Machine learning in trading: theory, models, practice and algo-trading - page 2662
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
But a human can see them, but a primitive algorithm cannot.
Of course, the market is not the same model as on the example of a ready-made elephant which was taken apart and put on the shelf in parts.
The very principle of construction is interesting.
A market chart consists of nodes and parts. The task is to correctly form the model. And the model must be known in advance. I have 2D models formed. I need to form a volumetric model. I can't find people who would be interested in it. There is an understanding of the model, but no such skills in software execution.
Of course, the market is not the same model as the example of a ready-made elephant which was disassembled and put on the shelf in parts.
The very principle of construction is interesting.
Yes my post was just about the fact that the disassembled is just very bad, a linear rsa would be much better I think, but the data of the elephant is unfortunately not available(
A market chart consists of nodes and parts. The task is to form the model correctly
yes, the task...
I have 2D models formed. I need to form a volumetric model.
why? if everything works, why? is there a third dimension in the model? is it necessary? .... ... ... ... ..
I don't find people who would be interested in this. There is an understanding of the model, but no such skills in software execution.
If you have made the products you have, you have enough skills, there is no desire to dive into something new....
And no one will do a complex idea better than the one who came up with it, or even no one at all except the first....
...
why? if everything works, why? is there a third dimension in the model? is it necessary? .... ... .... ... ... ..
Need.
2D is not a complete model.
Needed.
2d is not a complete model.
Well, if you want, you can tell me, we'll think about it.
Yeah... the algorithm is so bad...(( Or maybe I'm missing something...).
If it doesn't distort, it's still pretty good.
Yeah... the algorithm is so bad(( or I'm missing something...).
If it doesn't distort, then it's more or less okay.
Well that's too complicated a horse to predict :D
unfortunately for me, fortunately for the rest of the world, I don't know how to program anything)) I use what smart people have done as I can....
2d is not a complete model.
I may say the wrong thing, but this, as I understand it, is a trader's main problem with fitting any system to history.
Let's say we have a set of 1000 trades. If we open all the trades, for example, in Buy, we will get the result of 47% of profitable trades.
Now let's divide the aggregate into two parts of 500 deals by some feature (pattern), and open Buy in one part and Sell in the other, the result may improve, for example, up to 49%.
Now we have a patchwork, in which we open Buy or Sell with a very good overall result for 1000 deals.
And it seems that the division signs are chosen correctly, but the number in each "patchwork" is not statistically significant. And here it is, retraining.
I may say the wrong thing, but this, as I see it, is the main problem of a trader with fitting any system to history.
Let's say we have a set of 1000 deals. If we open all trades, for example, in Buy, we will get the result of 47% of profitable trades.
Now let's divide the aggregate into two parts of 500 deals by some feature (pattern), and open Buy in one part and Sell in the other, the result may improve, for example, to 49%.
Now we have a patchwork, in which we open Buy and Sell with a very good overall result for 1000 deals.
And it seems that the division signs are chosen correctly, but the number in each "patchwork" is not statistically significant. And here it is, retraining.
You need history.
You need it to understand the model. How price changes over time.
But the testing itself is an adjustment to history.
MA cannot anticipate events, but only relies on a flat chart from history.
We need to see the future of the price. The future is made of bricks of the past and will differ significantly from the past sizes of the pattern. It will depend on fundamental data or random events that MA does not yet know about.
The price model itself does not change, but the size and shape of the plots change, like when an elephant turns.
Therefore, it is desirable to have a 3-dimensional chart.
But here it is necessary to have a special view or a non-standard vision of the market. That's the way it is.
And the market, in my opinion, is predictable by much more than 50%, but there will still be problems of psychology in manual trading.