Market etiquette or good manners in a minefield - page 101

New comment
 
Neutron писал(а) >>

I have data on the frequency of patterns (fig. right) and their predictive validity (fig. left) as a function of H:

The data are given for d=5. The red colour shows the larger values, the blue one the smaller values.

Almost exactly what I need. Only I would like to make sense of it in numbers.

For example, if we take only 10 patterns from the right figure with H from 10 to 20 (red zone by eye)

и

see how the rule support changes (what you call pattern repeatability).

What % is it equal to there?

Set the min. support threshold

Then for the values of support above the thresholds see interestingness. (Please describe what you call "reliability of forecast", it may be the attractiveness),

and also choose the interest threshold.

We'll get a set of "working rules", i.e. those that we trust:

  • both in support (significance of the condition in the overall sample)
  • and interestingness (significant link between the condition and the consequence).

And on them we can already test the TS for profitability.

Moreover, the GA will help to "fit" the thresholds of support and interest


Let's look at the table of support and attractiveness.

Can you export it to Excel?

 
paralocus писал(а) >>

to Neutron

Have a good trend and large profits.

Thank you, Fedor!

Solve your current affairs and come back.

M1kha1l wrote(a) >>

Can you export it to Excel?

Yes, I can. Anything.

Here's a text file, for example. In it I will show the RT built on EURUSD tick history for one year for various H's (columns).

The format of the file is as follows:

  • the zero line contains the horizon of H segmentation in points,
  • the first line shows the number of PT segments.
HideYourRichess wrote(a) >>
1000

I have several non-contradictory hypotheses regarding your post... The most likely one is that you got the zero wrong:-)

Files:
kagyh.zip  687 kb
 
paralocus писал(а) >>

The idea is to detect the switching pattern +H/H and to skip one or two PT samples, after each committed transaction. Definitely, there should be duration (lifetime) statistics for the +H and -H strategies. A strategy should be changed after n RT samples from +H to -H(after 1 step pause) and vice versa after n RT samples from -H to +H. From my observations with kagi - tick series splits, there is one strong and repeating pattern: when top of previous RT falls in delta neighborhood (not more than 3-5 pips) of current (last) kagi top - you should change strategy from +H to -H, and to catch this pattern you should skip 1-2 RT samples after transaction - not trade on them, but analyze.

Fedor, thank you for:

  • confirmation of my attempt to make my point about intraday H-volatility instability (see 'Market etiquette or the rules of good manners in a minefield')
  • and, as a consequence, the need to change H strategy even intraday (or consider its shift)
  • and a BIGGEST THANK YOU for the suggested non-statistical method to detect the moment of H-strategy change.

Next, please criticise me for the following idea:

  1. for this approach it is inevitable to choose Н so that volatility would be as close to 2 as possible (at least for statistical stability reasons)
  2. Suppose we received H that volatility equals 2.1H, respectively we chose H+strategy
  3. then all we will get on each transaction = 0.1H-Spread

I hope 1-2-3 is correct.

If so, from this we can draw, imho, at least. 2 conclusions:

  • H > Spread/0.1 (of course, we can try to get closer to 2H less than 0.1, but is it practically possible?)
    i.e. with a EURUSD spread of 0.0002 H should be at least > 20
    (look at Neutron 13.07.2009 16:47 on the right picture - this is the value from which support for the most credible 10th pattern starts falling)
  • There is no need to wait until after opening a trade the kagi m.b. goes over the extreme pursuing the goal of "being in the trade all the time" for at least two reasons:
    1. You will increase your capital efficiency over time (the money will be "frozen" in the open trade for less time)
    2. The probability of positive closing of the trade will increase, because we can not say in advance about (using Fedor's terminology) "... the top of previous RT reference does not fall in the delta neighborhood (no more than 3-5 points) of the current (last) CAGI top...".

      So we can draw the following general conclusion:

      it's much more practical to open a deal with TP calculated as excess of the current volatility over 2H, well, and, if possible, offset the trade, though for small excesses of the current volatility over 2H it will strongly interfere instrument volatility.



      The confirmation or rejection of these thoughts would be very interesting because I want to avoid the idea degenerating in the pipsator.

       
      Neutron писал(а) >>

      I can, in anything.

      Here, for example, is a file in text format. It contains a PT built on the EURUSD tick history for a year, for different H's (columns).

      Sergey, if it's not difficult, would you be so kind as to put in 3 normal form in two fields

      H and Length_of_the_Cag_In_H

      in order of cagi segments in time from top to bottom


      I'm afraid, at night a table of 100 fields is not strong enough to deploy :)


      I've looked through the file, can you explain why there is no sign-variability for the construction cage?

      5
      37127
      0
      6
      8
      11
      13
      12
      8
      -10
      12
      ...

      5 - clear value H

      37127, i.e. 37128 is the number of cage segments.

      and below, imho, they should go "favorite" with an alternation of the sign

      Or is k.t. a different idea in this set?

       
      Neutron >> :

      I have several consistent hypotheses regarding your post... The most likely one is that you got the wrong zero:-)

      There is no mistake whatsoever.

       

      Explain to me please, I have been working in finance for I don't know how long and the term transaction has always been used.

      And now I look on wikipedia, supposedly in banking it is a transaction. It's very strange, liposuction...

      Who can comment?

       
      M1kha1l писал(а) >>

      Sergey, if it is not difficult, please be kind enough in 3 normal form in two fields

      H and Length_Cag_of_the_Segment_in_H

      and below, imho, they should go "favorite" with an alternating sign

      Hi. I've been on a business trip.

      Michael, I have given a Transaction Series (these are entry/exit points, not Kagi-building vertices), it should not be alternating and the average value of links for it is not equal to 2H (as for Kagi-building).

      HideYourRichess wrote >>

      There is no error.

      Then, what does your mean - 1000?

      gip wrote(a) >>

      Please explain, I've been working with finances for I don't know how long and the term transaction was always used.

      And now I look in wikipedia, supposedly in banking - transaction. It's very strange, liposuction...

      Who can comment?

      Transaction, I read it in Shiryaev's two-volume book. I thought it would look so smart. Then I myself have repeatedly encountered different spellings of the word in the literature. For the sake of certainty, I should stick to the more widespread variant - transaction.

       
      Neutron >> :

      Then what does yours mean - 1,000?

      Obviously the 1,000th post in this thread, 100 pages of 10 posts. You can sum it up for the anniversary.

       
      Neutron писал(а) >>

      Hi. I was on a business trip.

      Mikhail, I have a Transaction series (these are entry/exit points, not Kagi-building vertices), it should not be sign-variable and the average value of links for it is not equal to 2H (as for Kagi-building).

      Sergey, let's do by your methodology - step by step and simplifying.

      We are looking for Kaga patterns, and then we make a decision about transactions - they are secondary.

      So I suggest we start with the kaga patterns first.



      Attach the cagi segments on the ticks, I'll process them quickly,

      As I already wrote a Kaga handling software for PERIOD_M1 - it will be interesting to compare.




      This is what it looks like in Excel

       

      Michael, I'll attach the Kagi-building file a bit later, if it's warranted...

      Now about its usefulness (relevance). By construction, it's always an acquiniscence series. What do you propose to look for in its patterns, Kagi zigzag aspect ratio? To my mind, it has long been a trampled through and through theme that is of no practical interest. And here the PT is even very interesting! It is the basic one for trading, because it defines the market entry and exit points. In his dissertation, Pastukhov exploited the simplest property of PT - its "almost" sign-variability. He devoted the most part of his work to this question. At the end of his work, he touches briefly on more complex PT-pattern configurations, and it is this aspect that I studied in my research, the results of which I have given above.

      Are you suggesting that we start all over again, from Kagi's constructions?

      1...949596979899100101102103104
      New comment