Machine learning in trading: theory, models, practice and algo-trading - page 1456

 
Maxim Dmitrievsky:
is a Galton board, not a neuron. They teach it in school.

I envy you, I didn't learn it at school.

I read the description, it should be a normal distribution, but the video shows a different effect - the distribution is almost uniform and the black balls are in the center.

 

a normal distribution can be anything, it does not have to be standard

almost equal - there is no such distribution

black balls are heavier, they are not centered because of color, take my word for it

 
Maxim Dmitrievsky:

a normal distribution can be anything, it does not have to be standard

almost equal - there is no such distribution

the black balls are heavier, they didn't hit the center because of the color, take my word for it

Yes it's clear that there is a peculiarity in these balls, that's why they are different - it's just a layout in which I saw the allocation of data.

 

About market volatility...
I'm trying to teach the forest to trade with fixed TP/SL.

Saw an interesting backtest (i.e., history matching)

Before Jan 17, 2017 the set TP=120 and SL=80 were bringing good profits, and after that they stopped working. Apparently the amplitude of price movements has changed, e.g. stopped reaching 120 pts from similar points in 2016. The forward was about the same as the whole 2017, i.e. 50/50

 
Maxim Dmitrievsky:

A normal distribution can be anything, it doesn't have to be standard.

Psychiatrist.

 
elibrarius:

About market volatility...
I'm trying to teach the forest to trade with fixed TP/SL.

Saw an interesting backtest (i.e., history matching)

Before Jan 17, 2017 the set TP=120 and SL=80 were bringing good profits, and after that they stopped working. Apparently the amplitude of price movements has changed, e.g. stopped reaching 120 pts from similar points in 2016. The forwards turned out about the same as all of 2017, i.e. 50/50.

This is a good "fit", you can already build a profitable strategy.
 
Sergey Chalyshev:
This is a good "fit", you can already build a profitable strategy.
Forward 50/50 is not an interesting strategy
 
elibrarius:
Forward 50/50 - it's not interesting to trade like that
You can't get any better from a neural network, apply a good mm and you'll be happy
 
Sergey Chalyshev:
You can't get the best from the neural network, apply a good mm and you will be happy.
You can't do it with averaging, there are too many losing trades in a row.
 
Elibrarius:
I can't average, there are too many losing trades in a row.

Judging by your picture, you got the normal distribution that so many people dream about.

It is necessary to apply a clever mm, rather than stupidly doubling up.

p.s. thank you is a lot, no need to thank me, the percentage on my account is just right ))