Interesting topic for many: what's new in MetaTrader 4 and MQL4 - big changes on the way - page 35

[Alexandr Bryzgalov]

tol64:
Well, you'll probably not only notice it, but you'll also manage to pinch it off. )))

Of course, while you are writing code to follow this rule, while you are testing it on a 100 year old story, the rule disappears, so it's real

[Vasiliy Smirnov]

hrenfx:

What tick spread? Do you even realise that there should be no spread at all in the tester? And the ticks are 99% unnecessary.

It is only necessary to have HighBid and LowAsk. No ticks or spreads. Stop talking nonsense.

HighBid and LowAsk are made up of this, what is the nonsense? The ticks in this case are more complete information.

[Mykola Demko]

sanyooooook:
of course, and while you write code for this legitimacy, while you test it on 100 years of history, the pattern disappears, so only real

Oppa, so that means that MQ wrote the tester for nothing?

[Alexandr Bryzgalov]

Urain:
Oppa, so MQ has written the tester for nothing?

I only use the tester to check if the code is correct, searching for patterns through the tester is nonsense to me.

I think that's why they wrote it, to quickly test the code for possible errors, the conditions in the tester are very perfect.

We should check pips strategies out of it (the one in MT4 now, MT5 has not tried much, but judging by the theme the same, someone is missing something) makes no sense, in real life it's different anyway, only if the code is correct for the algorithm to check.

[Vladimir Gomonov]

hrenfx:

What tick spread? Do you even realise that there should be no spread at all in the tester? And the ticks are 99% unnecessary.

It is necessary to have only HighBid and LowAsk. No ticks and spreads. Stop talking nonsense.

There is one subtle problem with extremum testing.

The point is that we never know in real time that a certain "now-price" is an extremum, if it is - it will be visible on the history, but nobody knows it "right here-now" (at least reliably). On the other hand, the tester knows it in advance. Thus, trading on extrema in the tester (for example, on OHLC) is a kind of tip from the tester, and it is a very important (I would say critical) information.

For this reason I, for example, do not particularly trust the attempts to tune the TS "directly" using OHLS testing. At least I do not practice it due to the above reasons. Now it seems to me that such optimization is possible but special efforts are required not to become a victim of self-deception.

[Mykola Demko]

sanyooooook:

I only use the tester to check if the code is correct, searching for patterns through the tester is nonsense to me.

I think that's what they wrote for, to quickly test the code for possible errors, the conditions in the tester are very perfect.

If you want to check pips strategies out of it (the one in MT4 now, MT5 has not tried much, but judging by the theme the same, someone is missing something) makes no sense, it's different in real life, only if you check the code for the correctness of the algorithm.

The only difference is if the code is checked for correctness of the algorithm.

And you are hammering nails in with a microscope :)

[Mykola Demko]

MetaDriver:

There is one subtle problem with extremum testing.

The point is that we never know in real time that a certain "now-price" is an extremum, if it is - it will be seen on history, but "right here-now" it is not known to anyone (at least reliably). On the other hand, the tester knows it beforehand. Thus, trading on extrema in the tester (for example, on OHLC) is a kind of tip from the tester, and it is a very important (I would say critical) information.

For this reason I, for example, do not particularly trust the attempts to tune the TS "directly" by testing on OHLC. At least, I do not practice it for the above reasons. Now it seems to me that such optimization is possible but special efforts will be required not to become a victim of self-deception.

So you need a ticking history? (don't answer this is a provocation :)

[Alexandr Bryzgalov]

Urain:

Well, people have tried to saw, there were all sorts of deadlines, hurry up, hey whoopee.

And you use a microscope to hammer nails :)

That's the way it's done.)

ZS: lied a bit, some strategies run in the tester for a general idea of the system, but only if the system is interesting

I do not look at the status quoted in the tester (I'm lying a bit), I look at the chart from afar, if the blue icons are below the red ones the system will work if not, then it will not.

[Vadim Konyaev]

MetaDriver:

...

Here are more histograms:

EURUSD M1:

...

is this a bug or is robofx messing with it?

[Vladimir Gomonov]

Urain:
So you need a ticking story anyway? (don't answer that provocation :)

You only think it's a provocation, in fact there is no professional alternative, bars are a surrogate.

Open/Close prices have nothing to do with the logic of the quote, they are randomly pulled out of the row. // And one of them (any of them) is unnecessary.

As for intra-bar extremums - I have already stated it.

They are not perfect (there are, for example, an awful lot of them), but there is nothing to replace them without falling into tester idiocy.

--

You can suggest many ways to thin out tics. Ivan suggested it and I've tried it - it seemed reasonable, but I can make up some more. If anybody likes it, let them test/optimize it using Renko/Kagi; if I want equitemporal bars, go ahead and test them; if I want equi-capacity ones, let them generate and test them; if I got a hankering, let the optimizer use raw ticks.

HOW to realize it is another quite difficult question (taking into account the necessity of multicurrency synchronization!), but it can be solved if desired.