Machine learning in trading: theory, models, practice and algo-trading - page 1375
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
Then check these points, I once had a good hit on them on Si in training, but it turned out that it is not realistic to close at 10:00 on the stop without pulling through, which greatly distorted the result.
Naturally, all this distorts the results. I should cut all this stuff out of my training and trading. And the NS will not be able to cope with the vespers on its own, either.
Now it's just a demonstration of the possibility of using only NS for quotes prediction. It seems to work).
Naturally, all this distorts the results. All this crap should be cut out, both from training and from trade. And the NS will not be able to cope with the vespers on its own either.
Now it's just a demonstration of the possibility of using only NS for quotes prediction. It seems to work).
If you do not know what to do with it, you may use it to make some profit.
I do not mind to listen to the details.)
I think it is possible to manage on Si on the evening, in terms of liquidity now is not so bad, it depends on the deposit of course.
And the fact, that it worked, it's good, I don't mind to hear about details :)
All code in front of you. Excluded only preparatory work loading data, etc.. Learning, as it was and remains, I think you worked with the original data. Similar.
About the details, I do not know what to say, ask.
Made a prototype of the TS on the NS. Closing the deal 5 min after opening (prediction time). There is no monitoring of the deal.
Here is the first result:
For x - trade number, for y - profit in pips. Commissions etc. are not considered. The test interval is 3.5 months.
There is no need to trade up to the 60th trade as it is before the previous futures closing, the forecast is not very much possible there. The sharp jumps are, I suspect, the intraday gaps.
Well, and Python code. It couldn't be simpler.
As someone said, the TS description should fit on a matchbox.)
As someone said, the description of the TS should fit on a matchbox)
This is not the most important, the most important is not the TS, but the trader's personal qualities, his psychology.
The most important thing is not the TS, but the trader's personal qualities and psychology. Some people have mash-ups with martin that work fine and bring in billions, while others are not helped by any forest with perseptrons, bad karma.
If karma is bad, then no matter what you do, you'll always be in trouble.)
Here are a lot of Python/R cheat sheets. Useful to have at hand so you don't have to flip through the folio
Good luck
All the code is in front of you. Only preparatory work loading data, etc. is excluded. Training, as it was and remains, you seem to have worked with the original data. Similar.
About the details, I do not know what to say, ask.
How stable is the NS training, is there a wide spread of results from training to training?
How steadily does the NS learn, is there a wide variation in results from training to training?
There is practically no variation. Training on a random sample of 5000 strings (have you seen such a live). The array itself is 55 - 60 thousand rows - the history of TF 1m 3.5 months. Test on it.
There is practically no scatter. Training on a random sample of 5000 strings (have you seen this one live). The array itself is 55-60 thousand rows - a story of 3.5 months. Test on it.
Why training on less than 10% of the whole sample, shouldn't increasing the sample lead to improvement?