Machine learning in trading: theory, models, practice and algo-trading - page 502

 
Dmitry:

It is possible to fit the SB to the signs of the price range.

What is the point?

After fitting it statistically (characteristics) it will not differ from the real price series. But we know that there is no profit on SB.

That's why I asked about the differences. People here are serious, must know/see. I don't believe that >500 pages and such a high priority question has not been considered, but I couldn't find it by searching.

 
fxsaber:

After fitting statistically (characteristics) it will not differ from the real price. But we know that there is no profit in SB.

That's why I asked about the differences. People here are serious, they should know/see.


After the transformation it will be no longer SB

 
Dmitry:

After the conversion, it will no longer be SB.

Terminologically so, but in essence it will be a row without the possibility of profit.

 

It's all bullshit, fat tails, seasonality, etc. SB can easily be fitted by these simple statistical characteristics, but the MAIN thing is the Predictability of the future direction, sign and value, based on past increments and other non-linear functions (indicators) of the series and other series. In the market you can get 52-53% of correct predictions of direction and 60-70% of volatility, and SB is 50% random, that is (the market is losing the spread). Volatility can be seen "by eye", it's very seasonal and autocorrelated, and the direction... that's the whole point, everything there is complicated and a little bit.

 
fxsaber:
I would like to ask the respected participants of this branch

The question, as I see it, is key: how much does the real price data differ from the SB? If I understand correctly, the greater the difference, the more opportunities to squeeze out a profit. And vice versa, up to "no difference - no profit".


As a rule, the higher the difference, the higher the possibility to squeeze out profit. There are a lot of different quotes charts with different properties, compare for example a bitcoin and eurodollar chart and they will have different properties

 
Andrew:

It's all bullshit, fat tails, seasonality, etc. SB can easily be fitted by these simple statistical characteristics, but the MAIN thing is the Predictability of the future direction, sign and value, based on past increments and other non-linear functions (indicators) of the series and other series. In the market you can get 52-53% of correct predictions of direction and 60-70% of volatility, and SB is 50% random, that is (the market is losing the spread). Volatility can be seen "by eye", it's very seasonal and autocorrelated, and the direction... That's the point, everything is complicated and only a little bit.


))) This is the thread you need to go to.

Give me ANY portion of the SB and I'll give you such functions "indicators", which will give you not 52-53% but 70-80%

 
Dimitri:

))) This is exactly where you should be - you're like Maximka.

Give me ANY part of the SB and I'll give you the functions "indicators", which will give you not 52-53%, but 70-80%


Why me again? It's not my fault people are trying to argue with me and then they're so eh... well, maybe...

 
Maxim Dmitrievsky:

Why me again? It's not my fault people are trying to argue with me, and then they're like this... well, maybe.


))) I'm sorry. I'll erase.

 
@Maxim Dmitrievsky do you have odd logic on the whole RF or on each tree individually?

Respectfully.
 
Andrey Kisselyov:
@Maxim Dmitrievsky- do you have odd logic for the entire RF or for each tree separately?

With respect.

now only on the output - I pass through it a few targets, let's say from 3 bars, and it gives 1 cumulative result, and this result I teach the woods