Testing real-time forecasting systems - page 82
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
Tick modelling... I think minutes are enough for profitable trading (exceptions - movements during the news, which is better not to work on). Our task, or rather your research, is to predict the direction, approximate trajectory and reference point in terms of the final calculation point.
Good luck with your research!
P.S. I am amazed at the very possibility of such predictions.
Conceptually I understand where the main error is - in inaccurate definition of the length of the historical time series, maximally influencing the future, i.e. "memory" of the series.
For some reason I got 4000-4500 bars optimum at min.error on test mn. and also res.forecast better on M15 than on H1 (visually on the chart)
Tick modelling... I think minutes are enough for profitable trading (exceptions - movements during the news, which is better not to work on). Our task, or rather your research, is to predict the direction, approximate trajectory and reference point in terms of the final calculation point.
Good luck with your research!
Thanks a lot for the wish! I am sure everything will work out :o)
P.S. I'm amazed at the very possibility of such predictions.
I have always said, one should not limit the nature in possibilities of its manifestation, moreover, often not understanding it completely. By limiting the nature (impossibility of something) we limit ourselves and then there is no chance to find something.
For some reason I got 4000-4500 bar optimum at min.error on test mn. and also res.forecast better on M15 than on H1 (visually on the chart)
I take a series of 25000 counts, in which I start searching for "memory". Turns out there are several levels of market memory, i.e. literally "short term" and "long term". A serious error comes out when forecasting short term (about a runoff) using "long term memory". Anyway, the theory is still in its infancy with me, working. :о)
I have always said that we should not limit nature in the possibilities of its manifestation, especially if we often do not understand it. Limiting the nature (impossibility of something) - thus we limit ourselves and then exactly, - there is no chance to find something.
Yeah, and without limiting it, there's a healthy chance of getting lost to death in endless variations of realisation.
There should be an optimum behaviour in this mode of search, and here is a question - what is this optimum?
That's roughly what I meant about the system "grabbing a long memory":
to Neutron
Also true, that's where gut feeling and intuition come in :o)
Good afternoon everyone!
Dax picture is too contradictory at the moment:
Most likely to open with a gap to the upside, then a small rise and, as the main move, a fall.
But I would refrain from trading today.
Hello! I am a newbie in this business! I decided to put my EUR/USD forecast I made, I hope somebody will point out my mistakes if there are any =)
I've decreased the number of bars and I've got such a forecast! The forecast of both indicators is identical, just pleases the eye, but how right they are time will tell
Can you tell me what to pay attention to when optimizing (training) in order of priority?
1.root mean square error on a training set
2.Maximum squared error of a learning set
3.percentage (or number) of recognised examples at a given minimum error value on the training set
4. the root mean square error on a test set
5.Maximum square error of a test set
6.percentage (or number) of recognized examples at a given minimum error value on a test set
Zar wrote >>
Can you tell me what to pay attention to when optimising (training) in order of priority?
1.Mean square error on the training set
2.Maximum quadratic error on thetraining set
3.The percentage (or amount) of recognized examples with a given minimum error value on the training set
4.The root mean square error on thetest set
5.Maximum squared error on thetest set
6.The percentage (or amount) of recognized examples for a given minimum error value on thetest set
On the optimal length of the training sample, at which the squv on the test sample is minimized when the squv error on the training sample reaches a given level.