From theory to practice - page 119

New comment
 
Vladimir:

.

A stably dysfunctional model is not usually used for real trading. And it does not create risks.

I am just discussing models that work, even work for a long time, but necessarily drain the deposit. Let's watch the signals. There are no other ones.

That's why the proof of an Expert Advisor's performance lies in a theoretical proof. For the discussed GARCH, it is making decision on the stationary series.


Let the address of Alpari (UK) Limited 201 Bishopsgate London, EC2M 3AB United Kingdom be on the small banana island of the UK, fine. However, I wouldn't argue with the view that the European peninsula gives bad licences either. It's a matter of taste.

Haven't dealt with the London one, but judging by the refund I got - it was an offshore with a London sign with all that implies.


What follows? No government regulation

  • no deposit guarantee (20,000 euros)
  • ban on the use of clients' money in dealer trading,
  • obligation to take out transactions on the external market (prohibition for brokerage firms to play against clients)
  • in the event of withdrawal refusal there is someone to push through and it is not some caufort.


Everything else is trivial, and the above are the main risks.

 
ILNUR777:
SanSanych, tell me honestly. These Archie GARCHs of yours. Are they making you money themselves?
If they have no predictive power, then why are you running around the forum with them. Or maybe you just "don't know how to cook them". Some claim that they use Fourier, although its direct application in its classic form has no real predictive ability in the market.

And if they work for you, what's the problem with proving it. It doesn't have to be strict mathematical after all. There are some that are not strict. Experimental ones after all.

I don't know how to do GARCH, I'm trying to set up a machine learning-like hangout.


Machine learning has a real EA - a mixture of a random forest and indicators. My indicators, as well as all others, are based on NOT stationary data, so they periodically strive to drain depo. Although, financial problems I had 2 years ago I managed to solve them in a year. I'm trying to replace it with GARCH, for which there is some evidence of future behaviour. But GARCH turned out to be too complicated.

So join GARCH.

 
СанСаныч Фоменко:

I don't know how to make GARCH, I'm trying to organise a hangout similar to machine learning.


But GARCH appeared too complicated.

So join GARCH.

Where's the logic in that? It didn't work for me, but you join in. I feel like I've been subtly sent away.

And if this is one way to implement it, why, without even having your own results on it, to preliminarily pass it off as predominantly better than others. I thought that since you approve it, you have something to say.
 
ILNUR777:
Where is the logic in this? It didn't work for me, but you join in. I feel like I've been sent out in a subtle way.

And if this is one way of implementing it, why, without even having your own results on it, why give it away in advance as predominantly better than the others. I thought that since you claim it, you have something to say.

Logic?

GARCH is mainstream on financial markets, I`ve been using it for some months, the result will appear, but not so quickly, as I thought at the beginning.

 
-You said Garch Arch Aphirma's power...
Everyone here is weak.

-Garch is an evil force.
The rich man comes, becomes poor.
Garch takes away power.
There, you're gone.

-Here, take the mashek.
Come on, take it! There's plenty to live on!

-No, what's good for the common man, is bad for the nerd.

 
ILNUR777:
-You said Garch Arch Aphirma's power...
Everyone here is weak.

-Garch is an evil force.
The rich man comes, becomes poor.
Garch takes away power.
There, you're gone.

-Here, take the mashek.
Come on, take it! There's plenty to live on!

-No, what's good for the common man, is bad for the nerd.


Do you want to make fun of me?

Well, well...

 
СанСаныч Фоменко:

Logic?

GARCH is mainstream in the financial markets, I've been doing it for a couple of months, there will definitely be results, but not as quickly as I thought at first.

I always had the impression that you've been going with them for a few years. And it's been a century since the topic was raised about adding R to MT, because it's more convenient to use them. Oh, come on. Most likely these models are not suitable in their pure form either. And that means making your own bikes anyway. And the advantages of these velopedal ones over others are not visible and it is not clear yet.
 
And now. Where's Alexander? Where are the deals. There's a mouse hanging from boredom in the arena.

I went into ticks to make a couple of trades a day. I ended up taking even longer than I would have if I had analysed the minutes. It's been a week. We'll never get Russian science off our knees like this.
 

And I'm in trouble - yesterday after the publication of Non-Farm data I should have had a great deal on AUDCAD, but because of an error in the program (see post #1141, where AUDCAD opening/closing commands were written to AUDCHF file...) I got negative on AUDCHF... This is what New Year's Eve drinking and complete stupefaction due to it... I'm going for a walk with my wife to VDNH now. If you see a man and his wife crying, that's me.

 
Vladimir:

Where am I to fall back on results obtained by simply measuring real data?

https://www.mql5.com/ru/forum/221552/page19#comment_6168925"Before multiplication, the values on the curves shown differed by a factor of 18, after multiplication they differed by a factor of 20. It's about the law of the square root."

To the unrealistic, or what?

Pretty sure your law of the root of t is only true for your method of data reception and no more. This law is not true for receiving 5-digit ticks and it is more accurate to use this exact formula for price deviation:

S = sqrt(N*mean(|Ask(t)-Ask(t-1)|)), where N is the number of ticks during observation time t.

1...112113114115116117118119120121122123124125126...1981
New comment