Machine learning in trading: theory, models, practice and algo-trading - page 534
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 is strange to hear that there are no real ticks in the metatrader's tether, one can feel an optimum of working with the platform
Really? Sorry, maybe I missed something, is it possible to download historical ticks? Or are they only available for "proprietary purposes"(rnd)?)) I'll be glad to be wrong indeed, I like to be wrong when I can get some ticks for an indefinite period in a lot of FX instruments))) I'm not a good sucker myself and I'm paying a lot of money for ticks....((!)
To prove my assumption, I tested my TS for generalization of the market and came to the conclusion that in general the TS is overtrained, and in order for it to start earning it is necessary to set such parameters that lead to this picture in the training area.
After all, as we know, the training area is not so important, so there is nothing wrong with the TS losing. But in real work it will gain, and the first two days are an example.... let's see what will happen next....
And here is the actual test section for market generality. This is where we choose how to orient the network so that it starts to gain in the future. The principle is simple. In the process of testing, the TS should not fail, as we see in this picture. About how to conduct the final test of the obtained model, for the generalization of the market, we will talk a little later....
Really? Sorry, maybe I missed something, is it possible to download historical ticks? Or they are available only for "official purposes" (rnd)?)) I'll be glad to be mistaken indeed, I like to be mistaken when I can get ticks for an indefinite period in a lot of FX instruments))) I'm not a sucker myself and I'm paying a lot of money for ticks from time to time....(((
You can use ticks from brokers` servers in mt5 for a long time already. ctrl+u, export ticks
Well in mt5 from the servers of brokers/exchanges have long been available, of course you can. ctrl+u, export ticks
You can download ticks for the last day from dukascopy.com/swiss/russian/marketwatch/historical/, but you need at least for a year and for the whole market (~150Gb), if you look at ticks from the kitchen server for previous days they have a completely different number, in order of number as minutes. It's just not even the evil intention or greed, but saving traffic, it's impossible to use "real ticks" for testing, a year of ticks for one instrument is about one gigabyte, look how much weight a folder with MT5 has during testing, where are those ticks? It's just marketing.
You can download ticks for the last day from dukascopy.com/swiss/russian/marketwatch/historical/, and you need at least for the year and the whole market, if you look at ticks from the kitchen server for previous days there is a completely different number, in order of number as minutes. It's just not even because of malice and greed, but to save traffic, it's impossible to test it with "real ticks", a year of ticks for one instrument is about one gigabyte, look how much weight a folder with mt5 has, during testing, where are those ticks? It's just marketing as usual, though.
during 2 years 600mb of ticks in the terminal folder of eurusd, 25mb every month is in .tks
and if you export it to csv, it took me 2 months to download 800mb
Now for the whole 2017, let's see how much it will be
97555367 ticks... 4.85 gb... not enough?
it's clear that not all brokerage companies have such a depth of tick history, some have more and some have less
For 2 years there are 600 mb of ticks in the terminal folder of eurusd, each month 25 mb in .tks
And if you export to csv, then for 2 months 800mb have been downloaded
Now for the whole 2017, let's see how much it will be
97555367 ticks... 4.85 gb... not enough?
mmm... well, it seems to be true)))) maybe I was wrong. But all the same, until I get 1 in 1 identical results with my tester, I wouldn't trust whatever the hell is written for the benefit of the kitchens. The tester itself is a fairly simple algorithm, the algorithm must be transparent and its results should be reproducible, as for example for a function of some kind of mathematical.
mmm... well, it seems to be true)))) maybe I was wrong. But all the same, until I get an identical result 1 in 1 on my tester, I would not trust who knows what written for the benefit of kitchens. The tester itself is a fairly simple algorithm, the algorithm must be transparent, its results must be reproducible, as for example for a function of some kind of mathematical.
If you take into account slippages and others in forex, then I think the quality of tick history is quite acceptable for testing
Well, the results are at least clearly different if you test on simulated and real ones, of course I cannot say anything about the quality of tick history itself... if you take into account slippages and other things in forex, then I think the quality of tick history is quite acceptable to test
Well, if you start to criticize MT tester, it's first of all for its speed, I don't know what they've done wrong, only conspearological ideas about kitchen influence come to mind, but once again, the tester is an ELEMENTARY ALGORITHM in a loop: on the signal - add / subtract to the balance lot * price (the next tick), everything, maybe a couple more insignificant operations, as many signals as iterations, it should take 1-10% more time than calculating the signals themselves, hundreds of millions (> tens of minutes) bars should be processed in tenths of seconds. To explain how a salaboin "strategy" on the wagons one run on minute candles for a month period, tested for many minutes, is not possible, it's a diversion. And it's not much more significant that "on real ticks", if signals are generated from candlesticks, and all ticks are in memory, then the execution is simply a selection from the maps on the time key following the signal of the tick or a lag later. In short... You got it)))
Well, if you start criticizing the MT tester, then first of all for the speed, I do not know what they have done there, I can only think of consperological thoughts about the kitchen influence, but once again, the tester is ELEMENTARY ALGORITHM, in the loop: on the signal - add / subtract to the balance lot * price (the next tick), everything, maybe a couple more insignificant operations, as many signals as iterations, it should take 1-10% more time than calculating the signals themselves, hundreds of millions (> tens of minutes) bars should be processed in tenths of seconds. To explain how a salaboin "strategy" on the wagons one run on minute candles for a month period, tested for many minutes, is not possible, it's a diversion. And it's not much more significant that "on real ticks", if signals are generated from candlesticks, and all ticks are in memory, then the execution is simply a selection from the maps on the time key following the signal of the tick or a lag later. In short... you got it)))
the standard MACD sample Expert Advisor with simple logic, 1 minute at 2 open prices for the year... well, the feeling is 4-5 seconds... and that's for the year on minutes
i don't think it's that slow + he reproduced the trading environment of floating spreads, drew a chart and showed the report