Non-fitting system - main features - page 6
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
What's constructive then?
Constructive is where the nightstand is where the money is.
Let's not deviate from the topic. I am against discussions applicable in stationary markets and not applicable in non-stationary markets and I am only interested in non-stationary markets.
Constructive is where the nightstand is where the money lies.
Let's not deviate from the topic. I am against discussions applicable in stationary markets and not applicable in non-stationary markets and I am only interested in non-stationary markets.
>> originally the sabot is about stationarity in results in real markets.
eh.... I wish you would have stopped at the thesis: "Can't add anything"... but you didn't, ....
Repeated it! Not added. What to do - if you missed it (judging by the question). I don't mind.
what do you call .... just mathematics and geometry after all? I take it these are some kind of TA fields?
So why are you talking nonsense? If you use a mathematical apparatus in sociology (mathematical statistics, for example), it - mathematical statistics - does not become sociology. And where will you take the arithmetic mean (simple MA) - also out of TA? Or you will assert that arithmetic according to my logic is the field of TA?
In short, I am amazed by your brilliant arguments. I do not find any answer - I give up.)
===
One more time? Price, time, volume, open interest, tiers of the cup, etc., whatever you find in the flow, this is the object of TA. How to analyse it - TA methods.
Constructive is where the nightstand is where the money lies.
Let's not deviate from the topic. I'm against discussions that apply to stationary markets and not applicable to non-stationary markets and I'm only interested in non-stationary markets.
I don't seem to be getting an answer.
A constructive one.
And one more amateurish opinion of mine.
But if we consider deviations from a correctly identified trend (non-linear in general case), then IMHO the presence of stationarity is an assessment of the correctness of approximation.
I've already tried to talk about this.
I did! I didn't add. Well, what to do - if you missed it (judging by the question). I don't mind.
So why talk nonsense? If you use a mathematical apparatus in sociology (mathematical statistics, for example), it - mathematical statistics - does not become sociology. And where will you take the arithmetic mean (simple MA) - also out of TA? Or you will assert that arithmetic according to my logic is the field of TA?
In short, I am amazed by your brilliant arguments. I do not find any answer - I give up.)
===
One more time? Price, time, volume, open interest, tiers of the cup, etc., whatever you find in the flow, this is the object of TA. How to analyse this is the methods of TA.
Bullshit? And who wrote the nonsense: "Anything that analyses the content of the quote stream is a TA."
OK, I suggest we come to a compromise. - You may analyse it as you wish, I don't mind, I just take my hat off to your analytical skills. :о)
PS: I don't understand one thing - why did you keep repeating yourself?
The debate is about terminology and the issue is practical.
Do you have any real ideas on how to account for non-stationarity in the tester?
I have already corrected the first post by adding there:
"Robust TS must be (almost) stationary at different optimization areas with a constant lot by the following criteria: expected payoff and profit factor. At the same time it may be absolutely non-stationary by Z-score.
I.e. one should not pay any attention to such Z-Score dependencies as, for example, drawdown or number of simultaneous (without) losing trades. As Leov correctly pointed out, a suspiciously small drawdown may hint at a history fit."C-4 писал(а) >> Свинозавр, пожалуйста больше не комментируй мои посты. Просто я не могу вести аргументированые споры с дураками (никто не может).
Svinozavr wrote(a) >> You know, if you've said something stupid, it's the last thing you should do. You seem to be a sane person.
:)
On that optimistic note, let me return to the topic.
It would be nice to formulate all the requirements for such a system. So far there is only one. I would like to hear more opinions. There must be a lot of collective experience. All you need is the desire to share your knowledge. Otherwise it will be a clamour as usual.
Sounds like it. :)
Reshetov wrote(a) >> No part of the combinations will ever give a sustainable profit. Markets are non-stationary by nature.
Well, if you're talking about Single Layer Reshetron, then you're 100% right. :) Just like Minsky's famous work. This thing can't handle the XOR problem and its variants.
Reshetov wrote(a) >> So all sorts of mythical terminology about supposedly "stability" etc. >> Therefore all mythical terminology of supposed "stability" and other "guarantees" in this context is irrelevant at all.
What's appropriate, Yura? What then would be the criterion for a non-adaptive system, if notgreater robustness to changes in the trading context? :) Speak your word of wisdom already!
// You should not have started this topic in vain. What if someone else finds the solution? It could be simple... Terrible. That's awful. This is a suicide mission... /// Shh. :) :)
--
Ok, I will get to the constructive part.
Overcomingthe" overlearningphenomenon"[FP] (fitting to a dataset is its synonym) is obviously at some meta-level. That is, you need to understand it (FP) well,
understand "what it consists of", and most importantly "what its limitations consist of". If this succeeds, solutions may follow. I will permit myself an attempt at formalisation:
1) There is a set of indicators (or TS - doesn't matter at this stage)
2) Everybody sometimes tells the truth, sometimes lies , sometimes speaks the truth, sometimes babbles nonsense. (just like Reshetov. Or me - it makes no difference at this stage) :)
3) We need a set of indicators of truthfulness/ falsity context (2) for each indicator (TS) of Level 1.
4) You need techniques for pairing indicators of set 1 and set 2 to obtain thus set 3 - a set of adaptive, context-sensitive indicators (CSE).
5) Further, one can deepen the method by going back recursively to point 1, i.e. each new idiom (context-sensitive) is contextualised again to further enhance
"coefficient of rightness" (aka "coefficient of luck").
I.e. we need to learn how to highlight the rightness/ wrongness contexts for each and every signal. More precisely, to make such an allocator, teach it (set it up) and use it intelligently. That's all.
True, there is some distance between "device in principle" and "device in a box". :)
I have found some approaches for myself, but here I will keep silent like a plum. ;) I've already told you a lot... :) :)
I'll listen to what people have to say.
I don't think I'm going to get an answer.
A constructive one.
And another of my amateurish opinions.
But if we consider deviations from a correctly identified trend (non-linear in general), then IMHO the presence of stationarity is an assessment of the correctness of the approximation.
I have already tried to talk about this.
I once came across a paper with the following text: the initial premise is that BP is non-stationary. Take its model (it does not matter if it is stationary or non-stationary). The confidence interval of this model with respect to real BP is 0.97, etc. This is a theory. I don't think there is any need for such a mann.
We all work on patterns. If the pattern conditions give a signal, we enter. And it doesn't matter what is the next section where the pattern is executed: stationary or non-stationary. And so on. The problem is that our pattern can work once in a lifetime. Then we start talking about over-optimization, and we shouldn't do that. The next section can be different, you have to recognize that it's some new section, you can apply an old pattern to it, a new one, or we don't have a pattern at all. And the problem is recognising and anticipating transitions from one section of a non-stationary market to another. As it seems to me without adequate BP model this issue cannot be solved.
A vegetable garden? And who wrote the nonsense: "Anything that analyses the content of the quote stream is a TA."
Bullshit? ))) Then what is TA to you? Ok. If for you TA is a simple MA, then yes - bullshit. Why am I talking about it? It's maths - it's not TA. Then it's not clear at all. Oh... I get it - TA doesn't exist! No. It doesn't work either.
All right, I suggest we compromise. - You're welcome to analyse, I don't mind, just take my hat off to your analytical skills. :о)
Easily. You stop talking nonsense and we come to a compromise. Anyway, okay - hats off to you for your knowledge of the various ways of analysis. Technical!!!))
PS: one thing I have not understood, and why you something repeated?
To the fact that the quote flow is not used in solving geometry problems at school. You keep missing that somehow. Have you been drinking?