Discussion of article "Optimal approach to the development and analysis of trading systems" - page 3

[Aleksandr Masterskikh]

I like Eugene's articles - sincerity and simplicity in presenting complex problems!

I will only add the following from myself:

development and optimisation of trading systems should correspond to the nature of the market, namely it is a non-stationary process (not only the amplitude, but also the frequency of price fluctuations is constantly changing). To be a science, analytics must use an elementary structure, the parameters of which could be used at any moment of time.

And such a structure is not defined in traditional methods. That's why developers helplessly search for the grail, creating hundreds and thousands of lines of code, but the result is the same - profitable traders from 5 to 15% (regardless of the market - official statistics). The reason is the same - a scientific approach needs an "atom" (as in physics) and a "molecule" (as in chemistry) for analytics to become a science and not a guessing at coffee grounds.

[Evgeniy Ilin]

Aleksandr Masterskikh:

I like Eugene's articles - sincerity and simplicity in presenting complex issues!

I will only add the following from myself:

development and optimisation of trading systems should correspond to the nature of the market, namely it is a non-stationary process (not only the amplitude, but also the frequency of price fluctuations is constantly changing). To be a science, analytics should use an elementary structure, the parameters of which could be used at any moment of time.

And such a structure is not defined in traditional methods. That's why developers helplessly search for the grail, creating hundreds and thousands of code lines, and the result is the same - profitable traders from 5 to 15% (regardless of the market - official statistics). The reason is the same - a scientific approach needs an "atom" (as in physics) and a "molecule" (as in chemistry) for analytics to become a science, not a guessing at coffee grounds.

Thank you for your support, your book would be still to read tightly ) I so started a little bit, but I just do not have time yet. By the way, about profitable traders you should subtract random results ) in reality I think there will be 2 per cent of those who understand what they are doing ).

[Aleksey Vyazmikin]

The article demonstrates an interesting approach to finding a good basic algorithm. I understand that if such an algorithm is found, it is analysed and improved more deeply. I agree with the author that many people have been working with their idea for years and systematically develop and improve it, even if it does not bring income, but it is their own. I also relate to such people, but here it is just a question of the character of a person and it is difficult to simply give up such behaviour. In the end, the approach is good if you have the ability to generate a lot of different ideas that give a signal to enter the market.

Regarding the metric indicators of testing results, I agree in general, but it is necessary to specify that they are valid for a fixed lot, in the case of network building (gradual accumulation of positions) the interpretation and significance of the indicators will change.

Regarding the code - lately I'm seeing more and more bugs in the strategy tester, and that's why I think it's desirable to make checks. But the code is secondary here, and you can always find something to pick on. I don't understand why forum members can't be kinder and instead of attacking and helping to correct a clear error in the code or logic - why such aggression - I don't understand.

[Igor Makanu]

Aleksey Vyazmikin:

Regarding the code - lately I've been seeing more and more bugs in the strategy tester, so I think it's advisable to make checks. But the code is secondary here, and you can always find something to complain about. I don't understand why forum members can't be kinder and instead of attacking and helping, fix a clear error in the code or logic - why such aggression - I don't understand.

OK, let's omit the code, we consider it to be an original writing of the code, as they say MQL-style.

Can you give us your opinion about formulas? - This is the section "Mathematics of Optimal Search".

[Aleksey Vyazmikin]

Igor Makanu:

can you give us your opinion on the formulas? - This is a section called "The Mathematics of Optimal Search".

I did not analyse the formulas themselves and did not check for correctness of their construction.

As for the logic for their construction, I have already written earlier that this is an approach that has a right to life if its user has the ability to generate new ideas.

Of course, any Expert Advisor has its own logic, a set of functions that pass from one Expert Advisor to another and they are constants, but the additional logic that generates market entry can be different, and I agree that if it brings profit at the initial stage with simple rules and a sufficient number of entries, the chances to squeeze more profit out of it will be greater with the same number of additional inequalities (lines of code) to the basic strategy. This genetic selection of ideas simply does not mean that the discarded ideas are bad, it means that it will take more time to develop (evolve) them, while humans have a limited supply of it.

[Denis Kirichenko]

It was this passage that got me thinking:

 В итоге все сводится к некой зависимости, достоверно определить которую невозможно, но можно примерно прощупать путем опытов:

• Ps=Ps(L)

Where "L" is the number of lines of working code. In other words, the probability that the first indicators of the currently created system will be in an acceptable range depends directly on the amount of code we have written. It may seem to many people that the more lines the better the system is, but in fact it is not always so. It should be understood that

• T=K*L

The more lines, the longer development takes, but we should think first of all not about how many lines are in our code and whether we will not be killed for 2 lines of code that work like 2000, but about how efficient our code is and how many acceptable systems we can write per unit of time. After all, all other indicators will depend on this indicator:

• E= Ps/T
• E --> Max

This is a rather bold statement. Imho, induction - from particular to general - takes place here: the author has found something small in code size and profitable and concludes that it is right... but who knows? Maybe it's just an accident. To make such bold conclusions, you need a whole statistical research here...

Another stylistic question. I wonder what universities teach how to present formulas unmarked in this way? For example, what is K?

[Denis Kirichenko]

Forum on trading, automated trading systems and testing trading strategies

Discussion of the article "Optimal approach to the development and analysis of trading systems"

fxsaber, 2020.10.24 12:08 pm

I have not watched the article yet. I don't understand this comment. I don't practice checks for Tester either. And I do it intentionally.

Colleague, come on. Even without checks is a normal record?

void CalcAllMQL5Values()//array recalculation
  {
   ArraySetAsSeries(High, false);
   ArraySetAsSeries(Low, false);
   ArraySetAsSeries(Close, false);
   ArraySetAsSeries(Open, false);
   ArraySetAsSeries(Time, false);
   ArraySetAsSeries(Volume, false);
   CopyHigh(_Symbol, _Period, 0, CandlesE, High);
   CopyLow(_Symbol, _Period, 0, CandlesE, Low);
   CopyClose(_Symbol, _Period, 0, CandlesE, Close);
   CopyOpen(_Symbol, _Period, 0, CandlesE, Open);
   CopyTime(_Symbol, _Period, 0, CandlesE, Time);
   CopyTickVolume(_Symbol, _Period, 0, CandlesE, Volume);
   ArraySetAsSeries(High, true);
   ArraySetAsSeries(Low, true);
   ArraySetAsSeries(Close, true);
   ArraySetAsSeries(Open, true);
   ArraySetAsSeries(Time, true);
   ArraySetAsSeries(Volume, true);
  }

It is specified in the Documentation:

...Неважно, какое свойство имеет приемный массив - as_series=true или as_series=false, данные будут скопированы таким образом, что самый старый по времени элемент будет в начале физической памяти, отведенной под массив. Существует 3 варианта функции...

So 12 lines of code are meaningless. This is a question of efficiency and optimality :-)

The last 6 lines should be simply moved to the DimensionAllMQL5Values() function.

[Denis Kirichenko]

Eugene, by the way, there is such a function CopyRates().

Yes, to the question of prototypes. There is a wonderful article "Prototype of a trading robot". If you want to program effectively, of course, and not crumble an olive salad ;-)

I don't know, maybe I haven't reached the level of understanding yet.... but so far there is such a feeling that somehow linking the number of lines of code with the profitability of an algorithm is like putting warm and soft....

[secret]

Eugene, did you by any chance buy market training materials from reputable sellers)?

[fxsaber]

Denis Kirichenko:

Colleague, come on. Even without any checks, it's a good record?

I don't work with bars, so I don't know.

ZЫ Yesterday I came across that some standard symbols were giving out excessive profits on real ticks. I made a custom symbol from real ticks. The grail disappeared...