Machine learning in trading: theory, models, practice and algo-trading - page 490
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
Three layers, each with 9 neurons. In the picture there is a very long section from 2004 to 2016. I chose the long history in order to check whether there will be a stable result, over the whole interval. In general, the drawdown on the forward is the largest, but on the other hand the bot started to earn on the second half of the forward.
In general, earnings in the second half of the forward is not obvious.
I trained for only 3 months (history -1 min TF - futures Sber - 3.17). Tested for 3 months - Sber 6.17 futures.
A simpler network - 15,15,15,10,1 (41 neurons) on the training sample gave awesome results, but on the next futures (Sber-6.17) the results are in the plus, but not so bad.
In general, earning a 2nd half forward is not obvious.
It's too early to tell right now. The horizon is big. I'll see in half a year, a year, after the test)
In four layers of net, stomping.
It's too early to tell right now. The horizon is a long one. In half a year, a year, I'll see for sure, after the test).
No, it has to be here and now. Really? - I need a working system now.
That's why, by the way, I choose a short training interval (3 months) and the same test. Imho, 1 min more and not needed. It is even unhealthy.))
In four layers of network, stomping.
Yes, the NS environment is what? R?
Yes, through random woods, very fast
Sincerely.
No, it has to be here and now. You don't say. - I need a working system now.
That's why, by the way, I choose a short training interval (3 months) and the same check. Imho, for 1 minute more and do not need.
Well, I do not know. A few months for the backtest will not be enough. So you can take some flea and good fit, but then it will be on the forward in most cases.
Forts are much more predictable than currencies. Of course it is not easy, but there is more chance for success on futures.
Yes, what is the NS environment? R?
MQL4)
I don't know. A few months will not be enough for the bestest. This way you can take some flea and make a good fit, but then it will pour out on the forwards in most cases.
Forts are much more predictable than currencies. Of course it is not easy, but there is more chance for success on futures.
I'm not pouring on forward, I've already written - 3 month forward - flight normal and no signs that the performance ends. 100 trades + training methodology.
So go to Forts if it's so predictable. I don't think so though. I've been there since '08.)
Actually profit per unit of investment is the same - Forts leverage is less ~1:10, but the moves are bigger. On forex the leverage is huge, but the movements are small.
So go to Forts, if it is so predictable. I don't think so, though. I've been there since '08.)
In fact the return on investment is the same - on Forts the leverage is lower, but the movement is greater. On forex, the leverage is huge, but the movements are small.
I do not have money for now, but I will trade there when I will have it.)
In general the most predictable markets are stocks.
I should correct, I have 9 weights and other in my network, and activation neurons turn out to be three, to be exact.