How do I make an automatic multiple-choice test? - page 2
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
It will disappear like that. I'm telling you, the EA is on a neural network and with each run it learns and improves its efficiency. Consequently, with each run the results are completely different.
OOPS! You are confused about something. There is no learning of the network on the test ))))
OPA! You are confused about something. There is no network training in testing ))))
I'm in testing my usual.
I've been on the road with my custom.
>> How's that?
How come? I'm not confused about anything. I may be new to MQL, but I know a thing or two. The Expert Advisor is called the Self-learning Expert Advisor and it is torn from the code base and slightly modified.
Here's even an abstract:
...program can be trained in two ways:
// A) As a result of preliminary testing on Strategy Tester....
Man, you live and learn, I've never encountered this before )))) >> Then it is not clear how it works in real trading, does it always learn?
Then it's not clear how it works on real, does it always learn?
On the real account it works using the ...\MetaTrader\expert\files folder, where files from the ...\MetaTrader\tester\files should be placed after testing. It reads them and overwrites them in case of an unprofitable trade. Yeah, well, it's learning all the time. In short, it's a full-fledged neuron.
Ohhhh, me too, that's why I asked for help. >> How do I test it for infinity? I've got a code for testing, but I don't know how to use it. I tried it all through my Expert Advisor, it did not want to be used. I added variables and defined constants. What do you think?
I can warn you right away about one pitfall that you may not be aware of - retraining. With this approach, there is a very high probability of overtraining the network. That is, it will learn the story too well that it will perform poorly (with a loss) on unfamiliar data(in the future).
I can warn you right away about one pitfall you may not be aware of - retraining. With this approach, there is a very high probability of overtraining the network. That is, it will learn the story too well that it will perform poorly (with a loss) on unfamiliar data(in the future).
No, it uses bar history as inputs, looks for analogy and outputs corridor. If the analogy and the corridor coincide - it opens. If there is a force majeure during support - it closes. At least in my demo with small training it shows very good results.
On the real account it works using folder ...\MetaTrader\expert\files, where files after testing from ...\MetaTrader\tester\files should be inserted beforehand. It reads them and overwrites them in case of an unprofitable trade. Yeah, well, it's learning all the time. In short, it's a full-fledged neuron.
You're getting a clean fit.
It's not worth the effort.