Machine learning in trading: theory, models, practice and algo-trading - page 2664
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
Yesterday I dreamt about a cool remedy for overfit (I also dreamt that I was picking blueberries on another planet, but let's omit that).
If it is a share included in the index - to make markup of deals taking into account this index, i.e. unidirectional deals should be in profit
For forex, take a package of correlated currencies and check the direction of trades on them. If there are losses on some of them, then reject the deal on the traded instrument.
The point is not to overtrain on noise, but to catch general trends.
Subtle? I tried it, it's pretty good. Now I'm trying to find the blueberries I picked.
Yesterday I dreamt about a cool cure for overfit (I also dreamt I was picking blueberries on another planet, but I'll leave that out)
If it is a share included in an index - to make markup of deals taking into account this index, i.e. unidirectional deals should be in profit
For forex, take a package of correlated currencies and check the direction of trades on them. If there are losses on some of them, then reject the deal on the instrument being traded.
The point is not to retrain on noise, but to catch general trends.
Subtle? I tried it, it's pretty good. Now I'm trying to find the blueberries I picked.
The unidirectionality of correlated instruments is a kind of trend strength. There's a logic to that, of course.
I made a simulation of buy and sell limits.... It is interesting that when the trend is down sellers are not allowed to sell but absorb buyers' limits and vice versa.
Is this situation inside your simulation or did you somehow save and compare data on price changes? If you did, did you do it on the stock exchange or forex?
Unidirectionality of correlated instruments is a certain strength of the trend. There is logic in this, of course.
A trend, which you have already seen post factum, is the same as looking at a mouse....
Is this situation inside your simulation or did you somehow save and compare data on price changes? If you did, did you do it on the stock exchange or forex?
inside
of a trend that you've already seen post facto, it's like looking at a car.
inside.
Mashka's got a delay.
That's what I'm talking about, if you see a trend, you see it with a delay, and it doesn't matter what you look at, whether it's Mashka or a tricky indicator
Through other pairs as a non-delayed noise filter, it's simple, there's no other sense there.
Through the other pairs as a non-delayed noise filter is obtained simply, there is no other sense there as if it were not provided
If the other pairs are more than 100 and their noise is Gaussian, you can do it
with a time code
https://youtu.be/wqD892r-wfo?t=345
On the same principle works Rendom Forest and other ensembles of rules, only there instead of noisy sinusoids, noisy rules, the sum of rules suppresses noise, this is the output of the model. And nothing ingenious))) the usual DSP of 100 years ago...))))
If there are more than 100 other pairs and their noise is Gaussian, you can.
time-coded
https://youtu.be/wqD892r-wfo?t=345
Rendom Forest and other ensembles of rules work on the same principle, only there instead of noisy sinusoids, noisy rules, the sum of rules suppresses noise, this is the output of the model. And nothing ingenious))) the usual DSP of 100 years ago...))))
Well, the point is a little different. Just remove atypical fluctuations from the markup that do not correlate with the rest of the symbols.
then the TS will sort of automatically work on the other symbols as well, if the signs are normalised correctly. But this does not exclude group fitting to history
as you rightly wrote, the more symbols, the more averaging.
There is a more generalised consequence here, when you can mark not only other tools, but also the chips we are training on, make a kind of filter on them. You can increase the correlation of the fiche with the target fiche through this. One way.