Machine learning in trading: theory, models, practice and algo-trading - page 721
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
I have only one theme left to finish, and it will be a grail, but there is a lot to do :)
Well, if anything about distributions or how to calculate the sample volume, or how to take into account the intensity of the trades - ask me. I'm an old man, and I have no secrets from the youth.
No, if there is anything about distributions, or how to calculate the sample volume, or how to take into account the intensity of trading - ask me. I'm an old man, and I have no secrets from young people.
No, there is a self-learning crap that studies the market, I'll describe it in more detail later, if I finish it.
Hi all!
I broke my head creating a training sample. I am trying to find the right direction to find good entry points or a direction to think.
The network I use seems to be able to predict on the basis of 5 candles the appearance of a rising bar, but it is not enough for a successful transaction. Stops and profits blow everything away.
How do I make a training sample determining the right moment to enter?
I marked all of the rising candles in the historical sample (took the tool SIH3-18 minutes) this formula:
I trained the network to predict the appearance of such a bar, but to enter is always higher than the Open and the stop-loss is higher than the profit and is often triggered. Most candlesticks after the open first tend downwards and only then start to grow and thus affecting stops first. So I am puzzling over how to show neuronics entry points that immediately go up.
I'm making a little progress, rebs.
on the auetsample.
Just to support the theme.
And a question right off the bat. What conclusions can you draw here? Is this a drawdown or has the model exhausted its capabilities????
And right away the question is a head-on one. What conclusions can you draw here??? Is this a slump or has the model exhausted its potential????
Oh, I don't know, I don't like the curve either. If you do that at least for 5 years to grow. Here it's a fluke.
What I know for sure and irrevocably - to use NS on their own marks is a total failure... chance that luck is 1 in a million. The whole topic is proof of that.
Oh, I don't know, I don't like the curve either. If you do that at least for 5 years to grow. Here it's a fluke.
What I know for sure and irrevocably - to use NS on their own marks is total failure... chance that luck is 1 in a million. The whole topic is a confirmation of that.
I actually asked why. When TC begins to plummet, the question arises, and whether it will recover? Will the plummeting stop and start the growth phase of the balance? Should we stick to the strategy or it is time to retest? These questions exactly arise at the moment of plummeting. Is it a drawdown or a total loss? This is just... Just thinking out loud...
That's why I asked. When the TS starts to drain, the question arises, will it be able to recover? Will the plummeting stop and begin the growth phase of the balance? Should we stick to the strategy or it's time to re-balance? These questions exactly arise at the moment of plummeting. Is it a drawdown or a total loss? This is just... Just thinking out loud...
You'll never know without crossvalidation.
I was watching the maximal number of losing trades per history, and if it was more than that, I would switch it off. In the end the drawdown is minimal, but there are not many profits. But my bot has opted for 3 months in tulle and trades over 1000, and still it is much weaker at OOS :)
But it is not so interesting since I have added memory cells, but it is still not the same, I have to make it more complicated.
I trained the network to predict the appearance of such a bar, but to enter is always higher than Open and the stop-loss is higher than the profit and is often triggered. Most of the candles after the opening first tend to go down and only then start to grow, and first touching the stops. That is why I am puzzling over how to show the neural network entry points that immediately go up.
First, it would be good to check this kind of opening strategy in the tester of mt5. If the tester shows losses, you should not train the neuronics, it will not be of any use.
For example I'm teaching the neural network to predict the price increase per bar (Open[0] - Open[1], Open[1]-Open[2], etc.) it is called regression. When you get a good result from neural network, you can hope to make a profit.
To begin with, it would be a good idea to test such a trade opening strategy in the mt5 tester. If even the tester shows losses with it, then you should not train the neuronics, it will be useless.
For example I'm teaching the neural network to predict the price increase per bar (Open[0] - Open[1], Open[1]-Open[2], etc.) it is called regression. When you get a good result from neuronics you can hope to make a profit.
It recognizes correctly, but I can not choose the right entry point. I want to teach the network to recognize not the candle's appearance, but the real entry point without falling down. It takes everything away from me at stops.
I cannot write such a condition. Maybe someone will tell me?
I can't check it in the tester, because it does not work there due to integration with neuronics. It's written in python and the information is exchanged through a file, and the tester doesn't create this file.
Do you think there is a prospect for automation: