Machine learning in trading: theory, models, practice and algo-trading - page 503
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
now only on the output - I pass several targets through it, let's say from 3 bars, and it gives 1 cumulative result
Regards.
Thanks, I got it. I put the odd one in each tree, see what happens.
Regards.
When I get back from Altai next week, I'll try to finalize the logic of the article, and finish it, and see what the result will be
I'll see how it turns out :)
))) This is where you need to go.
Give me ANY sample of SB and I'll give you the "indicators" which will give you not 52-53%, but 70-80%
Out of sample 70-80%? I can give you the SB, if you will set...
Although why bother with traffic, take SB on Gaussian noise, a simple cumulative sum, or geometric SB it doesn't matter, a million samples, do your alchemy with it, then take another piece of a million samples and try to model obtained on the first at least 51%, if you don't cheat somewhere and cheat yourself then the error variation will be around +-0.1% over 50%
Thanks understand. I wrote an odd one in each tree, see what happens.
Regards.
What's the lib, where did the wood come from? (trees)
There is a soft randomness in 2 shocks, there is a violent randomness in 3 and more. 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.
In fact we have to think about the meaning of the word "may". It is not obvious to everyone.
Watch at 14:10.
When I get back from Altai next week, I'll try to finalize the logic of the article, and finish it, and see what the result will be
I'll see what happens in the end :)
somebody needs bullshit too.
With respect.
out of sample 70-80%? I can give SB, if you will set it up...
Although why bother with the traffic, take a SB on Gaussian noise, a simple cumulative sum, or geometric SB it doesn't matter, a million samples, do your alchemy with it, then take another piece of a million samples and try to model the first one at least 51%, if you don't cheat and cheat yourself somewhere, the variation will be around 0.1% (50+-0.1%)
))) I'll take ANY piece of the SB range, break it down into 70+20+10% and pick up a set/model that will give 70%-80%.
You don't confuse "a chunk of a row" and generation.....
In fact, we should think about the meaning of the word "may". It's not obvious to everyone.
Watch from 2:10 p.m.
https://habrahabr.ru/company/itinvest/blog/280954/
here's some article, just came across (there's a link to the original article)
It seems that the participants of the branch do not know the differences between the (knowingly unpredictable) generated series and the price series.
Apparently, it is the Defense Department that is searching for these differences.
Don't poke me, I am not confused, I generate SB 2 million samples with the same stat characteristics, I give you a chunk of one million samples, you build a model on it, the second part of the series (1M) you do not see at all. To check how well your model will work on the second part, you can only check in realtime (via a socket, element by element, you forecast your next vector, etc.) or if an independent person gets your model and runs it on new data, on the second million samples. Other options are not valid.
PS: 70%-80% on SB without "tricks" is as realistic as to make a perpetual motion machine or a time machine
) I'll repeat my suggestion again:"Give me ANY part of SB and I'll choose for you such features "indicators" which will give you not 52-53% but 70-80%" (c).
P.S. Suggestion - give me an apple and I'll make a pie out of it. The answer - here's two nails and a piece of string and make four pieces of soap out of them.