Machine learning in trading: theory, models, practice and algo-trading - page 1820
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
increments, indicators, anything on the input. On the exit the direction of trades.
In addition to the direction correctly add expectation))) If the period after the trade is long enough and is a bit larger than the Stooploss, then it's a flat.
In addition to the direction correctly add expectation)))) If the period after the trade is long enough and a little more than a Stooploss in scope, then it's a flat.
it's all random, including the expectation. If you do it randomly many times, the pattern will be found according to the principle of large numbers.)
This is another selection block, not related to sampling. The maximum number of bars back is taken, and then the best combinations are selected. This is a problem of qualitative approximation, not sampling.
As a result, if properly sampled, the optimal number of lag predictors is selected that maximally describe the direction of trades.
Then it is monte carved and the best options are chosen. I don't think anyone has done it here, but I've had it for a long time.)
I don't think I can find anything for this task. I'll just try, random, regular, dependent on the indices.
It is unlikely that you can find one for your tasks here. Only try, random, regular, dependent on turkeys.
I googled multistep sampling. Choice of probability based on another probability, which is based on another... etc. I don't think there are any other options.
I.e., for example, then trades will open often, then rarely, then not at all. And the time of holding trades will vary greatly.
It may be related to volatility, no one knows how it should be done.
it's all random, including the expectation. If you do it many times at random, a pattern will be found by the principle of large numbers )
As long as it is stable / stationary. By logic, a pattern of clearly stationary series is not long, or rather the series is short term. A stable pattern on a long series is rare.
unless it is stable / stationary. Logically, the patterns of clearly stationary series are not long, or rather the series is short-term. And a stable pattern on a long series is rare.
For this is responsible 3rd block, which checks the quality of the model on new data)
the 3rd block is responsible for this, which checks the quality of the model on new data )
This is understandable, but we already have in our conclusions that there is no complete model, if MO makes casts and recognizes them, then it is for all series, and if patterns, then it is for certain series. If one is looking for an optimal sliding window learning model. then yes, but in any case the model falls on an unpredictable lag before learning.
In general the problem of lag minimization is eternal.
The problem is "is there a jeep in the picture".
The correct answer is NO jeep!
If you see it, then you either know how to see through the swamp, or you have schizophrenia.
I understand that when the brain sees the roof of the jeep it realizes that it is somewhere down in the swamp, but the question is "is there a jeep in the picture" in the picture there is no jeep, it is under the swamp.
It's like showing a network a picture of a field and making it say it sees a gopher, we know it's there for sure, but the picture is a field!!!
you need to think more, gentlemen...
That is, if the hood and windshield or half of jeep or full jeep but without back door would be seen on the picture, then it would not be on the picture either???? Just interested to hear an opinion...
I think the wrong questions for the wrong answers.
The task of the sorem NS is, like reflex, to solve without thinking about philosophy. That is the way modern NS pass old games (it does not even know what and why they do - and why it is necessary to be wary of opponents or jump over pits)
The most interesting thing is to simplify training before passing a new game - NS for some reason cleans..... although in my opinion the best gathered some GA for faster next learning in this direction
The problem is "is there a jeep in the picture".
The correct answer is NO jeep!
If you see it, then you either know how to see through the swamp, or you have schizophrenia.
I understand that when the brain sees the roof of the jeep it realizes that it is somewhere down in the swamp, but the question is "is there a jeep in the picture" in the picture there is no jeep, it is under the swamp.
It's like showing a network a picture of a field and making it say it sees a gopher, we know it's there for sure, but the picture is a field!!!
You have to think more, gentlemen...
With that approach, you won't get far. The problem is that in the market, if something is clearly visible, for example if you have recognized a trend, the money is already gone. But you have to learn to "recognize" the jeep just like on the picture. Or as above written, a man is sitting at a table in a room, but it is very likely that he is wearing slippers. The situation (context) is "so-so" The pattern (person)+context (room/table/evening) is the Grail. Correctly identify the pattern and the context in which it formed - that's it, nothing else is needed. While people are guessing, where is the euro going, the price goes up, down, collecting stops, but we are going up. Because in this flat "swamp" they have assumed "by a trunk with things sticking out" :))) With a very good probability (approximately 92% for the main pairs) the trend is up. And when the "jeep", i.e. the trend starts to show from the swamp, it is too late, only the blind and absolutely stupid would not understand, it is already time to go out with a profit.