again on the tester - 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
No one has cut it. There is a rarely reproducible bug in the local editor. It happens when you write a post, then click on a link in the same browser window to open it in a new window or in an adjacent tab, then go back to writing the post. Once the post has been submitted, it's cropped. Our webs are aware of this, but can't reproduce it.
I write responses in wordpress. I compile everything there. Then I paste. Then I add it. Call for editing. May be missing while editing is going on. Now I was noting that the phrases are quotes. Since it's just plain text in the vdas.
Yeprtst, I looked it up, Prival, if you don't mind, drop the proof in a private message and let's see how private it is... I came across this funny thing, maybe it's American, but it's a mystery to me... ...and I'm sharing it with you.
Yeprtst, I looked it up, Prival, if you don't mind, drop the proof in your personal, and let's see how private it is... I came across this funny thing, maybe it's American, but it's a mystery to me... and I shared it with you.
There are various ways to prove it. I'll try the simplest one. A graphical one using an example.
A bar of 5 ticks. Here is its drawing.
The task is to model an adequate arrangement of points on the XOY plane. Initial data OHLC and V.
A priori loss of temporal arrangement of points.
Point 3 corresponding to H can be on any of three places 2, 3 or 4. Same for point 2(Low). First inadequacy.
Second coordinate X (time), apart from mixing up the places, their exact location is unknown. In the tester, they are at equal intervals of time. And ticks can come in different ways. The first tick at the beginning of the minute, and a pack of the remaining 4 ticks at the end. Perhaps, there can be a pack in the beginning and one in the end - the sea of variants.
For example, it is impossible to construct the indicator of the flow strength on the basis of history. What is the flux intensity I wrote on the first page here. Roughly it is the number of ticks per unit time.
This is not right, you can do it, but it will differ from the real one, that's for sure. At first glance it is a Volum, but only at first glance. At the beginning of the bar it is Volume = 0 and at the end of the bar it increases, which is not correct. It turns out that at the beginning of the bar we have no flow, because its intensity = 0, but by the end of the bar we have the maximum intensity (stationarity :-)). To correctly build the bar, we should count in the window, shift by 1 second, count, shift by another 1 second, count, etc.
This is only the simulation of 1 currency pair (plane). But the multi-currency one is not a plane - it's an n-dimensional space. How developers will adequately allocate ticks is a mystery. Which tick has come first in EURUSD or EURJPY? The opposing bar has already closed because the next tick has arrived, while for another currency it hasn't. And I don't see how it can be adequately simulated without changing the format of data storage.
Figar0 if you don't mind sharing what you've got there.If you do not want to make it public, any way. Personal. Skype.
If it's not too much trouble, tell me what you've found. If you don't want to be public, any way you can. Personal. Skype.
Yeah, I kind of dropped it in my inbox, check it out... Nothing much there, but I was amazed at this approach, and when I started digging in this direction, I got even more amazed, by the way, a little on the subject of "Do we need reliable tics?". If that doesn't work, I'll try it another way tomorrow.
Example of how the tick simulation algorithm affects the final result of the strategy in the tester:
Tester simulated ticks on this bar (11:43) :
(You can see that the tick volume of the simulated bar (39) is smaller than the tick volume of the real bar (42).
Actually the simulation could have been different, if the algorithm was different.
For example, the sequence of ticks in this case could be: Open->High->Low->Close. Then the given limit order would have closed by TakeProfit and not by StopLoss...
P.S. For webmasters: If you click on the first picture, the original picture will open, which weighs 12Kb. A smaller copy (automatically generated by forum engine) eats 183Kb (15 times larger). Perhaps something should be changed to save traffic.
For example, in this case the sequence of ticks could be as follows: Open->High->Low->Close. Then the given limit order would have closed at TakeProfit and not at StopLoss...
And what can this give us when moving from a tester to a live sequence? It seems that there is technical capability for gathering and analyzing real ticks, but no one has caught "fish" there yet... Do you want to catch patterns in a synthetic tick sequence formed from several sources with DC filters, whose algorithm is self-similar and can change every minute? Of course, working with real tick sequences will increase the reliability of testing up to 99%, but it will not give us a real quid. It's not like we're trading in a tester.... I don't understand where you friends are going with this, although I'm honestly trying... If we get MT5 with a tumbler, the question will disappear by itself.
And what can it give us when moving from a tester to a "live" sequence? It seems that there are technical capabilities to collect and analyze real ticks, but nobody has caught "fish" there yet... Do you want to catch patterns in a synthetic tick sequence formed from several sources with DC filters, whose algorithm is self-similar and can vary even every minute? Of course, working with real tick sequences will increase the reliability of testing up to 99%, but it will not give us a real quid. It's not like we're trading in a tester.... I don't understand what you friends are getting at yet, although I'm honestly trying... If I get an MT5 with a tumbler, the question will disappear by itself.
The example above is on a suggestion from here:
stringo wrote >>
Suppose the tester can be fed data from different sources - generated or derived from real tick collection - during testing. Prove that for testing, real ticks are better than generated ticks.
Regarding real ticks: there is publicly available recorded historical data on ticks with volumes: tick arrival time (milliseconds), trading symbol, best Bid price and its volume, best Offer price and its volume. In other words, this data allows the analysis even of multi-currency ticks at the same time.
If you want historical data of the whole tick, you have all possibilities to record them yourself. Such data will take up approximately 10 times as much space.
The example above is on an offer from here:
Regarding real ticks: there is publicly available data on ticks with volumes: tick arrival time (milliseconds), trading symbol, Bid price and volume, Offer price and volume. I.e., data which allow to analyse even multi-currency ticks simultaneously.
So no one is arguing that they are there. MT doesn't have them :- (. And it's bad that developers think they don't have to. Some of them, not all of them, think they need this data. That is all. MT's competitors have already started providing tick history. It would be good if MT5 had a tick history. After all, 90% of the questions would disappear. The tester uses a flow of ticks, take it and do whatever you want. And there is no need to prove its adequacy. There was a story like that and that's it. You don't like it, you can generate another one or simulate it.
So no one is arguing that they are. In MT they don't :- (. And it's bad enough that the developers think they don't have to. Some, not all, think they need this data. That is all. MT's competitors have already started providing tick history. It would be good if MT5 had a tick history. After all, 90% of the questions would disappear. The tester uses a flow of ticks, take it and do whatever you want. And there is no need to prove its adequacy. There was a story like that and that's it. If you don't like it, you can generate another one or simulate it.
Write your own tester. Make it faster than the universal MT5 tester and with any features you are interested in.
Write your own tester. Make it faster than the universal MT5 tester and with any features you are interested in.
No problem, I have been using Matcad for a long time. Everything is simple and transparent. I opened a file with quotes, created an array and do whatever I want with it.
The time of testing is not the most critical and important value. The development of any ATC consists of several stages. Idea - algorithm - implementation of the algorithm (in any programming language) - testing - analysis of the results.