To follow up - page 44
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 I like about Bateson is the stimulus-reactive terminology.
That is, in stock market parlance, there is no obligation on the trading system to predict anything.
Only to make hypotheses and make bets, collecting statistics by trial and error. // a genetic algorithm in our parlance.
It is also well observed that the trading error can be interpreted in two ways: either as a response error to the stimulus or as a context classification error.
And it is rightly added that this dilemma is insoluble analytically.
// Which, in fact, is the main problem (imho). And not the system of phase coordinates at all.
// About the self-reinforcing traps of learning-II - a song. This phenomenon is observed all the time in the surrounding society.
// I call them "bias traps" among myself. (Well, this is my favourite topic. I will speak out again, maybe).
Thus, in the proposed language the problem is formulated as follows: It is required to build a computer model, capable of decent multidimensional Learning-II.
Assuming that most traders (including large ones) play in "flatter" models, such a model may have a statistical advantage in the market.
That's enough for now. I'll wait for fresh thoughts from the people. :)
I don't really know what this thread is about... it's just a random movement of incoherent thoughts...
Что касается отклонения остатков машки, то лично мне хотелось бы для начала увидеть собственно распределение этих остатков. Вы такое делали ? Оно действительно близко к нормальному ? Неплохо бы также понять какой смысл в таком отклонении. Иначе что будет означать такой показатель ?
The vine has budded.
I've budded a vine.
Unfortunately, I can't say anything. It is not clear how distributions, especially the lower ones, what amount of data was used for that, what does it have to do with ZZ and how it was used. When it comes to practical operations, (personally) I need a more complete description to understand.
And what do you say about Neutron's remark there ? His graphs seem to suggest otherwise.
К сожалению, ничего сказать не могу. Непонятно как строились распределения, в особенности нижние, какие объемы данных для этого использовались, при чем там ЗЗ и как он использовался. Когда речь идет о практических действиях, то (лично мне) нужно для понимания более полное описание.
А что вы скажете о замечании Neutron'а там ? Его графики вроде говорят об обратном.
I suggest that these subtleties be discussed there, and report here only the conclusions.
-- ZZ is not yet, but its place in the future evaluation is obvious. the data preparation is known from the previously written posts, I have only changed the conclusion:
aSpeed=iMA(NameVal,Tf1,15,0,0,1,i);
FileWrite(hFile,i,NameVal,dataConv(iTime(NameVal,Tf1,i)),
iOpen(NameVal,Tf1,i),iClose(NameVal,Tf1,i),
iLow(NameVal,Tf1,i),iHigh(NameVal,Tf1,i),
iVolume(NameVal,Tf1,i),aS[i],
aForecast[i],aZZ,bS
, (aForecast[i]-aZZ),
(aForecast[i]-iClose(NameVal,Tf1,i)),
(aZZ-iClose(NameVal,Tf1,i)),
aVol,aSpeed,
aVol-iOpen(NameVal,Tf1,i),
aSpeed-iOpen(NameVal,Tf1,i)
);
There are 5488 observations.
That's all for now.
--
Ran the minutes. 98331 observations.
Didn't test the hypotheses, but I did post the frequency pictures.
Closer to normal as far as I'm concerned.
Таким образом, на предложенном языке задача формулируется так: Требуется построить компьютерную модель, способную к приличному многомерному Обучению-II.
Исходя из предположения, что большинство трейдеров (включая крупных) играют в "более плоских" моделях, такая модель может иметь статпреимущество на рынке.
I'm afraid the market is behaving precisely so as to make learning as difficult as possible. There is a danger that such a model will go mad.
The article also has this to say: "One gets the impression that Training-II is a necessary preparation for a mental disorder". :)
Боюсь рынок ведёт себя именно так, чтобы в максимальной степени затруднить обучение. Есть опасность, что у такой модели поедет крыша.
Об этом в статье тоже есть: "Создается впечатление, что Обучение-II является необходимым приготовлением к психическому расстройству". :)
So?
Conclusions ?!
;-)
Выводы ?!
What are the conclusions now. Here's one in three years :) .
In general, the language of the article is quite unfamiliar to me. Without examples I wouldn't have understood anything at all. And with examples I understood almost nothing :)
I think that first of all it is necessary to do Learning I, like a fetus goes through stages of evolution during development. What to consider Learning I depends on the approach, it may be the ability to find entries and exits in TRS, or a trading context in TRS. If we stop there, the bot's psyche will not be threatened because it is absent.
Most likely, in order to trade more or less constantly, several "patterns" are needed, in the RTF these may be parallel entry-exit systems, in the TRS it is impossible to immediately formulate, different sorts of entry-exits.
Finally, it would be good to teach the bot to switch behaviours according to the situation. Can you call it a Learning II, I can not think, but that here the bot can go crazy easily I can imagine.
And in general it would be interesting to hear from an expert :)
Ну какие сейчас выводы. Вот года через 3 :) .
Вообще язык статьи для меня достаточно непривычен. Без примеров я бы вообще ничего не понял. А с примерами я почти ничего не понял :)
Надо полагать, сначала нужно таки сначала делать Обучение I, типа как зародыш при развитии проходит стадии эволюции. Что считать Обучением I зависит от подхода, можно умение находить входы-выходы в ИВВ, можно торговый контекст в РВВ. Если на этом остановиться, психике бота ничего угрожать не будет, по причине её отсутствия.
Скорее всего чтобы торговать более или менее постоянно понадобится несколько "моделей поведения", в РВВ это могут быть параллельные системы входов-выходов, в ИВВ так сразу и не сформулируешь, разные сорта, что ли, входов-выходов.
2) И, наконец, хорошо бы научить бота переключать модели поведения по ситуации. Можно ли это назвать Обучением II я сообразить никак не могу, но что здесь крыша у бота может поехать легко могу представить.
А вообще было бы интересно специалиста послушать :)
:)
Learning-I don't threaten the bot's psyche, that's true. But it does threaten the deposit. Examples are not needed here, they are at every turn, because that's how the vast (overwhelming?) majority plays. :)
2) Yes, this can and should (in the context of the article) be called training-II. And how the bot can go crazy is not difficult to imagine. That's if you make it necessarily choose the right context.
But it seems to me that it doesn't have to. Even averaging (after predicting all variants) in most cases will be enough to maintain stat advantage.
// Although it is possible and desirable to come up with something better.
People believe (for the most part) in the existence of the right context and the need to act in that context (the dominant principle). That is why they (people) have inner conflicts that are hard to bear. I call it a lack of a sense of humour. :)
A bot doesn't have to be trained in the absence of it. ;) It's better to have extra availability... // I mean, contextual conflict zones may turn out to be especially interesting for trading.
So here are the considerations.
To follow up with an anecdote (contextual shift demonstration):
"
- I hear you're looking for a new accountant?
- Yes. And the old one too..."
;-)
Практикующий маг не отучается от своего магического видения событий даже если его магия не работает. Фактически, утверждения, управляющие пунктуацией, имеют общую характеристику само-подтверждаться [9]. То, что мы называем "контекстом", включает поведение субъекта, равно как и внешние события. Но это поведение контролируется предыдущим Обучением II, и поэтому оно будет такого сорта, чтобы загнать общий контекст в шаблон, соответствующий ожидаемой пунктуации. В целом, эта само-подтверждающаяся характеристика Обучения-II делает это обучение почти неискоренимым.
Ay-yi-yi, how familiar that is. On the other hand, I can't see how the previous Learning-II can turn into a preparation for mental illness. It's just learning "the way to see a blot" (see below), which in a magician's value system can't be contradictory. The contradictions will probably be revealed already during the process of studying III, when, say, as a result of some dramatic event he will be forced to reorient the entire previous system of life values in such a way that he will stop being a magician and will turn into a vulgar materialist "a la Vogt" ("the brain secretes thought as the liver secretes bile"). I get the impression that Learning-III (the ability to change punctuation) does happen in people, and quite often with a complete reassessment of values.
Either I am misunderstanding something, and it would just be Learning-II. And it will become a Training-III only when he is capable of changing the system of values and corresponding actions between magic and vulgar materialism at will. Then he will definitely become a lunatic :)
A reminder about punctuation a little earlier in the article:
We assume that what is taught in Learning-II is a way of breaking down (punctuating) events. But the way of punctuation is not true or false. There is nothing in the statements of this training that can be tested for reality. It is like a picture seen in a blot; there is nothing right or wrong about it. It is only a way of seeing the blot.
P.S. Read a little further. Exactly: Learning-III is a profound reorganisation of punctuation.