To follow up - page 37

New comment
 
Mathemat >>:

Давай посмотрим, что получится. Этот подход тоже имеет право на жизнь.

That's what I want people to try and I would like to see what happens :) . On a more serious note, there is no point in maximising profits from unrealistic deals.

But I am very alarmed that avatara wants to go through history in the opposite order to the natural order, i.e. he intends to go from bar zero downwards. On the contrary, I would go from King Gorokh to the present, carefully avoiding looking into the future.

Well it's technically easier for him to look for ZZ tops, I think. That is, he is not serious about this method yet, he does not believe that he will be able to cut off unnecessary things from the phonetic TS :)

Curious and sobering. Maybe try to find the Intersection of the optimum zones on the whole story? Watching them move now I think is inadequate, as it is simply a consequence of a particular implementation of the quoting process, nothing more. With a different implementation these zones could move quite differently.

I don't know, I trusted in good parameters right away, and I still do. But the crisis has stirred up the swamp, so I am now calculating the diffusion radius for different times:

2006 100 minutes 18.7 old pips, 1000 pips 55.7, 10000 pips 160.9

2007 100-17.11 1000-51.6 10000-164.0

2008 100-37.6 1000-111.8 10000-391.07

2009 100-31.6 1000-93.2 10000-257.47

In 2009, it seems to have started to calm down, with more rapid diffusion on the larger horizons.

It would of course be more correct to calculate the time periods by half a year, but I do not want to yet.

 
Candid >>:

Так я же и хочу чтобы народ пробовал, а я бы смотрел что получится :) . Если серьёзнее - "но" по выходам остаётся, нет смысла максимизировать прибыль от нереальных сделок. Ну ему так технически проще вершины ЗЗ искать, я думаю. То есть он ещё не всерьёз к этой методе, не верит что от отфонарной ТС удастся отсечь лишнее :)

Exactly. And it was you who launched the idea.

Non-significant FC should also have been investigated to see if the vicinity of 0 would interfere.

And the introduction of F-C allowed you to introduce a criterion. Let it also be perfect.

 

Going back to basics.

Share your opinion on the volatility 'measurer'.

Kind of 'stovepiped' in the approaches.

;)

 
Mathemat >>:

Вот я только не понял, где начало истории, а где конец. Давайте лучше говорить в терминах нумерации баров, принятой в MQL4. Итак, с нулевого бара в бар с максимальным номером? почему именно так?

The marking needs to be done from ts.Goroch to us. First there were causes, then there were consequences.

Mathemat >>:

I like this approach too (Andrew, I probably misunderstood you at the beginning). It turns out that we are completely abstracting from the specific system giving out the pre-signals (it just doesn't exist), and drawinga"potential utility map

" of the price series. OK, let's give it a try. Do I understand you correctly now?

Exactly. We investigate the terrain, and only then build the car according to the information obtained

Mathemat >>:

Only I'd like to say right away: the utility map should be real, not an ideal markup a la ZZ.

Of course real, taking into account real spreads etc. I wrote: "(100% sure, it won't look like a zz by far)"

 

What I meant was that both inputs and outputs should be real, not at extremes. Only using the information on the left.

 
Mathemat >>:

Я имел в виду, что и входы, и выходы должны быть реальными, а не на экстремумах. Только с использованием инфы слева.

Well, that's what I meant. :)

Only using the information on the left.

 
Mathemat >>:

Я имел в виду, что и выходы должны быть реальными, а не на экстремумах. Только с использованием инфы слева.

I agree. But you could look at the inputs first.

If there is no principle in theirs either, there is no way out.

An ideal QC evaluation system is only the first step.

---

Especially if the TS is rollover, or net. Like in MT5.

;)

 

I'm getting a bit confused. If using only the information on the left - then how are such inputs better than the inputs of the real system? What makes them perfect?

 
Mathemat >>:

Что-то я зарапортовался. При использовании только инфы слева - чем тогда такие входы будут лучше входов реальной системы? В чем их идеальность?

I was hoping we were making them based on earlier analysis using perfect outputs. ;)

Now let's assume that any input with the opposite sign is an output for the inputs previously made.

But without a measure of ideal quality of QC and parameters forming the input - how to manage?

--

Here I propose to first determine the estimated (however bad) QC and only then move on.

Looking ahead at the history will provide an opportunity to filter and scale back the coordinates of FP.

Otherwise, I can't imagine how.

Juneteenth so far.


I'd like to see something similar...

 
avatara >>:

Но без мерятеля идеального качества КК и параметров образующих вход - как обойтись?

QC parameters do not form input, they only confirm or reject it.

Let me give you an example. Let's say I wanted to get away from the city for a couple of weeks and sunbathe in Turkey. Today is the first day of a two-week holiday. This is a preliminary signal of my system.

But we should also look at QC: is it the season in Turkey or not? If so, the AAC is working and I get my system's signal, i.e. go to Turkey. If not (it is winter in Turkey, snowing and generally cold), then the signal is rejected. But I won't make a decision based on QC alone if I have no signal: even if it's +35 in Turkey, I won't be able to go there to sunbathe until my holiday starts.

I.e. the general scheme is as follows: accurate preliminary signals are formed by the primary system, while QC as a coarse filter allows you to make the final decision. But not vice versa. The "+35" coordinate is not perfectly accurate, there could be a "+38" and a "+32" - and these different QC coordinate values will still allow me to execute the same signal. There is no one-to-one correspondence between the signal and its confirming QC coordinates.

Juneteenth for now.

"Execution cannot be pardoned". At first I thought there was a comma in that remark.

And the scheme, please, make it clear.

P.S. Going back to the example and Candid's scheme: one of the problems with the scheme is that the QC coordinate is assigned to the whole trade, which, generally speaking, has some extension in time (and hence a fuzziness in QC coordinates even for one given trade).

1...303132333435363738394041424344...51
New comment