Machine learning in trading: theory, models, practice and algo-trading - page 732

 
Maxim Dmitrievsky:

I've been waiting for someone to sign up for a long time, I'll be glad to do it), that's why I urge them to hurry up and do it.

Unfortunately for my signal will not be able to subscribe. All that will be left is to watch and lick your eyes. Although it is possible to manually copy it now..... You need to know the scheme...

 
Mihail Marchukajtes:

Unfortunately, there is no way to subscribe to my signal. The only thing left will be to watch and lust. Although it is possible to copy it manually now.... You'd have to know the schematic...

this is very cruel of you

 

Yes, it could be tough, if Mikhail at least used normal data, but to get a non-random model for the market with 40 samples is as likely as the fact that the Earth is 6 thousand years old. We can all relax and be envious, while Michael collects 40 points a month and works with them using Reshetov's "vectorial machine".

And there the speed of learning in this vector machine is very slow, and he simply can not take large sets - will wait a month

 

Actually, the topic is exhausted. Alexander_K2, if not proven, has shown quite convincingly that everything is solved by the methods of statistical physics, and with acceptable results for practice. And at least not for years.

All sorts of MO may well be a nice addition to these methods, but not their basis. At least on amateur level it will hardly be possible to use MO.

 
Yuriy Asaulenko:

Actually, the topic is exhausted. Alexander_K2, if not proven, has shown quite convincingly that everything is solved by the methods of statistical physics, and with acceptable results for practice. And at least not for years.

All sorts of MO may well be a nice addition to these methods, but not their basis. At least at the amateur level, it is hardly possible to use MO.

There is no proof, not even a backtest.

 
Maxim Dmitrievsky:

There's no proof, not even a backtest.

He already has +75 on the real.)) Your belief in the backtest is a kind of religion. He, in fact, no one needs it.) There are other methods.

 
Yuriy Asaulenko:

He already has +75 on the real.)) Your belief in bktest is a kind of religion. He, in fact, no one needs it.) There are other methods.

So I do not believe, this is the minimum for the initial estimate of the applicability of a method in trading.

 
Maxim Dmitrievsky:

So I do not believe, this is the minimum that should be for the initial assessment of the applicability of a method to trade

For example, for my systems a couple of days on demo and a couple of days on real to understand that this will work fine for a few months.

It's true, I still carry out preliminary tests of a model, but 3-6 months on minutes. And systems on ticks.

 
My aim is to buy from the brokerage companies:

For example, for my systems a couple of days on demo and a couple of days on real to understand that it will work fine for a few months.

True, I still do preliminary tests, but 3-6 months on minutes. And systems on ticks.

To be able to prove something, I need some kind of history, that's what I mean.

And 20-30 trades a month is not enough.

 
Maxim Dmitrievsky:

In short, to prove anything, you need some kind of history, I mean it

And 20-30 trades a month is not enough.

I would test mine, of course, but statistically it is enough to say that the system works. Of course he was taking a risk:)).

In fact, it does not matter whether Alexander's system will work in the future, or not. Everything has already been shown. Well, he will fine-tune it, and that's all.