Optimisation and Out-of-Sample Testing. - page 2

 
leonid553:

In the light of the above, I see the following way :

To build a simple additional Expert Advisor, - and load all obtained sets of parameters into it after the first optimization.

Each set will have its own index. And then we simply insert this additional EA into the tester instead of the first one and optimize it beyond the sample, and the optimization parameter will be the LOCAL NUMBER of inserted sets!

It may be a bit tricky, but it's much better than manually out of sample...

The only thing we need is to consider the versatility of this add-on.

This is an interesting idea. As a variant of implementation: Expert Advisor in deinitadds its parameters and optimization criterion to the file intended for writing. After optimization is complete, the script sorts the data from the file by optimization criterion, keeps the set number of best parameter sets and writes them into the read-only file. When the optimization is launched, the Expert Advisor in the init reads this file, etc. In other words, you don't need an additional EA, but an additional script.
 
FION:
leonid553:

In the light of the above, this is the way we see it so far: ....

I think it won't work that easily, for each optimised parameter there will be several extrema identified when linked to other parameters . It may be possible to find a solution if these extrema are fed to the input of the neural network.

In the last case we optimize only the Number, nothing more!

And we just get what we need. Or I misunderstood your post?

 
Guys, I've had it all working for a long time now.
But under TradeStation, and not for free ... :))
I don't see any point in doing it under MT, we're not used to paying for work.
 
Mak:
Guys, I've had it all working for a long time now.
But under TradeStation, and not for free ... :))
I don't see any point in doing it under MT, we're not used to paying for work.


I'm almost done too)))) And you do not have to embed anything in the Expert Advisor - the programmer gets a file with a set of parameters
 
I implemented this under MT4 a long time ago.
It really allows to estimate perspectives of different systems soberly,
And get rid of illusions caused by overoptimization.
After this second optimisation, we are left with only those vAriAs that yielded a profit outside the sample! <br / translate="no"> As a result, ideally, we get "ideal parameters" for further work and testing online!

Strangely enough, but parameters that have a profit outside the sample are not always profitable. Other selection criteria are also needed.
 
Integer, do you mean a command like
terminal.exe "\tester\MyTests\MACDTest.ini"
in a loop with the .set parameter file itself, which also needs to be updated somehow (if we want to run, say, 1000 tests with different genes)?
 
Mathemat:
Integer, do you mean a command like
terminal.exe "\tester\MyTests\MACDTest.ini"
in a loop with the .set parameter file itself, which also needs to be updated somehow?


Sort of. An external program creates a .set file, runs the terminal, monitors the process, then throws in a new .set file, runs the terminal again for testing, parses the report after each test...
 
OK, the general idea is clear. Well, then the last question to all those who have implemented this project (i.e. Belford, Mak, Integer): is it worth it? Of course, it is nice to have an "optimizer" which does not only curve fit (like metaquote) but also tries to test strategy on out-of-sample data, but does it really deserve a higher score than MQ optimizer (which is also good, but only as a curve fitter)?
 
Mathemat:
OK, the general idea is clear. Well, then the last question to all those who have implemented this project (i.e. Belford, Mak, Integer): is it worth it? Of course, it is nice to have an "optimizer" which does not only curve fit (like metaquote) but also tries to test strategy on out-of-sample data, but does it really deserve a higher score than MQ optimizer (which is also good, but only as a curve fitter)?

Everything is good for the household. There is no point in comparing with MQ, because this program does not test itself, it only runs a tester
 
leonid553:
FION:
leonid553:

In the light of all this, this is the way we see it so far: ....

I think it will not be easy, for each optimized parameter in connection with other parameters several extrema will be detected. It may be possible to find a solution if these extrema are fed to the input .

In the last case we optimize only the Number, nothing more!

And we just get what we need. Or I misunderstood your post?

I meant identifying the best combination of all parameters without choosing them by equity or drawdown or profitability. The experience shows that optimization by one criterion does not always find the best combination, that is why a neural network based on multivariate analysis can give good results.