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
For example, you can simplify the task by taking an input number divisible by 2...
It seems to me that an odd number of inputs is more promising, however, I have a single input added to the number stated in the parameter.
Me too.
A few thoughts on methods of "NS predictive capabilities"
( below is a private opinion to date, in response to the expressed interest of the author of the thread in taking in inverted commas, possibly containing inaccuracies and errors. I would be grateful for pointing them out and discussing the topic)
The title of the chapter of Pastukhov's thesis is on the picture, I bring it in order to underline two, in my opinion, key words "pattern" and "prediction".
To the first, imho, it is necessary to add the word "temporal", i.e. each member of a pattern from N is "located" in time after the previous one.
The N-th term will always be +/-1 and can be ignored, because the one preceding it will be opposite in sign and will determine it
(except for the problem predicting probability of prolongation of trend, however the author of the branch initially settled on the problem before predicting the sign of movement)
Second, "predicting" - defines the class of tasks, namely not recognizing h.p. for example, but exactly "looking into" the future.
So we get the task PREVIEW on the basis of TIME LOOKING AHEAD.
Having analysed the pattern sets for various H multiples of the spread (I will try to post later), it can be noted that:
- repeatability is very low,
- starts to appear from H = several tens of spreads,
- grows insignificantly with increasing H, and, of course,
- decreases with increasing N.
These conclusions are indirectly confirmed by Fig.3.1 on page 99 of the thesis with small values of N+ and N-.
From which it can be concluded that there is a need:
1. unification of the pattern
2. using DM.
The first is seen to be possible by the following methods:
It is appropriate to choose the DM model on the basis of the above formulated task "PREPARATION on the basis of TIME PATTERNS".
The most suitable, or rather - intended for its solution, are the following:
- asst.rules,
- correlation rules and
- decision tree (to a lesser extent, but, as experience has shown, sometimes with interesting results).
I would like to speak about assertion rules separately as they are the most "targeted" DM-model.
The nuance is that the pattern is by definition sign-variable and the model learns to always look for the motion opposite to the last one.
It might be worth modifying the pattern from:
+1,-3,+2,-1
в
+1,-1,-1,-1,+1,+1,-1
But then there are a few questions, such as about the dimensionality of the pattern.
Perhaps increasing the result:
- creating a committee based on heterogeneous models
- Use of models based on different H patterns
- introduction of additional (non-price) inputs into the model (volume, etc.)
Concerning the latter I would like to hear the opinion of those reading this thread
Imho, interesting EURUSD market conditions
and the need to choose H with a stable sign strategy
To: Neutron к последнему посту в личке
Colleagues, I suggest that in further discussion of the topic, we should refrain from unmeasured values by rationing on the spread. Agree, the spread is a characteristic of a particular instrument and does not fully reflect the dynamics of quoting of one or another instrument. In the future, if necessary, let us normalize the values by H, and all that should be kept in the dimension, let it be expressed in points (integers). Michail, please comment your figure, what is the axis and what does it show? As for your statement about the need to "...choose H with a stable in sign strategy ...", then, as I understand it, we are talking about the need to choose such H, for which the condition of stability of alternating series of transactions is satisfied, see fig..:
Right?
X - time in hours within a day
Y - cumulative monotonicity of part of the kagi in H
on the right side vertically H dimension in pips discretely to the spread
this is intraday analogue of Fig.3.8 and 3.9 in the dissertation
Right?
Please explain the picture in more detail
The dissertation shows that the boundary of the sign change strategy is H-volatility = 2.
In the picture it is difficult to determine the value of the H-volatility.
How / on what basis do you estimate the indicated zone of a particular strategy sign?
I understand correctly, that in the picture the rent-function is red, and the ticks are blue?
Reference for general information: The "Tir-Method" has been on the market successfully for years, it also constructs a zig-zag pattern on the basis of 5 points.
PT breaks should lie on the kagi +/- spread edges,
in the figure they are under/above the kagi breaks.
Is it the quality of the picture or...?
If so, please make the picture more readable.