Machine learning in trading: theory, models, practice and algo-trading - page 2619
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
is there anywhere I've tried to hack you?
Try it, you can certainly approximate the indicators, but the discrete logic with time gaps is not realistic to describe with a sliding window, it's a fact
The second model, which only learns to trade at those moments. You have to try it, it is not clear in advance
all clear)
I get it.)
Take responsibility for your own words and don't back down.
It's not that simple...
Profitable strategies don't trade in a sliding window, so you can't simulate them with MO because by standard AMOs work with tabular data, and tabular data is essentially a calculation of stuff in a sliding window...
Here's an example from the ceiling: Suppose it's a "Profitable Strategy": Wait for a breakout of the weekly low, then go back and wait for a candle configuration - enter...
How can you find such a pattern in the MO if you have table data, i.e. you look for the last n candlesticks, the answer is nothing.
Of course you can create traits for this "Profitable Strategy" to make it work, but you have to know this strategy to create traits for it and we don't know it...
There are only two algorithms that can solve these problems, maybe only one... But there is.
MO has a "+". When all the indices give a buy, MO remembers what was ok in that situation when it was sell. Purely statistics. But there is a "-" as well. Ideally, knowing the pattern makes things more interesting. To recognize the size of the pattern and the pattern itself you need a separate network. Which immediately exceeds the hardware and software (memory) capabilities of MT and will make training impractical in terms of time. And using third-party software in the Market is unacceptable, so we have to find a compromise. If we take a larger timeframe and fewer bars the picture loses its uniqueness. If we take a small TF and many bars then inertia of the main trend is lost. I prefer not to use any indicators in NS at all - it slows down the reaction time and leads to stereotypicality.
MO has a '+'. When all the indices give a buy MO remembers what was ok in that situation when it was sell. Purely statistics. But there is also a "-". Ideally, knowing the pattern makes things more interesting. To recognize the size of the pattern and the pattern itself you need a separate network. Which immediately exceeds the hardware and software (memory) capabilities of MT and will make training impractical in terms of time. And using third-party software in the Market is unacceptable, so we have to find a compromise. If we take a larger timeframe and fewer bars the picture loses its uniqueness. If we take a small TF and many bars then inertia of the main trend is lost. I prefer not to use any indicators in NS at all - it slows down the reaction time and leads to stereotypicality.
Ran that concept of mine... well, it's hard to grasp, the innards themselves, overfit
need to read the Prado articles the other day ) want a TC on the MO!
there's a ready-made
https://github.com/fernandodelacalle/adv-financial-ml-marcos-exercises
Ran that concept of mine... well, it's hard to grasp, the innards themselves, overfit
need to read the Prado articles the other day ) want a TC on the MO!
I think you first need to understand the flaws of using the sliding window for the market, and then you need a server to calculate something.