Machine learning in trading: theory, models, practice and algo-trading - page 1824

New comment
 
Maxim Dmitrievsky:
Look, well this is not an academic group on proving anything :) there are different approaches of working with time series, most of them are garbage on quotes. Unless proven otherwise, why the fuss?

that's it, I'm stopping ))

 
Maxim Dmitrievsky:

multistep sampling googled. Choice of probability is based on another probability, which is based on another...etc. I don't think there are any other options.

I.e., for example, then trades will open often, then rarely, then not at all. And the time of holding trades will vary greatly.

it may be related to volatility, no one knows how

Probability on probability cuts too much. A weak signal may be just noise, it should either be fixed for a long time or it should not be weak)

Volatility and volume, that's right, we should figure out how to fix it. Volatility comes in all sorts of ways. )

 
mytarmailS:

with the approach - the market is chaos

The market isn't chaos after all, and the noise isn't (always) white))))

 
mytarmailS:

that's it, I'm done ))

There are patterns that last for a year or two. Would I want to invest my money in that? Nope. It's still an uncertainty. It may work for a risky person who is willing to wait for the result.
 
Valeriy Yastremskiy:

probability to probability is thinning too much. Weak signal can be just noise, it either needs to be fixed for a long time, or it shouldn't be weak)

Volatility and volume, yes, we have to figure out how to screw it up. Volatility comes in all sorts of ways. )

I don't have more qualitative approaches, because there is not even a theoretical basis for them. And there's no Runtforms, except for lag signs.
 
mytarmailS:

Yes I understand, you don't understand me... you said that:

Not most, but all... And everyone works in a sliding window, and almost everyone with returnees, and everyone does NOT work with the market. I'm wondering why you don't ask yourself the question - why is that? The easiest thing to say is that the market is chaos, the hardest thing to change your mindset...

Any time series forecasting is based on the regularity of fluctuations, minus the trend. If there are no cycles, there is no forecasting.
 
mytarmailS:

What if the market is not a time series? but an event model? purely hypothetically...

The frequency of events is the cycle that carries the pattern
 
mytarmailS:

What if the market is not a time series? but an event pattern? purely hypothetically...


Look, we have this pattern, for example:

there's an extremum that the price breaks up, then the price breaks the extremum down-- what are we doing?

Did you retrace? - You killed the pattern.

you subtracted the trend? -- you killed the pattern.

you didn't normalize the price correctly? - you killed the pattern.

you look at the chart in the sliding window? -- you killed the pattern.

A pattern is a repeating pattern, no matter how you present it, with or without the trend
 
mytarmailS:

and this pattern in reality, not bad?

And this is considered its most elementary type, in two actions, and if the actions are 15?

So conclusions, what matters in the market is events, their sequence and their price.

What a mess.
 
mytarmailS:

Max, are you drinking there or what? ))

How do you see this pattern when you subtract the trend? Don't you even think about what I'm talking about?

If you subtract the trend from the same patterns, they will still be the same
1...181718181819182018211822182318241825182618271828182918301831...3498
New comment
Reason: