Market etiquette or good manners in a minefield - page 63
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
The phrase about the NS alternative is unclear. Why did you come to that conclusion? And why is the condition of determinism in the market not important to you? And why is there no connection between the bars? What data do you take, for real trading, if not bar history?
I have conducted many statistical studies within the linear model to determine the existence of statistically significant relations between bars. The result is as follows:
1. Dependencies exist.
2. the existing dependencies increase as the TF decreases.
3.Existing dependences are weak (in the sense of outweighing the commission of DC).
Conclusion: there is no significant relationship between the bars (in the sense as stated above).
I did not use non-linear AR-models for the study. There are two reasons for that. One - they are too complicated for my understanding. The second is that circumstantial data indicates that there is no non-linear relationship between the bars. All this served as a motivation for switching to a "universal approximator" - NS.
I don't understand about the determinism of the market. If the market were completely non-deterministic, it would support the Efficient Market Hypothesis, which contradicts our presence in this Forum.
I use a vertical breakdown of the price series for forecasting.
I suppose it's more of an emotion than a real sense of the market with a "nerve-racking". Try to be more careful, hopefully the fate of the character depicted on your avatar should warn of a fundamentally possible danger.
You're probably right about emotions, but the fate of the "avatar hero" will one day be shared by all of us - one way or another.
You're probably right about the emotions, but the fate of the "avatar hero" will one day be shared by all of us - one way or another.
But some people want it to happen later.
But some people want it to happen later.
I join in -:) there is no need to rush.
to Neutron
Do you take the prediction error inside the cycle by epoch? Or are these two different procedures on the same graph?
Yep, and the statistics are dialed in from the outside.
As the experiment shows, there is always a risk of over-training the network and one will always have to visually control the optimal number of epochs of training. In addition, I have plotted the learning quality of the network on a test sample (that did not take part in the training) for the EURUSD charts:
![]()
It can be seen that as the number of inputs d of a single perseptron increases, the quality of the prediction increases (minimum of blue dots). But this process asymptotically tends to a constant and it is reasonable to limit the maximum number of perseptron inputs.
It can be seen that as the number of inputs d of a single perseptron increases, the quality of the prediction increases (minimum of blue dots). But this process asymptotically tends to a constant and it is reasonable to limit the maximum number of perseptron inputs.By the way, for some reason, as the number of inputs grows, the learning error grows and asymptotically approaches 1