Arbitrage - page 19
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
Can I ask, just out of curiosity, what methods are used to evaluate the effectiveness of the source code?
I've been preparing an article about this for some time now. That's exactly what it's about, i.e. 90-95% of the mechanical and semi-mechanical Grails that periodically appear on trader forums. A couple of weeks ago I thought it could already be published, but recently, thanks to Rosh, I saw some unexpected problems and decided to put the brakes on the article for now. In 2-3 months, hopefully I will pick it up again and finally finish it (I would like to think so). This "long-build" is entirely based on my own research and therefore the article is progressing slowly. I do not want to give away too crude a product. But I can already say with complete confidence: the article will be very pessimistic.
I need the source code to really see how the trading system generates signals. Of course, you cannot see it in a black box. And even the most optimistic results of testing the black box will not convince me of the system quality more than studying the source code.
I have never been able to understand the reason for the frenzied success of this advisor.
I am not so competent to argue about arbitrage-not arbitrage and I am not so confident
to prove to professionals that another grail has been found.
Just continuing testing the very first version to see how it will die.
At the same time I apply input filters to the code with my awry little hands and test it on demo-programs too.
For example, variant #7 has raised my deposit by $2586 since 17.04, today my profit by open positions is +210$ (current values reached +1200).
And I have high hopes on you, dear Mathemat, because I am not clear, for example, how to correctly test the finalized Expert Advisor.
The results are good on demo and I can wait for its fall and then I am afraid to use it on real trading but I am afraid to throw it away.
To make a decision, we need a testing methodology.
I have been preparing an article about this for some time now. That's exactly what it's about, i.e. the 90-95% of mechanical and semi-mechanical Grails that periodically appear on trader forums.
Well, you need the source code in order to really see how signals are generated in a trading system. Of course, it cannot be seen in a black box. And even the most optimistic results of testing the black box will not convince me of the system quality more than studying the source code.
And I have high hopes for you, dear Mathemat, as it is not clear to me how, for example, to properly test a finalised EA.
maksaa wrote: And I would also like to question the necessity of using "complicated" indicators and advisors in trading. Is it really necessary? What is the likelihood that a fatal error will not creep in behind the complexity of the formulas?
Show me an example of a "simple", really stable and profitable system - and I'll agree with you. But please note that I am highly skeptical of such "proof" of its stability as results of testing on history and even optimization. The only view in which I am willing to consider it is its open source code. Ask kompostera, who has written 300 custom EAs, or Rosha, who seems to have experimented with everything under this sun. Maybe the original tools may be "simple", but their interpretation (i.e. signals) cannot be.I think the Barispolz SC is a stable system, you must have heard about it. It also seems to me that Reshetov's system will be stable, when everyone makes it up to his own level.
When you write an article, please let me know here.
Do I understand you correctly, that you can analyse this code to conclude that the system is stable and robust? And you don't even need to test it?
I don't have any unambiguous and overarching criteria for sustainability. There are only a few hypotheses about the necessary conditions of stability ("if a system is stable, then it has so-and-so property"). But none of them is sufficient ("if the system has this property, the system is stable").
For example, I took a DB of one-minute quotes and rewrote Reshetov's code in Builder to see an objective picture of events. But even studying the minutiae of events I see that I still lose much information. And the result of the test will be only estimated and preliminary.
So, your article will indeed be interesting.
Regards, Fed.
The fact that it is unstable when working with only one given pair is evident from testing results posted by the author himself. But that doesn't necessarily mean that some combination of unstable systems won't become stable.
P.S. Eh, Yuri, what a style you have. Was it really so hard to divide a huge start() into several small logically closed blocks? That's 172 bloody strings...
P.P.S. Yuri, would you mind if I suddenly included your Expert Advisor's analysis into my article? I don't guarantee that this analysis will necessarily get into the article, but such a possibility exists. There will be no swearing in your direction, don't worry. But if the analysis does pass judgement on the EA, don't be sorry... If you don't want to reply on the forum, write to my postal address, it's in my profile.
In fact, I'm genuinely disappointed that the release of the article will probably still be delayed. I suggest you first release part 1 - without multivariable testing, and then the subsequent ones. Frankly speaking, there is a great shortage of competent packs on this topic. Personally, I am now moving on my own guilt when developing a tester. That is, I have experience in programming and working with the database, including on satellite positioning data. The quotes are even worse. The material is as follows: I have taken the major currency pairs, minute quotes, I obtain crosses by calculation (the initial quotes of crosses have holes for 15 minutes), the Close price is a minute one. I imitate Ask as Close+spred, and Bid is vice versa. Well, here is already an obstacle and error.
The mql commands, such as OrderCloseBy, have to be rewritten using own functions. Fortunately, Yuri has no indicators and the code is simple. If I dealt with tests professionally, the mql commands should have been transferred to dll.
Yuri's code is simple, but far from being primitive. On the one hand it's clear, but on the other hand it's hard to figure out how it all works in a group. I put all variables calculated (every minute) into log table and look through these arrays with my eyes and see what's going on inside. But I haven't finished transferring his code to the end yet. When I've mastered it thoroughly, of course, I will reduce log to readable state. But for now I'm afraid of making a mistake somewhere. It's important for me to select the optimal group of currency pairs, then I'll improve it for myself directly in Builder, and only then I will transfer my developments to mql. But my actual decisions (on improving the code and assessing the composition of groups) will be made on the basis of historical data, which does not guarantee....., etc. So what to do? Another drawback - it takes me a long time to calculate multicurrencies (I tried another system before). 60 minutes - 1 minute is calculated (+ also swap data for calculations from time to time, so that further all moves faster and it is still about a minute). SQL I know well and everything seems to be done optimally in calculations. But I have no patience for constant testing, I have 10-20 day portions.
But if there was a really good practice in testing - I would use it. Maybe I would cut down on some things, or pay more attention to something, or even do things differently.
So: Waiting for the article!
Reshetov's code is indeed interesting. Maybe I have not seen much of it in my life, but the approach is really innovative and unusual. Even if my test shows that it is not worth risking on real, I am still very grateful to Yuri - he generates a lot of other thoughts.
Regards, Fed
Yuri, would you mind if I suddenly included an analysis of your Expert Advisor in my article?