Floating market parameters - page 6

 
Rorschach:

Wavelets


What have you been laying out?

 
Valeriy Yastremskiy:

What have you been laying out?

Euros and random

 
Rorschach:

euros and randoms

Similar)))) random is even more stationary.)

 
Valeriy Yastremskiy:

Similar)))) random is even more stationary.)

Yeah, only the stationarity (volatility?) is different.
 
Rorschach:
Yeah, only the stationarity (volatility?) is different.

It turns out that you have to find the minimum areas to define the model and look at it in a sliding way. If it stops describing reliably, then ouch...

 
Valeriy Yastremskiy:

It turns out that you have to find the minimum areas to define the model and look at it in a sliding way. If it stops describing reliably, then ouch.

This is exactly the direction I'm looking at. I build the model on 100 bars sliding and look at 101 bars for prediction error, then I draw the average error, if it is less than the price increment, it means that the method removes uncertainty from the price.

 
Rorschach:

This is the direction I am looking at. I build the model on 100 bars sliding and look at 101 bars for prediction error, I get the average error, if it is less than the price increment, it means that the method removes uncertainty from the price.

It's an interesting idea to look at the probability of price range and if it goes out of the range change anything)))

On different TFs or on one?

 
Valeriy Yastremskiy:

Interesting idea, watch the probability of the price range, and if it goes out of the range change anything))))

On different TFs or on one?

On one
 
Rorschach:
on one

100 bars makes no sense. 120 - 132 makes more sense to me. 10 years, 2 years, quarter, 3 weeks, working time of the week)

There is something about zooming out))) Haven't found the truth yet. The sergeant's truth against the older TF is not something to go by, but maybe there is something)

 
Rorschach:

This is the direction I am looking at. I build a 100-bar moving model and look at the 101-bar prediction error, then I deduce the average error, if it is less than the price increments, it means that the method removes uncertainty from the price.

It will not work with bars (candlesticks), time sampling introduces a random component, you need to use other sampling methods.
Reason: