Machine learning in trading: theory, models, practice and algo-trading - page 1222
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
MS IE 11
Got it, thanks, sorry, good luck with everything!
The market is a little different, there is no hierarchy, large investors do not consist of small investors, and those are even smaller, there is no connection between them, the big guys act their own way, the small guys their own way, the hedgers their own way, etc. There is fractality but no hierarchy.
The markets have the most brutal hierarchy, with reinforced concrete links between the "levels".
The markets have a brutal hierarchy, with reinforced concrete links between the "levels.
And in general, who is wagging who: Forex exchange or Forex exchange?
I tried RNN and LSTM, the results are mediocre, this is my personal experience, someone may succeed, I personally chilled to all neural network with backprop, very much trouble, unstable results, terribly slow, for CNN and RNN have their own niche, pictures, video, sound, speech, etc. and there is an explainable reason for this, namely, the hierarchical structure, one part of the third, and so on. In the market it is a bit different, there is no hierarchy, big investors do not consist of small ones and they have even smaller ones, there is no connection between them, the big guys have their own way, the small guys have their own way, the hedgers have their own way, etc. There is fractality but no hierarchy.
In general replication is very interesting in its context, I sometimes look into it, but its applicability to the market is skeptical at the moment.
You argue about deep learning and neural networks as a trend, but like, all our tasks can be solved by boosting without any problems.
About the hierarchy in financial markets it is questionable, but another hierarchy - boosting->forest->decision tree - is obvious to me, as well as related purely technical problems.
For example, targets in models as constants are stored in tree leaves and it makes no sense to put there continuous values like prices or signal levels, only discrete ones, like class labels.
At the same time predictors as well as constants are stored in branching nodes and limit possibilities of the model for extrapolation, and smoothing of intermediate results of its prediction can only take place at weighted voting in ensemble.
In other words, it turns out that by definition, we are limited to narrow classification boundaries, while with models even based on a primitive perseptron, we can already solve regression problems with multiple outputs. Even the main advantage - the speed of learning, due to fast, recursive algorithms of building trees, which has always accompanied boosting compared to neural networks, has recently become controversial.
Do you think - are these levels tied to the round forex numbers or do they "float" along with the CME prices?
So why do you need it?
Here's why. I have one order working (always at any time during the speculation, the lot is always the same - the minimum) as if the grid is working on the trend and on the rebound at the same time.
I'm worried about the red areas.
They appear when the price "accelerates" so that it is impossible to gain profit neither against a trend nor against a correction, because there is a
There is a historical noise (3000-4000 points), and there is no way to detect it with usual methods.
It took me a month to restore the first fixed drawdown (the first red square).
The robot always works the same way, there are only a few adaptive statistical parameters and that's it. The order system is always the same.
The only thing left here is the MO, since all parameters jump like crazy with such noise.
THIS IS AN EXAMPLE WITH NEGATIVE PROBABILITY.
I intentionally did it in my trading robot to clearly show the drawdowns.
This is how it works in real life:
By manipulating the probability I simply highlighted the weaknesses with the equity graph...
That's why.
You're not on salts, are you? ))
Leha went abroad, apparently to some hedge fund, now there is no one to troll you about the negative error and ZZ, they monitor even the phones and home computers, you can not bitch on the forums, you can go to jail.
Ha-ha-ha... So we believed it, your Alyosha is running away from angry investors, but they will find him abroad and fuck him, no doubt.
What the fuck 100% accuracy?))) They must have laughed at someone, probably after knocking on the wall with peas for a long time, 53-55% is all, stories about 70-90% - bullshit, well, how many times do I repeat ...
Glad to be wrong, but IMHO "different approach" does not exist, in this case, everything is mainly determined by the data, their quantity and quality
This is bullshit, demotivation of the forex-community, start again about qualified investors and their trillions, harvard and insiders.
In general, the main thing is to know the market conditions, when the trend is likely to occur, when the flat is going to happen, the channels, grids, etc. work during the flat, the trend is going to work with a wrecker.
PS: you should go... after Alyosha from here...
Show me an example...
You're not on salts, are you? ))
I don't see what's so hard about it... I don't know what salts you mean)
I've already made my own information recognition... object-signature matrices as far as possible...
I've only got a neural network to work with... It's just for fun really... just when you do something yourself you get a better idea of how it works)
I got used to the fact that there are only a few people who can really make something... the rest came here for pee-pee-pee... excuse my French.
PS: If you'll take this last phrase personally it means you'll get it for a reason;)