Not the Grail, just a regular one - Bablokos!!! - page 263

 
Joker:

are the same as everyone else's: take or trawl.

In most cases, the equity level is the take.

Sensei, can I see 1-2 fresh "textboxes" from you for analysis? I can't take apart the old ones because I don't have such a deep history, and the last state, which was kindly published by you, as I understand, has already been optimized by you, which we "earthly people" are still very far away from.
 
seedormatrasch:
Sensei, is it possible to see 1-2 "textboxes" fresh from you for analysis. Old ones are out of hand for there is no such a deep history, and the last state, which was kindly published by you, I understand already with your author's optimization, which we "mundane people" are still very far away from.

There's already 250 pages of them...

There's more than enough material, go ahead...

P.S.: I'm not going to post the pages. Here's yesterday's market exit:

2015.11.19 03:17:4455712505USDCADsellout0.051.32766700960460.00-0.12-16.3610 382.68
2015.11.19 03:45:4555714495USDJPYsellout1.37123.415700977970.00-0.3175.1010 457.47
2015.11.19 03:45:4555714497USDCHFsellout1.041.01469700977990.000.43295.7610 753.66
2015.11.19 03:45:4555714500USDCADsellout1.191.32635700978030.00-2.76-507.2210 243.68
2015.11.19 03:45:4555714502NZDUSDbuyout1.950.65406700978060.00-63.90-985.629 194.16
2015.11.19 03:45:4555714505GBPUSDsellout1.281.52930700978090.00-1.181 044.2810 237.26
2015.11.19 03:45:4555714507EURUSDsellout1.551.07165700978110.00-3.14646.8510 880.97
2015.11.19 03:45:4555714508AUDUSDsellout2.450.71641700978130.0066.931 148.9512 096.85

P.P.S.: I won't post anything else here...

 
Joker:

There's already 250 pages of them...

There's more than enough material, go ahead...

P.S.: I'm not going to post the pages. Here's yesterday's market exit:

2015.11.19 03:17:4455712505USDCADsellout0.051.32766700960460.00-0.12-16.3610 382.68
2015.11.19 03:45:4555714495USDJPYsellout1.37123.415700977970.00-0.3175.1010 457.47
2015.11.19 03:45:4555714497USDCHFsellout1.041.01469700977990.000.43295.7610 753.66
2015.11.19 03:45:4555714500USDCADsellout1.191.32635700978030.00-2.76-507.2210 243.68
2015.11.19 03:45:4555714502NZDUSDbuyout1.950.65406700978060.00-63.90-985.629 194.16
2015.11.19 03:45:4555714505GBPUSDsellout1.281.52930700978090.00-1.181 044.2810 237.26
2015.11.19 03:45:4555714507EURUSDsellout1.551.07165700978110.00-3.14646.8510 880.97
2015.11.19 03:45:4555714508AUDUSDsellout2.450.71641700978130.0066.931 148.9512 096.85

P.P.S: I won't post anything else here...

It's true - there's plenty of material. And thank you again for the information.
 
Joker:
https://www.mql5.com/ru/code/1146

Guys, I'm not very good at maths. Can you tell me how to select weights in synthetics to minimize dispersion on given window via regression, for example in lrbuild we need to feed both regressors and values of synthetics itself to find coefficients. Then which libu to solve optimization problem to find a synthetic with min. variance? I would be grateful if you could show me an example with code adaptive to mql5 (4).

 
kirillka:

Guys, I'm not very good at maths. Can you tell me how to select weights in synthetics to minimize dispersion on given window via regression, for example in lrbuild we need to feed both regressors and values of synthetics itself to find coefficients. Then which libu to solve optimization problem to find a synthetic with min. variance? I would be grateful if you could show me an example with code adaptive for mql5 (4).

In lrbuild you need to feed either:

(1) all regressors against some table function

or

(2) a set of regressors against some asset


in general case:

CAlglib::LRBuildZ(MATRIX,points,roots,info,LM,AR);

the last MATRIX column must be a tabular function or asset data (spread legs)

 
transcendreamer:

in lrbuild we have to feed either:

(1) all regressors against some table function

or

(2) a set of regressors against some asset


in general case:

CAlglib::LRBuildZ(MATRIX,points,roots,info,LM,AR);

the last MATRIX column must be tabular functions or asset data (spread legs)

This is clear.

The question is how to search for a spread with minimum variance.

Of course, you may try to search for the minimum of a function by equating particular derivatives ofσ(Y(A1..A4)) to zero, or you may really bother to find a solution to the system of equations for finding the minimum of a function.I'm just sure there's a ready-made library to find the solution in a couple of lines of code). There's a discussion of the concept, but nothing is written about the implementation)

 
kirillka:

This is understandable.

The question is how to search for a spread with a minimum variance.

Of course you can try to do a brute force operation with some discrete step or really freeze and find solution for system of equations for finding minimum function by equating partial derivativesσ(Y(A1..A4)) to zero.I'm just sure there's a ready-made library to find the solution in a couple of lines of code). There's a discussion of the concept, but nothing is written about the implementation)

I'm not sure that the minimum variance is so necessary, it is more important to analyze the behavior of the spread curve

If you want, you may write a generator of spreads and check it on a cycle, the number of combinations can be searched numerically.

but it is a good idea to optimize the length and position of the calculation period window, because it strongly affects the appearance of the spread.

I do not quite understand the system of equations, there is only one equation (regression)

 

With 4 tickers and minimum step of 1% the number of runs will be 100^4/2.

There is a simple function globalMin.portfolio() in r for example, but I haven't found anything similar in Alglib.

UPD: In general there is no ready solution for minimum variance portfolio in Alglib, but it's possible to perform operations with matrices, so half-custody is already there. My question is removed.

 
Joker, how were you born with this system?
 
GerbertX:
Joker, how did you come up with this system?

Probably through hard and painstaking work. In the commonwealth of high knowledge of certain areas. He did not dream it. )))