Machine learning in trading: theory, models, practice and algo-trading - page 914
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 confess that I thought the data was crap, but it turned out not to be. There are some useful things in them. If there are no errors when unloading. There are no peeks or duplicates, it's fine for a buy.... exit...
This is for sell exit...
I'm sending you the script. It's not the one I use, but I made mine out of this one, so it's a good start... Good luck!!!
Always wondered about this forum. I have only mq4 among the files that are allowed to be published. Just drop me a line. I'll send it to you.
)))) in short, what do you say or not, that the results will be random with this approach. (in the long run)
and your long term graph shows that your overall expectation is less than zero - exactly the same if you trained your model over the entire period, it is stupidly unrobust
Try adding it up in your head again.
We just have a different view of the market, you think that long-lived models exist, while getting a model that will work on a long enough OM is comparable to a bad luck. I, on the contrary, believe that in the market there are extremely small periods of patterns that require constant model building. The difference is that I'm not sure that after a month it may raise something without drawdowns, but not without errors, because I'm sure I will earn it in the next two days.
so make a virtual tester and retrain constantly at least every hour, and you will see that it does not work in the long term ))
I confess that I thought the data was crap, but it turned out not to be. There are some useful things in them. If there are no errors when unloading. There are no peeks or duplicates, it's fine for a buy.... exit...
This is for sell exit...
I'm sending you the script. It's not the one I use, but I made mine out of this one, so it's a good start... Good luck!!!
Always wondered about this forum. I have only mq4 among the files that are allowed to be published. Just drop me a line. I'll send it to you.
Thanks! It`s interesting, I`m still doing the manual analysis, the results seem to be the same, I`ll post it later for comparison. But, the predictor arr_DonProc is good by idea, but for sale it strongly cuts the inputs. Maybe I should compress it, not 10 variants but 3?
Thank you for the script, I will now write in person.
However, I think the file is not deleted, but simply not attached - you can zip it and it will be fine.
We just have a different view of the market. You think that long-running models exist, but to get a model that will work on a long enough OOS is comparable to bad luck. I, on the contrary, believe that in the market there are very small periods of patterns that require constant model building. The difference is that after a month of model creation you can't be sure that it may raise raise against what I'm sure to earn in next two days, without heavy drawdowns, but not without errors...
Education month - minimum, well, would be more than 2-3 months. The test - 1-2 months - quite enough.
If less, it turns out that we are trying to guess the multiplication table, knowing the answers to only a small part of it. It is clear that we will not guess most of it.)
So make a virtual tester and retrain every hour, and you'll see that it does not work in the long term ))
There's no need to make any trouble. I work in online mode, and there you will see.....
Training for a month - minimum, it would be good to have more than 2-3 months. The test - 1-2 months - is enough.
If less, it turns out that we are trying to guess the multiplication table, knowing the answers to only a small part of it. It is clear that most of it will not be guessed.)
I say the quality of the model drops dramatically, while the model is adequate to the current market. It may stay for a couple of days making 5-10 deals on the plus side without too many mistakes. But the model knows much more about the market right now than the market itself (conditionally of course). Move to an older timeframe that's all, here's the clock and does not give me comfort last signal to sell. I think it will come when no one expects it in the market now....
Stop stealing my models :)
Come on... You saw my deal and stole it... Okay... I'm not greedy...
Right... I trade ONLY one instrument and that's the GBP. I wonder how people do it on several instruments at once...
Yeah, what's your opening time?
It does not matter, if we both are going to get elks :-)