Where is the line between fitting and actual patterns? - page 24

New comment
 
Vigor:

So the traditional approach is to take the MACD and look for a line between fit and real patterns? I only wrote that we should not look for patterns where there are none.

It is as simple as that. If a pattern is found, the Expert Advisor does not equal a loss. When parameters change, it also does not equal a loss.


1) The first part is answered in private.

2) Again not clear. Where does it not leak? Already on the real? Or on a test OSS?

You have to have a reasonable and presumed statistical advantage BEFORE you put the MTS on the real.

When the owls are gone, it is possible to analyse, but difficult.

 
lasso:

2) No. It is necessary to have a reasonable statutory advantage BEFORE putting the MTS on the real.

The daily no-loss Expert Advisor in fixed lot trading(the number of trades should be large - at least 2-3 per day) - this is already a statistical advantage over the spread, slippage and non-stationarity of BP.
 
IgorM:
Not losing an EA when trading a fixed lot is already a stat advantage over the spread, slippage and unsteadiness of BP


Sorry. Stereotypes again.

It's just that I'm exploring EA performance on several planes.

And by statistical advantage I mean "not only positive MO".

Many people here have already asked: -- Give me an EA with stable MO=spread*-10. ))

-10 is even cool, they just asked for a steadily draining one. No one gave.

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

Crisis and stingy people, however..... (c)

 
lasso:

just asked for a steady leaker. No one gave.

I'm not a problem to steadily lose - even an expert can sit out losses, not to mention people, the problem of identifying the moment (time) of making a decision to close a losing position is very relevant, the exit with a loss on the SL, the same adjustment on the history, as I wrote on the front pages - the optimal system should have a signal to enter and exit the market, and with this signal to decide which side to trade (BUY / SELL) - but at this stage, I have it all in theory so far (((
 

lasso:

to the question in the flowchart I provided

The alternative approach you suggest is just as unacceptable as the traditional approach in the absence of an expert with found patterns. But when it comes to tuning a "working" system (I think everyone has a couple or three working systems, but the drawdown/income level is not satisfying enough), what is called "traditional approach" in the diagram (2 OOS sections) allows to get rid of the over-optimisation effect better.
 
Vigor:
The alternative approach you suggest is just as unacceptable as the traditional approach in the absence of an expert with found patterns. But when it comes to tuning a "working" system(I think everyone has a couple or three working systems, but the drawdown/income level is not satisfying enough), what is called "traditional approach" in the scheme (2 OOS sections) allows to get rid of the over-optimisation effect better.

1) What is the advantage? Specifically, please.

2) You have not answered the central question:

lasso:
What super important information (or invaluable data) do you get by going through steps 2T and 3T and 4T,
and which cannot be obtained in any way at the point 2H with the alternative approach?
 
IgorM:
I can't lose consistently, even an Expert Advisor can wait out losses, not to mention people, the problem is in identifying the moment (time) when the decision to close a losing position is quite relevant, the exit with a loss on the SL, the same adjustment on the history, as I wrote on the first pages - the optimal system should have a signal to enter and exit the market, and with this signal to decide which way to trade (BUY/SELL) - but at this stage, I have it all in theory (((


An example of a stable losing EA, on 10 years of history with MO = spread * -7, well, at least 1000 trades (over 10 years) and a stable lot

STUDIO!!!

 
lasso:

What super important information (or invaluable data) do you get by going through steps 2T and 3T and 4T,
and which cannot in any way be obtained at 2H with the alternative approach?

Why complicate things. Double screening is always better than single screening. With single one you can match the FN combination to the first period of OOS. With double elimination the probability of fitting is reduced. You can go on like this... A successive tenth OOS will leave the studio with real working parameters from the original set of 1T sets.
 
Vigor:
Why complicate things. Double sifting is always better than single sifting. With a single one, you can match the FN combination to the first OOS period. With double elimination the probability of fitting is reduced. One can go on like this... With successive tenth OOS, the studio will be left with the actual working parameters from the original set of 1T sets.


You've got to be kidding me.

1)I've just made things as simple as possible. A point and two rays in different directions... How much simpler could it be?

.....

2) How much less likely is it to fit with a double dropout? And with a quadruple? How did you work that out? Formulas?!

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

3)What will you see in ten consecutive OOS that you can't see in 2H? Mind you, this is the central question....

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

Double drop,triple drop,decimal drop..... Oh, bollocks. You can keep doing that.

I'm going to bed.

 
lasso:


Example of a consistently losing EA, on 10 years of history with MO = spread * -7, well trades at least 1000 (over 10 years) and lot stable

TO THE STUDIO!!!

http://imglink.ru/pictures/23-01-11/0f344543065272980a134c23718d8968.jpg

http://imglink.ru/pictures/23-01-11/aac98709ba2f6a53965d01ee5059891d.jpg

Do you need one of these?

1...171819202122232425262728293031...64
New comment