Algorithmic ''centrifuge'' - page 4

[Реter Konow]

Maxim Kuznetsov:

indicators considering a history of depth N can be represented as a functional product of SMA 1...N, therefore

even for a pair of elementary indicators with period 32, without taking into account constant coefficients and excluding symmetric solutions,

number of variations C(32,16)=601080390

live with it.

I think there is a mistake.

You should count the number of variations of input/output parameter assemblies based on the total number of available formulas and indicators in the base.

The number of configurations of parameters-components of Signals on the input/output is large enough, but not cosmic.

[Aleksey Mavrin]

Реter Konow:

I think there is an error.

The number of variations of Input/Output Signal parameter assemblies should be counted based on the total number of available formulas and indicators in the base.

The number of configurations of parameters-components of Signals I/O is quite large, but not cosmic.

I think at some point you will come to a key point on which a lot depends - the definition of GI for parameters, then look into my topic)

briefly:

If the simplest TS is stable, its profitability can be further improved and then you need additional pairs

Suppose initially there are 10 of them. to roughly perform 5 steps of each, two at the boundary of GU and three inside the range, we have 10k runs. multiply by the number of elementary strategies which is no more than 100.

I think that for many strategies we can manage with a number of pairs much less than 10

[Реter Konow]

Aleksey Mavrin:

I think at some point you'll come to a key point on which a lot depends - defining a GU for parameters, then check out my thread)

briefly:

If the simplest TS is stable, its profitability can be further improved and then additional pairs are needed

Suppose initially there are 10 of them. to roughly perform 5 steps of each, two at the boundary of GU and three inside the range, we have 10k runs. multiply by the number of elementary strategies which is no more than 100.

I think that for many strategies we can get by with a number of pairs much less than 10.

I will read your topic tomorrow.

If you do not want to complicate things too much, my concept involves the following:

1. Base indicators, formulas and equations calculating key for all TS Parameters should be placed in a common program.
2. Basic input/output signals configurations contain no more than 3 or 4 parameters (indicators/formulas) which confirm each other. Consequently, Optimisation should not make longer configurations.
   

[Dmitry Fedoseev]

Aleksey Mavrin:

I don't mean strategy construction itself - but a shell to automatically enumerate combinations of all subspecies each with each, including in the MT optimiser.

I just didn't find information about such results apart from ideas, but maybe it really has been done before and I wasn't looking hard enough.

Maybe I don't quite understand what the conversation is about. If you have to go through certain combinations rather than all parameters when optimising, then make an array of strings (array of parameter sets), and optimise the parameter set number.

[Igor Makanu]

Реter Konow:

I think there is an error.

The number of variations of Input/Output Signal parameter assemblies should be counted based on the total number of available formulas and indicators in the base.

The number of configurations of parameters-components of Signals to I/O is quite large, but not cosmic.

This is another topic, and I suspect that the source code will not be available as usual.

I recommend you install the terminal to solve some problems on the fly instead of discussing what came up ;)

#property copyright "IgorM"
#property link      "https://www.mql5.com/ru/users/igorm"
#property version   "1.00"
input  int param_1    = 1;
input  int param_2    = 1;

int OnInit()
  {
   return(INIT_SUCCEEDED);
  }

void OnDeinit(const int reason)
  {
  }

void OnTick()
  {
//---
   
  }

no thanks!

)))

[Aleksey Mavrin]

Dmitry Fedoseev:

Maybe I don't quite understand what the conversation is about. If during optimization we need to search not all parameters, but certain combinations, then make an array of strings (an array of parameter sets), and optimize the parameter set number.

Yes, not the parameters to search through (or rather not only), but combinations of blocks that are sub-strategies and from which the final strategy is compiled. Apparently you haven't read my link. quote below . if you read it, write your opinion, it's interesting.

Explanation: I havedecomposed the overall strategy view into the following elements

- a strategy for defining an entry opportunity

- strategy for defining an entry point

- a strategy for setting a stop loss

- strategy for setting profit target - take profit target

- maintenance strategy - trailing stop

- tracking strategy - volume management (topping up and / or grid and/or partial closing)

- maintenance strategy - defining early exit

For each sub-step of decision making I make collections of elementary (and not so) strategies

And I want to go through all possible combinations of each with each. That's why it's so complicated.

Igor, why open the terminal, it's more interesting to think without it )))

[Реter Konow]

For example:

We use 9 indicators for enumeration. The signals are made up of three parameters (indicators). A total of 27 variations of the signal (if I'm not mistaken, I'm tired). Each signal is a strategy.

In each strategy we are searching for the best values for the signal input and output parameters (we know the signal parameters themselves, we only need to find the best values).

In addition to values of the signal parameters, we are looking for values of stops and lots. Everything is as usual. Then we move on to the next signal And so on.

At the end, we compare the results of all signals and their parameter values, find the best ones and get the strategy.

[Maxim Kuznetsov]

Реter Konow:

I think there is an error.

The number of variations of Input/Output Signal parameter assemblies should be counted based on the total number of available formulas and indicators in the base.

The number of configurations of parameters-components of Signals to I/O is quite large, but not cosmic.

With maths as well as programming, it's a cinch...

601m - number of unique pairs of indicators on 32 bars, without their parameters. Maximum - number of 32x32 matrices excluding reflections and squared

if each parameter has 2 (at least - just scaling linear+distortion), with integer values from 1 to 9 inclusive, then additionally multiply by 10^4

[Aleksey Mavrin]

Maxim Kuznetsov:

the maths is just like programming - great...

601m is the number of unique indicator pairs on 32 bars, without their parameters. Maximum - number of 32x32 matrices excluding reflections and squared

If each parameter has 2 (at least - just scaling linear+distortion), with integer values from 1 to 9 inclusive, then additionally multiply by 10^4

I think it's also a question of the GU settinghttps://www.mql5.com/ru/forum/329028

and of all 600 million I would choose one combination - the first is an indicator of how much the price has changed, and the second is an indicator of how much the price has not changed)) Next, it's a question of their pairs.

[Реter Konow]

Maxim Kuznetsov:

with mathematics as well as with programming, it's a cinch...

601 million - number of unique indicator pairs on 32 bars, without their parameters. Maximum - number of 32x32 matrices excluding reflections and squared

If each parameter has 2 (at least - just scaling linear+distortion), with integer values from 1 to 9 inclusive, then additionally multiply by 10^4

(Really, that's great.))

You're counting something wrong. What do bars have to do with it? What do the internal parameters of the indicators have to do with it?

Each Indicator can be represented by ONE parameter that is placed in theMarket Entry/Exit Signal.

9 indicators create 9 CONFIGURATIONS of the entry/exit Signal, with 3 indicators included into ONE Signal.

Why 9 and not 27?

Because, variations (Indicator_1, Indicator_2, Indicator_3) inside the cue can be mixed -(Indicator_2, Indicator_3, Indicator_1) and the essence of the signal will not change.

Why 601 million unique pairs of Indicators?)

I repeat: INDICATOR AT THE TOTAL OF ALL CALCULATIONS GETS A END-PARAMETER SIGNIFICANCE. THIS IS THE FINAL PARAMETER THAT WE PUT IN THE SIGNAL AND ADJUST DURING THE OPTIMIZATION PROCESS.