From theory to practice - page 18

[Mykola Demko]

Mickey Moose:

What equations, inequalities? You don't even know the basics.

That's called trolling. You either turn the accusation around, what things are missing? in maths, trading.

Or don't write anything.

[Mickey Moose]

Nikolay Demko:

Now that's called trolling. You either unfold the accusation, what things are missing? in maths, trading.

Or don't write anything.

The message was not addressed to you.

(In case you didn't notice).

[Mykola Demko]

Mickey Moose:

The message was not addressed to you.

(In case you haven't noticed)

The answer is not on the merits, you will litter the thread and pick on the participants I will call a moderator.

You can and should pick on participants' ideas, that's what this thread is for.

ZZZY And so you threw an accusation, and you are expected that the next message you turn it around and be more specific, if you would be kind enough to please.

[Yuriy Asaulenko]

Alexander_K:
!!!!!!!!!!!!!!! My respects. Richard Feynman! Who knows this man is definitely a colleague of mine. And the problem will be solved. I am writing less now - processing the results. Working, in general. The New Year is just around the corner - I have less and less time...

Disappointingly, I'm not a physicist at all). Although, physics is not the least of my professional activities. But what amazes me: a physicist coming to Forex has forgotten all about physics and is trying to "beat" it using only mathematical methods. Before applying mathematics it would be good to understand the physical meaning of the phenomenon first. At least some kind of at least quality model of the Forex service, and then the mathematics comes along.

While talking about Feynman I found another quote especially for you:

Approximately this is what mathematical methods will give.

[СанСаныч Фоменко]

Alexander_K:

Are we talking about the same thing but in different languages?

Let's be more specific:

1. The movement of Ask or Bid price itself is a non-stationary process whose characteristics change over time, which is described by the Fokker-Planck equation.

2. Returns increments, expressed in pips, of Ask or Bid prices are a quasi-stationary process having a particular t2 Student's distribution with a particularnon-parametric mean = 0 and a scale factor s not equal to the standard deviation.

You think differently?????

I see no point in discussing anything with you - you do not speak the terminology at the level of a 1st year student of the relevant specialty.

Good luck.

[Vladimir]

Alexander:

What is the best way to organise the collection of tick data??? At the same intervals, exponentially distributed? Is EVERY tick quote that important???

Traders fight for every tick only because of arbitrage strategies, right? I do not see any other physical sense in this method of data collection

Oleg, what's the point of working with every tick? Why traders are fighting for it?

I would gladly work with all ticks and probably like everyone here, would fight for obtaining each quote. But there is a problem - the model shows the best results on timeframes H4 and higher - i.e. I need to take more than 16.384 quotes in the FIFO buffer! So life itself makes me search for ways to optimize the model.

So, it's vice versa - chasing every tick is not necessary at all and it is OK to accept ticks in a certain time interval, if we're not engaged in arbitrage. Do I understand it correctly?

Nikolay Demko 2017.12.06 15:44 #185 "Average tick rate in many brokerage companies is 3 - 3.5 sec, by setting the parameter to 1Hz you are going below the discreteness of the flow."

!!!!!!!!!!!!!!! That's something to think about.

I will try to answer the above questions.

What is the best way to collect tick data?
- As it is required by your trading system. Maybe, it doesn't need ticks at all.

Traders are fighting for every tick only because of arbitrage strategies and that's it, right? I don't see any other physical meaning in this method of data collection

- Where did you read that traders are fighting for every tick? My impression is the opposite. They don't, and MQ trading platforms ask them to do so, ticks appear to them as something auxiliary and illustrative, the main thing - OHLC for different timeframes. As Petros Shatakhtsyan has already told you, ticks have one property that distinguishes them from absolutely all other data - it is the primary source. Just like primary accounting documents. In auditing a company, they are checked. The others are not trusted until the primary ones have been checked.

- I would love to work with all the ticks...you have to take more than 16.384 quotes into the FIFO buffer! So life itself makes me look for ways to optimize the model.

It's not life that forces you, but attachment to VisSim. Why not calculate the same, but with other software? You can't seriously assume that everything will be done, as you said, "at the push of three buttons"... If the median calculation function without sorting is of any help to you, I found one (haven't checked) http://caxapa.ru/170575.html:

This is the code that calculates the median for me:

 int median( int temp[], int length) 
{ int slit = length/2;
  for( int i=0; i < length; i++)
  { int s1=0, s2=0;
    int val = temp[i];
    for( int j=0; j < length; j++)
    { if( temp[j] < val)
      { if( ++s1 > slit) goto nexti;
      }
      else if( temp[j] > val)
      { if( ++s2 > slit) goto nexti;
      }
    }
    return val;
nexti:
  }
  return 0;  // формальность, досюда никогда не доходит.
}

where: temp[] is the array to search for the median length is the number of elements in the temp[] array
The array elements of the temp[] array are sequentially enumerated here as candidates for the median. By comparing the candidate with
all other elements, we separately count the number of those smaller than it (s1), and the number of those larger than it (s2).
A candidate is immediately declared the winner (the following candidates are no longer considered) if at the end of his
comparison with the remaining ones, none of the numbers s1 and s2 exceeds one half of the total number of elements in the array. The enumeration
is so arranged that achieving such a run on any of the numbers s1 or s2 during the comparison immediately removes
cadite from the race, terminating the comparison with the rest of them (goto nexti), and the next on the list
candidate. P.S. I made up this method myself once, and programmed it in x86 asseble myself (I needed median
filtering large arrays).

End of quote

Since even the average frequency of ticks was new to you, though you have analyzed them, I will give more detailed data. Last week 27.11-02.12.2017

Explanation of the picture.
Different names in capital letters and numbers up to 4 characters long mean different DCs. At the top you can see the quality statistics for different DCs. BreAllDCms=10000, BreOneDCms=60000 means that an individual break of communication with one DC was counted if there were no ticks for 10 seconds, total break of communication - if there were no ticks from any DC
within 60 seconds. In a trading week 120 hours = 7200 minutes = 432 thousand seconds. For the 431957 seconds when ticks were running, there are 7143 seconds when there were total disconnects, almost 2 hours. Not the best week, but bearable.
Below in yellow are the cells where less than 86400 ticks (number of seconds per day) came for the week, which means the ticks were on average no more than once every 5 seconds. The record belongs to AUDNZD in 1WOR, 18890 ticks, i.e. one tick every 22-23 seconds. 8 DCs are coloured almost yellow - probably they have a 4 digit quote.
Red coloured - where more than 432000 ticks came in during the week, i.e. on average more than once per second. Again, by eye EURJPY has a maximum of 1564587 ticks in ALP, or 3.6 ticks per second. And this is on average...

[Vladimir]

Alexander_K:

Yuri, I'm probably in a hurry - so I make assumptions somewhere, use terminology a little wrong somewhere, and suffer from academicism somewhere. But, I never forget about physics. I just REALLY have no time both at work and at home to finalize the program.

Alexander_K:

In Wissim, it's a dimensionless value. And the weighted average is calculated classically there - seehttps://ru.wikipedia.org/wiki/Среднее_арифметическое_взвешенное

https://www.mql5.com/ru/forum/221552/page14#comment_6155308 In the link you cited, the weights are dimensionless, just real numbers. In Wissim they are 1/pips, so the moving average becomes dimensionless. It turns out that if DC quotes with 4 digits, then WMA in Wissim will have 10 times lower values than with 5-digit quoting. Or is it still normalized in Wissim in some way to have comparable results?

Reminded me of my question. Since you remember about physics, please answer my above question. There is no way a physicist could forget about the units of measurement. Especially when moving to practice.

[Mikhail Dovbakh]

Alexander_K:

I appeal to physicists - until you understand this, friends, we will be laughed at by some economists. That's it!

This appeal made a lot of sense to me...
megalomania, however.
)

[Vladimir]

Alexander_K:

Vladimir - look at the pictures I have attached above on the subject. This is the actual flow of quotes from the ECN account.

Here are the statistics:

How can it be that Nikolay writes about average frequency of ticks coming in 1 time in 3 sec, and I got average = 3 sec. We are absolutely strangers to him personally, but the data are the same?

In addition there is some dip around 3 sec. as if the DC is deliberately cutting out the data.

Where do you see a contradiction? According to my actual measurements of 25 instruments in 34 BC during the previous week, the average tick periodicity ranges from one tick in 22-23 seconds to 3.6 ticks per second. Is the estimate of one tick every 3 seconds outside this range in your eyes?

Maybe you can still answer my question, which you just quoted yourself https://www.mql5.com/ru/forum/221552/page22#comment_6167453?

And the DC has nowhere to cut the data, it generates it itself. Over what period were the ticks taken? I'm not asking for the length of the period, I'm asking exactly, start date-time - end date-time.

What does the name ECN account type give you, why do you keep mentioning it? Are you aware of the fact that when executing on the market all brokerage companies call their quotes indicative, not real?

[secret]

Alexander_K:

Do you have a test of your system on historical data? can you post equities here? or at least give us the main parameters - average trade, profit factor?