetaQuotes Software Corp. would cease to be a software company and become one with Goldman :)

 
Yeah, discussing the intricacies of the compression method without really understanding how that compression will help forecasting. How is a powerful spike (say, Payrolls) different from an overnight flat?
 

The constant bitrate hypothesis, the use of compression for prediction - all this at this stage is nothing more than thinking out loud.

However, the reasoning was that there was an absolutely concrete criterion for determining the best basis for finding patterns - compressibility.

Step by step:

  1. Save the history of all the majors in such a way that the archiver compresses such data well. This may be CSV-files or HST-files of MT4. But you'd better use your own format to store all the majors at once, not separately.
  2. Compress the obtained file with the archiver and get a file of size N-bytes. An archiver is selected among the best ones by its compressibility index.
  3. On the basis of multicurrency analysis we have created currency indexes (whatever you want to call them) - new majors that form a new basis.
  4. Repeated steps 1 and 2 for the new majors and got M bytes.
  5. If M < N, M-basis is optimal for finding patterns, otherwise N-basis.

Random values:

This way of determining the optimal basis says nothing about the nature of its majors. If EURUSD, GBPUSD and other majors are random variables, i.e. they and their crosses cannot be predicted, then M will always be equal to N (with a certain margin of error).


 
getch >>:

P.S.

Однако, анализ статистики корреляций различных мажоров (например, куда пошла EUR/USD, туда и GBP/USD) имеет смысл, как некоторая вероятностная характеристика. Она носит такой же смысл, как анализ, например, EUR/USD вместе с RUB/USD. Т.е. смысл слабо прослеживается.

You can spend a lot of time fretting about the (un)usefulness of multicurrency exchanges. But it's better to see once than blurt a hundred times.


I finalized Spreader_v7. Today I created an account for it and set up output to the site via FTP


Reports will be available at https://www.mql5.com/go?link=http://spreader.heigh.ru/statement/statement.htm


Trading server: UWC-Demo.com
Account number (login): 235953
Main password: ******
Investor password: uqmj0va
 
Reshetov >>:

Poorly explained, since you didn't understand the first post.

Your Spreader trades a cross based on the information of its constituent pairs. There is nothing wrong with that.

Example:

If you trade AUDJPY and NZDJPY (regardless of weights) it is trading on AUDNZD (this is especially obvious if the base currency of the account is JPY).

All AUDNZD information is contained in the AUDUSD and NZDUSD basis, or the converted AUDJPY and NZDJPY basis. But in EURUSD, GBPUSD, USDJPY and many other pairs there is NO information whatsoever concerning the AUDNZD behaviour. That's what was meant.

Spreader:

Perhaps you should run Spreader through such a tester on history. Of course, not with the provided tester, but with your own one. The idea seems to be simple.

 
Reshetov >>:

Можно долго лясы точить по поводу (бес)полезности мультивалютников. Но лучше один раз увидеть, чем сто раз флудить


Я допилил Spreader_v7. Сегодня создал ему счет и настроил вывод на сайт через FTP


Отчеты можно будет смотреть по ссылке https://www.mql5.com/go?link=http://spreader.heigh.ru/statement/statement.htm


Trading server:UWC-Demo.com
Account number (login):235953
Main password:******
Investor password:uqmj0va

What's new in version 7 compared to the posted ones?

 
getch, what you are gradually approaching is called fractal dimensionality.

in short, its essence is that a trajectory or set of points embedded in space can have a non-integer dimensionality
.If you have a line then the dimensionality is 1, if you have a plane then the dimensionality is 2, the volume is 3 and so on. The dimension of a set can be 1.3, 2.5, any number. In fact, the dimensionality means the number of independent variables.

I.e. you have 8 currency pairs and the question "Do all these quotes correspond to each other?" To answer the question, you should impose (somehow) all these quotes in multidimensional space and define the dimensionality of this set. If it will be = 1 - then all 8 pairs can be reduced to one, if it will be 2 - then two pairs are "real" (independent), etc. I did it long ago I don't remember the result, but something like 1 and something. I.e. all this data equals little more than one scalar time series (one quote). In fact, it does not mean that you can watch only EURUSD and know all other rates.

If it is interesting, google this topic, if it is not clear at all, here you can watch a scientific pop-up http://rutracker.org/forum/viewtopic.php?t=2690789
You can also ask me, I will help you in any way I can.
 
Give proof, not a sci-fi reference.
 
irriss >>:

At school I had a good sense and understanding of the meaning of abstract dimensionality of spaces and the relationship to fractals. There's almost nothing left in my head. Thanks for the tip and the link. At the moment I still don't understand why the dimensionality of quotation spaces says they are dependent.
My botanical hypothesis, on the other hand, boils down to the independence of the majors from each other.

 
getch >>:

В школе хорошо чувствовал и понимал смысл абстрактной размерности пространств и взаимосвязь с фракталами. В голове почти ничего не осталось. Спасибо за наводку и ссылку. На данный момент еще не понимаю, почему размерность пространства котировок говорит о их зависимости.
Моя же ботаническая гипотеза сводится к независимости мажоров друг от друга.

Majors are probably as independent as "minors", but the only question is, what can you express their value in? If, for instance, you express everything in dollars, then the correlation between pairs will be present, because the influence of the dollar is evident in every pair.

In my opinion, "minors" are not much worse than "majors" in that respect. The only problem is to build an objective index of each currency in order to diversify the influence of other currencies in the pair.