From theory to practice - page 1049

[sibirqk]

Aleksey Nikolayev:

This is essentially a move to a Markov chain from SB. The step, although small, is towards the real price - without Markovianness it is difficult to model a flop. The problem here is non-stationarity - instead of unpredictability of prices we now deal with unpredictability of model parameters. The hope is that these parameters change more slowly than prices - I wrote above about "Newton's first law".

Thus, some local inefficiencies can be formalised as time stretches with slowly changing parameters of some price models.

Great! The only thing left is to find a way to diagnose these local inefficiencies and TC's dream will come true.

[Aleksey Nikolayev]

sibirqk:

Great! The little that remains is to find a way to diagnose these local inefficiencies and TC's dream will come true.

SoonMaxim Dmitrievsky will fully automate this process and then the all-powerful bankers will finally get rid of traders constantly revealing their sinister plans)

However, it turns out that TS is a hired banker? I suggest he should be ashamed and repent)

[Дмитрий]

When you read the forum, you get scared - the same thing over and over again...

We discussed it 2 or 3 times on four, I remember two times on five.

And every time new clowns come out with such a clever look and with such trivial ideas, which have been chewed up many times...

[aleger]

Uladzimir Izerski:

Price is a reflection of people's emotions on events.

The reason for price rises and falls is to be found there, not in physical formulas. Physics does not allow physicists to become successful traders).

Who cares what causes the price to rise or fall, you need the correct reaction of the TS (or trader) to these changes!

[sibirqk]

Aleksey Nikolayev:

SoonMaxim Dmitrievsky will fully automate this process and then the all-powerful bankers will finally get rid of traders who are constantly revealing their sinister plans)

I have serious doubts about this. With methods of machine learning computer can of course teach a lot, but for this, in my opinion, you must know exactly what to teach, i.e. very clearly describe how to determine the local inefficiencies, otherwise it will strongly resemble the physical exercises of Sisyphus. But if you describe exactly how to determine them, then it is easier to code it directly in the algorithm. That is, until you figure it out with your own head, I don't think MO is of any help.

[Yuriy Asaulenko]

sibirqk:

I have serious doubts about this. With methods of machine learning computer of course can be taught very much, but for this, in my opinion, it is necessary to know exactly what to teach, i.e. to describe very clearly how to define local 'inefficiencies, otherwise it will strongly resemble physical exercises of citizen Sisyphus. But if you describe exactly how to define them, then it is easier to code them directly in the algorithm. That is, until you figure it out with your own head, I don't think MO is of any help.

Just from the text of the post, you are absolutely ignorant of MO methods and their applicability area.

But, there really is no substitute for the MoD's head.

[Aleksey Nikolayev]

sibirqk:

Something I have big doubts about. Of course it is possible to teach the computer by machine learning methods, but in my opinion, for this you must know exactly what to teach, i.e. to very clearly describe how to determine the local inefficiencies, otherwise it will strongly resemble the physical exercises of Sisyphus. But if you describe exactly how to determine them, then it is easier to code them directly in the algorithm. That is, until you figure it out with your own head, I don't think MO is of any help.

My point is a little different. In economic terms, traders do both good and harm. The harm can be quite significant - this has been talked about since the flash crash, for example. If the beneficial side of their work can be fully automated, independent traders may become a thing of the past.

[Evgeniy Kvasov]

What's up with the theme )))) It's a long way off))

[Yuriy Asaulenko]

vladevgeniy:
What's up with the theme )))) It's a long way off).

On the subject, Kolmogorov was replaced by Einstein and someone else.

[sibirqk]

Aleksey Nikolayev:

My point is a little different. In economic terms, traders do both good and harm. The harm can be quite significant - this has been talked about since the flash crash, for example. If the beneficial side of their work can be fully automated, independent traders may become a thing of the past.

The concepts of harm and benefit are extremely vague, poorly formalised and highly subjective. And flush crashes are on the contrary a manifestation of trading robots, their involuntary self-organization at a certain moment. It is a reminder that besides instrument's price there is also a concept of liquidity, and when liquidity disappears, the price of an asset changes abruptly.