Interesting topic for many: what's new in MetaTrader 4 and MQL4 - big changes on the way - page 23
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
It's just algotrader ignorance to store the bar spread. They would only replace it with Low_Ask. And the accuracy would have improved considerably.
I doubt it. Low_Ask could be at any point in this stretch, in that respect the dickens are the same.
Next question to everyone: how is the historical spread used by the tester calculated?
I doubt it. Low_Ask can be at any point in this segment, in this respect, the dick is the same as the dick.
It's not a question of super accurate tick history. For almost all TC, the run with M1_Bid+Ask and the tick history will give a miserable deviation. It is very silly to spend huge amounts of time and computing resources for the sake of it.
As a practitioner, for almost all TCs M1_Bid+Ask history is more than enough. And it does not matter which OHLC Bid was earlier or later than OHLC Ask.
Adding Ask history to the tester is two lines for developers. It does not affect speed of testing in any way. Only the accuracy improves and a lot of market regularities appear. Profitability of TSs is adjusted more precisely and becomes even more profitable than it would be if they were adjusted on a simple bid history.
Next question: how is the historical spread used by the tester calculated?
There can be no hard feelings. Let's respect each other. Someone is actually doing something useful for others for purely ideological reasons. Stop being silent.
It's no fun to feel like an idiot that no one understands. It's a total bummer to be listened to just because of some stupid percentages shown somewhere.
Well, there has to be logic. There are people who speak logically, don't put them down without understanding.
If someone doesn't understand the importance of the asc story - just say you don't. Don't start arguing that it's not fucking necessary.
If someone doesn't understand that meta-testers suck at accuracy - again, say you don't understand it.
Perhaps someone will come along and explain the reasoning, and you will finally understand.
But the practitioners who are silent are lazy assholes, to put it mildly. Only a small number of them have their own research infrastructure. All the rest stupidly test everything on crude metatheaters, missing a huge amount of simple market patterns because of it.
Yes we understand everything, you do not understand that the tester is not made to test the HFT. And no asc history is not going to fix it.
For those algorithms you're talking about, you need to immerse MT completely in a virtual environment along with a virtual DC server, and test everything on the saved history of the tumblr.
I don't know when I told that this tester is not for programmers, not for checking the correctness of the algorithm, and certainly not for HFT, but for a rough estimation of TC.
As for me, instead of a tester we could create something like UML for checking strategies and that's all, it would be enough. You can calculate on the knee in the indicator where you open at what price and where you close the sum, and all this on a 10-year history for 1.5 seconds.
Anyway, no matter how much you scratch out this tester, you won't believe that all the functionality of "Events, graphical objects and any other new function" won't fail.
Hence the moral, we have a working tester, which costs a ton of time (read dough), which satisfies 99% of users, and only 1% is not satisfied. Yes, the most advanced percentage, but still 1%. And this percentage is trying to convince that by burdening the tester with processing additional data, and as a consequence, reducing the speed by half, there will be a gain in accuracy.
There will not be this gain, why I wrote here.
What the hell is HFT? I made almost all my money with my tester, which I assembled on my knees in a day or two, and which uses the " close prices" principle with M1 bid and ask history.
And it is due to this tester that my results are much stronger than others, even with more advanced ideas. The reason for this is simple. You take an idea and the metatesters show that it doesn't fucking work.
And my tester shows that it works. That's it. Some people make money, others keep looking.
Sometimes the opposite happens: metatesters show that the idea works, while my tester doesn't. In the end, the deceived user, in the worst case, drains his money.
...
For me it was possible to create something like UML for testing strategies instead of tester ...
...
I'm serious, after years I realized that to check the TS you need a very simplified calculation where you opened where you closed on what signal, everything is very light and as quick and simple as possible.
No need to emulate the execution, everything is tested on demos and micro-reals.
But the functionality of built-in statistics could be really tricky, starting with built-in neural networks and ending with making statistical hypotheses. Such a tester would be a real shot in the armoury.
You know very well my opinion on the asc story, and it's not negative, which by the way I pointed out in that thread, only above. About the tester I am also silent, show me where I would express my complete satisfaction with it. I'm just a proponent of real solutions, not sky-high dreams. Because the "if only" can wait a long time, almost indefinitely.
Yes there is no question of super accurate tick history. For almost all TCs, a run on M1_Bid+Ask and on the tick history will give a negligible deviation. It is very stupid to spend huge amounts of time and computing resources for the sake of it.
As a practitioner, for almost all CU's M1_Bid+Ask history is more than enough. And it doesn't matter which OHLC Bid was there before or after OHLC Ask.
Adding Ask history to the tester is two lines for developers. It does not affect speed of testing in any way. Only the accuracy improves and a lot of market regularities appear. Profitability of TSs is adjusted more precisely and becomes even more profitable than if they were adjusted on a simple bid history.
In the tester, the notion of spread should not exist at all.Yes, I agree in this context.
It's trivial and nonsense why Ask prices are so disadvantaged, they are no less significant than Bid.
I'm serious, after years I realized that to check the TS you need a very simplified calculation of where you opened where you closed on what signal, everything is very light and as quick and simple as possible.
You don't need to emulate the execution, it's all tested on demos and micro-reals.
But the functionality of built-in statistics could really be tricked out, starting with built-in neural networks and ending with statistical hypotheticals. Such a tester would be a real shot in the armoury.
Don't forget, there are "subtle" TS, which are built not only on the intersection of wipes and other nonsense.
It's not a question of super accurate tick history. For almost all bars of M1_Bid+Ask and the tick history will give a miserable deviation. It is very silly to spend huge amounts of time and computing resources for the sake of it.
As a practitioner, for almost all CU's M1_Bid+Ask history is more than enough. And it doesn't matter which OHLC Bid was there before or after OHLC Ask.
Adding Ask history to the tester is two lines for developers. It does not affect speed of testing in any way. Only the accuracy improves and a lot of market regularities appear. Profitability of TSs is adjusted more precisely and becomes even more profitable than if they were adjusted on a simple bid history.
The notion of spread should not exist in the tester at all."This is what it was like here before the Soviets"(c).
All you say there is in MT5, there is Ask accounting, it (Ask) can be received at any moment, the corresponding levels are triggered by Ask.
Yes the price is calculated from Bid+Spred and what's the difference if you don't care what was LowBid or HighAsk before ?