![MQL5 - Language of trade strategies built-in the MetaTrader 5 client terminal](https://c.mql5.com/i/registerlandings/logo-2.png)
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
I'm thinking about this - if the issues of defining the concepts of probability density W(x,t) and drift M(x,t) in the Fokker-Planck equation cause such a rejection, then what happens when we consider the following terms - current and historical variance? I don't know, I don't know... I'll think about it until tonight - maybe I'll drop the subject altogether...
In the meantime, I continue to write for those people to whom these concepts are close and familiar or who wish to become familiar with them. Well, for young people, of course - just read, maybe something will come in handy :)))
For the moment, by common efforts we have made two practical conclusions:
1. the algorithm of tick data reception - with a period of 1 second real tick quotes are accepted, i.e. when the deal was really executed. Not all ticks in a row, and not every value with 1 sec. frequency, but exactly.
2. The only true measure of the central tendency for a market price movement is not the MA (moving average), not the median or mode - but precisely the WMA - moving weighted average, where the weight w is determined from the probability density formula for the price increments returns for a particular currency pair.
I've already tried to point out somewhere in the discussion that the market is a bit different from what mathematicians imagine it to be :-)
In order to take data to build or test a hypothesis, you first have to sift out what you don't need. Here you have taken the incremental quotes you get and they technically come in two (maybe more, but two are obvious) states
- Active trading - the price that comes to you/us is largely shaped in demand/supply on a particular server and its immediate surroundings. The main actors here are the traders.
- Passive trade - the price which comes to you/us is generated mostly by the suppliers and piers and the local market does not participate at all. Market makers are the main actors in this market. There are not enough traders to trade with each other and the market makers have to play along.
There are also technical nuances - in moments of news release, session opening/closing, date changes ticks sent to the terminal will "disappear and stick together", i.e. increments will be different. Man, the spread jumps precisely at the border of the hours, purely technical nuance, independent of the market.
Plus all the pairs are related and only EURUSD may be studied/analysed - it is the most active and it is less affected by movements of others. But every movement of EURUSD has an effect on other pairs. If you take something else, you will accumulate a great amount of noise, that will only hinder in the analysis. Or you measure a specific rate of a specific broker at a given time :-)
There is also a point about "tick volume", which is neither the volume nor the number of passed deals, nor the amount of fluctuations of bid and ask outside tick_size, but for some reason it is different for everyone :-)
I've already tried to point out somewhere in the discussion that the market is a bit different from what mathematicians imagine it to be :-)
In order to take data to build or test a hypothesis, you first have to sift out what you don't need. Here you have taken the incremental quotes you get and they technically come in two (maybe more, but two are obvious) states
- Active trading - the price that comes to you/us is largely formed in the demand/supply on a particular server and its immediate surroundings. The traders are the main actors here.
- Passive trade - the price which comes to you/us is generated mostly by the suppliers and piers and the local market does not participate at all. Market makers are the main actors in this market. There are not enough traders to trade with each other and the market makers have to play along.
There is also a technical nuance - in moments of news release, session opening/closing, date changes ticks sent to the terminal will "disappear and stick together", i.e. increments will be different. Man, the spread jumps precisely at the border of the hours, purely technical nuance, independent of the market.
Plus all the pairs are related and only EURUSD may be studied/analysed - it is the most active and it is less affected by movements of others. But every movement of EURUSD has an effect on other pairs. If you take something else, you will accumulate a great amount of noise, that will only hinder in the analysis. Or you measure a specific market of a specific broker at a given time :-).
There is also a point about "tick volume", which is neither the volume nor the number of past transactions, nor the amount of bid and ask fluctuations beyond tick_size, but for some reason it is different for everyone :-)
You have actually outlined the key problem, which does not allow physicists-mathematicians to turn around. In fact - traders do not have a unified ALGORITHM of taking tick data for processing and analysis. I've read so many articles - there's no unanimity about the method of tick data sampling, and that's it.
Most tend to think that we should fight for EVERY tick. I think this approach is fundamentally wrong.
In this thread I have suggested another method for taking data - I think it is more correct from a physical point of view.
1. Last is a reflection of the actual transaction, the rest is a tambourine game. And no discreteness at 1 Hz will help, because we are not trading time, but price. This is not important, as long as the change has not occurred, the world of trading does not exist.
On 2. What is the sample size for calculating probability density function? for WMA
...Most are inclined to believe that you have to fight for EVERY tic. I think this approach is fundamentally wrong.
In this thread I suggested a different method of data reception - it seems to me that it is more correct from the physical point of view.
Alexander, hello!
Do I understand correctly, that the concept has changed a bit? :-)
Now we need to collect ticks with a certain periodicity, let it be 1 sec.
In MetaTrader4/5 this can be done in the Tester using a robot, a mule chaser [tick collector].
1. Last is a reflection of the actual transaction, the rest is a tambourine game. And no discreteness at 1 Hz will help, because we are not trading time, but price. This is not important, as long as the change has not occurred, the world of trading does not exist.
On 2. What is the sample size for calculating probability density function?
just sit in a flat for a few years for a tenfold jump and then tell me that time is not important.
Alexander, hello!
Am I right in assuming that the concept has changed slightly? :-)
Now we need to collect ticks with some periodicity, let it be 1 sec.
In MetaTrader4/5 it can be done in the Strategy Tester by a robot, a mule chaser [ticks collector].
Greetings Denis!
You see what the problem is.
If we want to solve this problem together, we need to speak the same language - work with the same data stream. Moreover, it's not the fact that the trading robot that will be tested and built will really work with other brokerage companies - the quotes flow is different.
NikolayDemko asks - what should be the sampling volume? What should I answer? Because I operate only with my brokerage company and according to my estimations FOR ALL currency pairs the volume of selection should be no less than 5 000 ticks . And for each pair an optimal volume of sampling is calculated manually by the formula I gave in this thread.
But this optimal volume can appear different for different brokerage companies. So, what should we do?
The answer - we have to work out a UNIVERSAL algorithm for data reception.
1. Last is a reflection of the actual transaction, the rest is a tambourine game. And no discreteness at 1 Hz will help, because we are not trading time, but price. This is not important, as long as the change has not occurred, the world of trading does not exist.
On 2. What is the sample size for probability density calculation?
Greetings Nikolai!
Last is a really cool thing. It simplifies the algorithm i.e. I don't need to work with Asc and Bid separately. But, I foolishly opened an account with some broker that has no Last.
So the unfortunate thing is that whatever I write and calculations I give - these calculations are valid only for me in my brokerage company. Each trader who wants to use my calculations will have to double-check them for his brokerage company, conduct his own calculations of volumes, etc. This is very long and tedious.
I stand by my opinion - it is necessary to read quotes in a certain sequence.
Greetings Nikolai!
Last is a really cool thing. Moreover - it simplifies algorithm, because it becomes unnecessary to work with Ask and Bid separately. But, I foolishly opened an account with some broker that has no Last.
So the unfortunate thing is that whatever I write and calculations I give - these calculations are valid only for me in my brokerage company. Each trader who wants to use my calculations will have to double-check them for his brokerage company, conduct his own calculations of volumes, etc. This is very long and tedious.
I stand by my opinion - it is necessary to read quotes in a certain sequence.
I'll tell you a more dramatic idea, treat it as an untested hypothesis: it is necessary to analyze Last from exchange-traded currency futures, because market makers (this is my hypothesis) look at the futures as a guide. Thus analyzing one flow it will be possible to apply the results of the analysis on any DC, because forex pricing (imho) is an echo of pricing on exchanges.
I will even say more, the futures themselves are traded with an eye on options. I don't know what the mechanism is, maybe the futures traders are also options traders, but there are real battles to hold or break one level or the other before the options levels.
Although the above mentioned does not cancel the very concept of problem solving. I assure you - in theoretical physics such problems, similar to price movements at Forex are solved quite often, it is just that real physicists have no time to deal with them - they consider it unworthy to deal with such nonsense. But I was carried away (does it mean that I'm not a real physicist? - hmm... I'm into some kind of philosophy, though). Well, it's just for general mood :)))