Don't tell me then that TA doesn't work - page 25

New comment
 
Reshetov:


Today I found an interesting feature in the EA which is in GD2. It turns out that in the future you should not trade in 3rd mode (pass = 3), but in 1st or 2nd.....

...............................

It's all good. But I've done an experiment with your Expert Advisor but it was altered a little bit.

1. Let it be 23 hours today 2009.10.12 (trading platform time)

I optimize the Expert Advisor according to your initial methodology with extreme right

date 2009.10.13, i.e. I am adding last 23 H1 bars 2009.10.12 - day.

2. I set an option in my Expert Advisor: "close all positions at 23 hours" and run the Expert Advisor on the interval

2009.10.13 - 2009.10.14 - fix profit/loss

4. I proceed to point 1, increasing the date by one day and repeating everything.

--------------------------------------------------------------------------------------

I managed to work 40 days like this, 35 of these days turned out to be profitable.

I made 1000$ initial deposit and traded 0.1 lot with 1600$ profit.

---------------------------------------------------------------------------------------

Now I am building an optimization-automation program to trade on the next day after the far right date of trade optimization.

 
more:

I'm currently building an optimisation-fit automation program to trade on the next day after the far right trade optimisation date.


Will this do?



Added in version 2.1, which can be downloaded from the latest version downloads page: http://gold-dust.info/ru/downloads.

 
Reshetov:


Will this do?



Added in version 2.1, which can be downloaded from the latest downloads page: http://gold-dust.info/ru/downloads.

I think that's it, I'll try it now, thank you.

It would be nice to be able to insert your EA with its own controllable parameters for optimisation-fitting,

But I know, I know - send it to Job immediately.

 
more:

That seems to be it, I will try it now, thank you.

It would be nice to be able to insert your own EA with its own controllable parameters for optimization and fitting,

But I know, I know - send it immediately to Job.

The thing is, in order to make changes or additions to an existing EA, the whole program has to be recompiled.

That is, there is no universality in the form of the ability to insert any TC. It is not so easy to build it. If it were easy, I would have built it long ago - I also need it to experiment with different TCs myself.

 
Mathemat:

Why so harsh (I mean in blue)? Don't you think you can reject a good system in this way? Yeah, I get it: you seem like an extreme perfectionist...

that's something to aspire to. It's just that there is your own quality control on the system. As soon as random plots appear, the questions immediately arise "will I be able to figure out when such a plot will come in the future and what its duration will be?". I have no such answers, so I block the strategy.

imho, the sequence of trades expressed as "success-failure" is a Bernoulli process in the vast majority of cases. So how do you remove randomness from there?

But that's not what I'm analysing, i.e. not "success-failure" and I don't know how to answer your question. I think it's not very informative and says little about the trading process itself. When I test in MathCAD I get balance state in each interval (bar), i.e. my output is two final samples of equal length, one is a quote and the other is balance state(balance process). There are no transactions (entry, exit, success, failure, etc.) as such. I'm trying to understand the quality of the function of quote-to-balance conversion. If the "fractality" of the balance is close to the "fractality" of the market, then the system knows nothing about the market, it virtually copies it and is bound to lose.

PS: by the way, to be formally objective, the respected author of this thread confirmed TA's inoperability :o) I.e. TA as an independent discipline cannot answer the main question, where the price will go :o)

 
Reshetov:

The point is that in order to make changes or additions to an existing EA, the entire program needs to be recompiled.

I.e. universality in the form of possibility to insert any TS is completely missing. It is not so easy to build it. If it were easy, I would have created it long ago - I myself need it to experiment with different TS.

I see, that is, the name and parameters of the Expert Advisor, the algorithm and the sequence of substitution of these parameters in the tester at each call of the terminal are written in the program body.

I apologize for the intrusion, but maybe you can sew my program as well. Here are its parameters:

// PASS = 1 – подгонка  на периоде (2 + 3)(по Решетову) - получим значения TakeProfit и StopLoss и Stop_0,
//                      на периоде 2                 
// PASS = 2 – подгонка  на периоде 3       
// PASS = 3 – фильтрация путем отсева противоречивых сигнлов, поступающих от ТС, подогнаных на периоде 2 и не периоде 3
//            в режиме тестирования  без оптимизации или в режиме автотрейдинга на демонстрационном или реальном депозите
extern int    PASS = 1;
extern int    x11 = 100;  // оптимизация Start = 0 Step = 1 Stop = 200
extern int    x21 = 100;
extern int    x31 = 100;
extern int    x41 = 100;

extern int    p   = 20;   // оптимизация Start = 3 Step = 1 Stop = 100

extern int    x12 = 100;  // оптимизация Start = 0 Step = 1 Stop = 200
extern int    x22 = 100;
extern int    x32 = 100;
extern int    x42 = 100;
       int P1_bar = 0; // значение perceptron1() при открытие нулевого бара
       int P2_bar = 0; // значение perceptron2() при открытие нулевого бара


//--- 
// TakeProfit, StopLoss ,Stop_0, StopLossTrailDist, StopLossTrailStep  заданы для 4-х разрядных котировок, если котировки 5-ти разрядные,
// то программа сама это обнаруживает и умножает заданные величины на 10.
//---
extern int    TakeProfit  = 50;// 4-х разрядная котировка
extern int    StopLoss    = 50;// 4-х разрядная котировка
extern int    Stop_0      = 30;// 4-х разрядная котировка, при достижение любым из ордеров такого профита  в пунктах его стоп-лосс переносится в безубыток.
                               // Если Stoplevel не позволяет этого сделать, выдается сообщение и звуковой сигнал тревоги.
                               // Если Stop_0 = 0, то никаких действий по переносу StopLoss-уровня в безубыток не производим.

At the first step, parameters are optimized:

extern int TakeProfit = 50; - Start = 50 Step = 1 Stop = 200
extern int StopLoss = 50; - Start = 50 Step = 1 Stop = 100

extern int Stop_0 = 30; - Start = 30 Step = 1 Stop = 100

Of course, also your parameter

extern int p = 20; // optimization Start = 3 Step = 1 Stop = 100

The rest of the steps remain unchanged.

If you will do me this favor, I would like to attach the program itself, just in case.

Optimization, as usual, on the basis of the pore model.

*************************

Files:
testgolddust.mq4  33 kb
 
more:

I see, i.e. the name and parameters of the EA, the algorithm and sequence of substitution of these parameters in the tester at each call of the terminal are written in the program body.

I'm sorry to bother you, but maybe you can stitch my program too. Here are its parameters:

No, I'm not taking the job. It's a lot of work. If I could just plug in the program and everything would work, then I would do it. But because, for each step, we have to correct the EA settings in *.set and *.ini files, it is too bothersome.

I don't have time to make separate programs for each EA - everything goes to my TS.

Therefore, you should definitely go to Zhoba with your EAs.

 
Reshetov:

No, I'm not taking the job. It's a lot of work. If I could just plug in the program and everything would work, then I would do it. But because, for each step, we have to correct the EA settings in *.set and *.ini files, it is too bothersome.

I do not have time to make separate programs for each EA - everything goes to my TS.

Therefore, you should definitely go to Zhoba with your EAs.

I see, I'll go and have a look at it.
 
more:
Right, I'll go and have a look at it.
If you order for money, it is better to order something universal, so that any TC can be inserted and adjusted without any problems. Otherwise, you will need to change the TC and again: find a programmer, pay money, wait until the work is done, and so on until the pulse goes out.
 
Reshetov:
If you order for money, it's better to order something universal, so that any TC can be inserted and adjusted without any problems. Otherwise, you will need to change the TS and again: look for a programmer, pay money, wait for the execution, and so on until you lose heartbeat.

Something universal, of course. I used to program in C++ in Windows, so I don't really see

I don't see any problem creating a universal variant.

1. Menu № 1 - it is proposed in a digestible form, using the "calendar" to determine the periods of adjustment-optimization and FI

2. Menu #2 - you can run through the catalogues and choose the right Expert Advisor.

2. The selected EA is read, formed and displayed in Menu 3

3. Menu 3 - all variables are listed and asked to make

check/uncheck the boxes for participation/non-participation in the optimization fitting at certain periods of the fitting-optimization

4. After Menu № 3 processing all necessary *.set *.ini files are generated

5. The terminal is called ...as many times as necessary with the required parameters.

For me personally, only the question of transferring parameters to the terminal and tester remains somewhat unclear.

1...18192021222324252627
New comment