Not the Grail, just a regular one - Bablokos!!! - page 239

New comment
 
ArgumEEEEEnt)
 
_new-rena:

I know that he traded with me and Alexander, and that a lot of things that were not in that thread went through his personal account. I know how he traded, not out of the blue.

And the correlation is not constant. The price of the market is higher than the price of the real one.

Evaluating the correlation is the best way to overlap pairs.

That's not what I was saying. I was saying that you have the wrong idea about correlation and its corollary - cointegration
 
qimer:
That's not what I was saying. I was saying that you have the wrong idea about correlation and its corollary, cointegration
Okay.
 
lob32371:

... Do you go on to do an overshooting of lots for every such combination? That's not understanding the essence of the portfolio.

Maybe, of course, someone does consider it an overshoot, but I don't indulge in it. There are other tools and methods (Recycle, MNCs, etc.). And in general, to be in the "subject" it is necessary at least part of the branch to reread, but yourself think up what we do here and criticize.
 
qimer:

You have written some bullshit. Correlation shows the similarity in the movement of pairs. It does not show whether the pairs are converging or diverging in any way. High correlation is cointegration, i.e. the same movement of pairs.

No, correlation is not cointegration!

Correlation shows the closeness of relationship between data series (the similarity of movement of pairs, as you write).

And to understand what cointegration is, you have to imagine a person with a dog on a leash. The dog can run away, but only by the length of the leash (both to the right and to the left), but it will go the whole way next to the person. Here

It means that in case of high correlation the rows can diverge infinitely long time and in case of cointegration will oscillate around each other.

 
Mislaid:

I have statistics on trending synthetics. I took this definition: a synthetic is promising to buy if it has been trading above the AMA for half a month on the day. That is, there has been a breakout of the AMA. The volume of all synthetics shown in the file is no more than 6.1 USD minimum lots.

How do you build the trend synthetics? As I see you are using not only majors but crosses as well, the question is: what units do you use to calculate the total equity?
 
pastor-ru:
A list of all the synthetics in seven pairs with four pairs each. There are 35 pairs in total. I use mt5 robot based on Recycle2 I earned on average 50-100% monthly with drawdown 10%, but there are drops and failures in the drawdown stability is still not enough. I do not fully understand the mathematics of Recycle2 and spread chart movement in the channel.

a little too much for portfolio trading, on average it should be around 20%, but it depends on the particular trading method

 
Mislaid:

I have statistics on trending synthetics. I took this definition: a synthetic is promising to buy if it has been trading above the AMA for half a month on the day. That is, there has been a breakout of the AMA. The volume of all synthetics shown in the file does not exceed 6.1 USD minimum lots.

Column A is the day of the year when the synthetic was identified as a prospect. Column B - weight of the synthetic in minimum USD lots. Column D - the synthetic itself. Columns K, G and F - day, day number of the year and number of synthetics identified on that day as a prospect and remaining as such as of the last date. Column I is the number of synthetics retired from the prospect list as of the last date.

It can be seen that many of the synthetics are kept above the AMA for months. New ones are identified on a regular basis. So, they are not all similar.

Many synthetics are highly correlated with each other. I distinguish between two types of correlation: vector correlation and trend correlation. Vector correlation is when synthetics are very close in terms of currencies. Trend is when synthetics are close in terms of movement.

Vector. Each synthetic can be contrasted with a vector whose components are calculated as the value of a currency bought (sold with a "minus" sign) in the same unit of measure. The cosine of the angle between such vectors gives the vector correlation.

I took the plan file for 18.11.14. This is how one of the synthetics looks like today. Clusters are formed by trend correlation. Strongly vector correlated synthetics are truncated: the pattern with the smallest weight is selected.

It's hard to parse this file without commentary, but I get the general idea.

I think you're using the term "trend synthetics" in your sense.

I meant that trend synthetics (in their sense, i.e. obtained by regression on a linear trend) are all similar, because if we take an arbitrary set of symbols, wide enough (over 6-7), then on one period the graphs of the obtained trend synthetics will most probably look very similar visually

 
transcendreamer:

It's hard to make sense of this file without commentary, but I get the general idea.

I think you are using the term "trend synthetics" in your sense

I meant that trend synthetics (in its sense, i.e. obtained by regression on a linear trend) are all similar, because if we take an arbitrary set of instruments (more than 6-7), then in one period, plots of obtained trend synthetics will most probably look very similar visually

yep

two practically unchanged camps (portfolios) over the last 40 years, if you look at the graphs at months. more precisely - with the help of the correlation coefficient you can break down



 
lob32371:

I couldn't get the joker code to work. Yours did:

The result immediately made it clear what the monster joker code was doing:

Let's not focus on the school algorithm of combining combinations. And let's not stress that it would look a hundred times better in OOP. Let's get straight to the point.

Why do you make yourself such an elaborate pain? Do you continue to make an enumeration of lots for each such combination? So this is not understanding the essence of the portfolio.

There is a wonderful example of lot (coefficient - right) - zero. Well, consider that you have a zero lot for some symbols. So go through them.

Definitely, with such a search can not go far - even the Cloud will not help test the TS. You have to do the maths, without brute force.

#include <spreadsmath.mqh>

extern int N=4;

extern string Symbols="EURUSD,GBPUSD,NZDUSD,AUDUSD,USDJPY,USDCHF,USDCAD";




void OnStart()
  {
      int K=GetSpreadsCount(Symbols,N);      
      
      for(int i=1;i<=K;i++)
         {string symbol_list=GetSpreadByIndex(Symbols,N,i);
1...232233234235236237238239240241242243244245246...650
New comment