:)) Attempt number 2. Or let's speculate about the essence of earning, but without specifics ... :)) - page 3

 
Serg_ASV писал (а) >>

http://www.biznesit.ru/forex/nolimit/statement.htm

I don't see the plum...

It's a demo, not a real one, judging by the size of the lots. What kind of pipsing results can we talk about on the basis of a demo? Only the real one. Period.

 

A speech on the appropriation of market money (about the essence of "making money"):

What do you think you are, a trader?
You think you're making money? 1
You're upset why you didn't find out about the market earlier)))

- You're a swamp trader. - Market Sucks.
Hold your grudges. You see that river over there?
Why do you think there's water in it in summer when it's dry and it's not raining?
- Because (5th grade textbook) there is a swamp where water is held back, enough to last all year.
That's how you are a trader - when you get money, you think you've earned it).
But no, you haven't earned, you just took the money to give it back at an uneven hour.
Don't cry, don't sulk, you better study Natural History. It's easier than 19th century maths anyway. Maybe not all is lost, maybe you will become an ecologist....

 

A few more words about the components of Holy Grail. In its development, the necessary components are:

1. Correct source data. The normal bars we see in the terminal are the wrong data.

2. Correct signals. Well, there is nothing to add here, because 90% of activity on the forum is searching for the right signals.

3. Correct MM.

4. Correct testing that statistically almost excludes the possibility that the result of the first three points turned out to be a fitting.

MT4 tester is only one component of correct testing. And the optimizer may do the job - only if it is used correctly, understanding how the optimal parameters are related to the real trading: if we say that during optimization we must choose the area of TS parameters, in whose neighborhood the profitability is stable, we must have a real mechanism, in which exactly this stability is directly exploited. If this mechanism is not built into the trading algorithm, this optimization is worthless.

 
Mathemat писал (а) >>

A few more words about the components of Holy Grail. In its development, the necessary components are:

1. Correct source data. The normal bars we see in the terminal are the wrong data.

2. Correct signals. Well, there is nothing to add here, because 90% of activity on the forum is searching for the right signals.

3. Correct MM.

4. Correct testing that statistically almost excludes the possibility that the result of the first three points turned out to be a fitting.

MT4 tester is only one component of proper testing. And the optimizer may do the job - only if it is used correctly, understanding how the optimal parameters are related to the real trading: if we say that during optimization we must choose the area of TS parameters, in whose neighborhood the profitability is stable, we must have a real mechanism, in which exactly this stability is directly exploited. If this mechanism is not implemented in the trading algorithm, this optimization is of no value.

П.1

By and large, this is true. But we cannot change it. Therefore, a discussion - where to get or how to make "correct" data, it makes sense to postpone for a certain time ... :))

П.2

Exactly... That's what I mean too... If the sell signal will be given ten minutes before the fall of price to the minestop in 99% of cases, and the buy signal inversely, then it is a grail... Everything else is secondary.

П.3

Hmmm... testing is not from the area of ideal models... If the indicator for example will just accumulate the difference in points between the OP-signals and CL-signals... And this difference will grow... And it will be at least 1% of the theoretically possible total difference, then how to implement this strategy within limits and resistances of CA, how to avoid Mr-calls due to large drawdowns? All this is secondary... I have not seen a single strategy, which even pre-optimised with the same historical data, could be run through ALL history from, say, 1998 to 2008, so that there was not a single absolute drawdown of more than $5000 with, of course, no matter what the starting deposit....

That is, in essence - yes it's all right! But this is in fact the second question - "how to improve?" or "how to make it perfect?"

 
NProgrammer писал (а) >>

... I have not seen a single strategy, which even pre-optimized on THEIR historical data, could be run through ALL the history from, say, 1998 to 2008, so that there was not a single absolute drawdown of more than $5000

What did I get!!!!

 
Integer писал (а) >>

'What I've got!!!!'

I wasn't quite clear, but from the context of my words it was clear that firstly, the strategy should not have any money management, neither explicit nor implicit... And with what you've got there.

Short positions (% wins) 2238 (52.32%)
Long positions (% of winners) 2262 (51.19%)
Profitable trades (% of all) 2329 (51.76%)
Loss trades (% of all) 2171 (48.24%)

to have a $1000 drawdown is not normal... Try to cancel buying, and leave only selling and vice versa.

That's the first thing.

Secondly I'm talking about dates from 1999 to 2008 exactly in this range...

And in the third, simply put, it is important that the length of the code of such a strategy was let's say 100 or 200 lines in normal style... In a word, the idea should be.... Although if you have it exactly like that, then state the essence of the idea... And let's start discussing it. It's quite possible something else will be born...

Well, I'll tell you honestly, but don't be offended - it's not very impressive...

 
Prival писал (а) >>

Now that's interesting. If you don't mind. Your view exactly in that vein.

Well, I haven't got very far yet). But I am working, and on the implementation as well. My current idea of the grail is the following:

1) Expert Advisors that use clear criteria of opening, closing, and maintaining a position, at best, are tester grails. The market is constantly changing; therefore, the grail must constantly change, adjusting to these changes or even trying to predict them. Hence severe restrictions on the use of classic indicators or standard variants of their use and the need to develop a complex and flexible decision-making apparatus.

2) As much data and tools as possible must be considered to find current relationships and patterns, and to select from among them those that work, or are expected to work in the not too distant future. Of course the grail has to do it all for itself.

3) The best "friends" in building the grail should be terver and matstat.

If someone would like to share his ideas, it will be interesting to compare them and get food for thought.

 
Mathemat писал (а) >>

It's a demo, not a real one, judging by the size of the lots. What kind of pipsing results can we talk about on the basis of a demo? Only the real one. Period.

So, no one says that it is a "real" grail. ;-)

 
NProgrammer писал (а) >>

I have not seen a single strategy, which even pre-optimized on THEIR historical data, could be run through ALL the history from, say, 1998 to 2008, so that there was not once an absolute drawdown more

I want to be a little more specific ...

1. To make it clearer -- let's take a range of two months, and by this range we move from 1999 to 2008... And then we do this - we pimeline at this range.... see the result... We shift, optimize again, look at the results again... ...and so on. The criterion for winning is simple - not a single plummet in any of your attempts... You can optimize as much as you want... But the balance or the account balance after you close all positions must be positive... In this case you have to implement one strategy in your Expert Advisor's code ... But to optimize coefficients/parameters.... again - :)) You can do that as much as you want...

I haven't seen any such strategy yet. So optimization is not evil... And we should not be afraid of it. IMHO.

 
NProgrammer писал (а) >>

1. To make it clearer - let's take a range of two months, and this range moves from 1999 to 2008... In this case we do the following - we pimeline in this range.... see the result... We shift, optimize again, look at the results again... etc.

Well, yes, the forward analysis described by Pardo. And that is the real basis for parameter optimisation, as it is a working mechanism for exploiting that very sustainability.

By and large, it is. But we, it is beyond our power to change that. So discussion of where to get or how to make the "right" data, it makes sense to postpone for an indefinite time...

And I'm not going to discuss it. But it's not something we can't do. Neural networks produce data too - and no one is saying we can't do that.