THE 2007 TRADING CHAMPIONSHIP! - page 5

New comment
 
xeon писал (а):
 I dare say it has certainly paid off, for example, with a surge in popularity of MT4.
For example I was surprised to hear positive feedback on the Championship from people who totally rejected automated trading. Plus some of my friends have changed their brokerage companies for those who work with MT4.
At least it is an indicator for me :-)


Exactly my point, I also know a few people who have changed their negative attitude to autotrading after the championship.

As for restrictions, I'm sure the margin call proposal will solve all problems, as I'm sure the next championship will be more competitive, and there won't be any random lucky ones in the front row, they will just get screwed by the margin call!

 
Yes, I think the idea of a 50% margin call would be a good compromise.

It would immediately reduce the number of blatantly reckless experts and lead to more people controlling losses.
 

And I don't think a 50% margin call is going to stop the developers of reckless EAs.
Who's going to stop them from opening with 5-10 lots from the start?
Here is a typical example:
https://www.mql5.com/ru/users/anatoly/
This Expert Advisor finished the Championship in 16 hours. Had there been a 50% margin call, he would have finished it even earlier.
Similar EA is in sashkena, but it opens in another direction :) (And it's still impossible to beat "double" registrations 100%)

Decrease leverage to 1:20 or even 1:10 would solve the problem of reckless Expert Advisors.

But I understand, for sponsors and organizers, it's desirable that the winner shows the best result, but a small leverage doesn't help.

 
I do not understand why everybody is clinging to artificial limitations (number of orders, lot size, number of accounts). IMHO if the championship is to be more close to the real conditions, then first of all the winner must be determined not by profit, but by several statistical parameters, including expected payoff, probability of ruin, volatility, etc. All these parameters can be combined into complex estimation like geometric mean or weighted average. That is, if the task is to eliminate "non-serious" Expert Advisors, no matter how artificial restrictions may be applied to the code, there is no guarantee that the Expert Advisor will be relevant while the criterion of success is only profit.
In general the ideal situation is when expert passes the following test:

a) good results on historical tests, criterion - "total" score on several indicators
b) good results in real, i.e. statistics of indicators in real situation does not critically differ from historical tests. Or for the better :)
 
Renat писал (а):
Yes, I think the idea of a 50% margin call would be a good compromise.
It would immediately reduce the number of blatantly reckless experts and lead to more people controlling losses.
I think if you add to that the requirement that the expert survives the test in the tester for the previous say 9 months, it will definitely weed out the reckless experts and eliminate the need to come up with artificial limits.

Survival testing can be made more objective, if one does not test all 9 months at a time, but three tests per quarter.
Sashken's expert showed excellent results in the test from July 1, but in the test from April 1 he lost everything.
 
SkullZZ:
I do not understand why everybody is clinging to artificial limitations (number of orders, lot size, number of accounts). IMHO if the championship is to be more close to the real conditions, then first of all the winner must be determined not by profit, but by several statistical parameters, including expected payoff, probability of ruin, volatility, etc. All these parameters can be combined into complex estimation like geometric mean or weighted average. That is, if the task is to eliminate "non-serious" Expert Advisors, no matter how artificial restrictions may be applied to the code, there is no guarantee that the Expert Advisor will be relevant while the criterion of success is only profit.
in general, the ideal situation is when expert passes such a test:

a) good results on historical tests, the criterion is a "cumulative" score on several indicators
b) good results on the real account, i.e. statistics of indicators on the real account does not critically differ from the historical tests. Or for the better :)
As if it's not possible to adjust to these conditions?

Unfortunately, in reality there will only be a few people who will work through the optimization for this integral index, and then come out as winners with discouraging results. For example, having made only 1 (and amazing) trade. But more likely some newbie will accidentally show amazing efficiency calculated by the formula in a couple of trades. And everyone will just be shocked.

Therefore, the solution lies in the area of maximum simplicity, but not in the area of complicating conditions and inventing complex formulas.
 
Yurixx писал (а):

 I think that if we add to this the requirement that the expert survived the test in the tester for the previous say 9 months, it will definitely eliminate reckless experts and avoid the need to come up with artificial restrictions.

Survival testing can be made more objective, if one does not test all 9 months at a time, but three tests per quarter.
Expert sashken'a showed excellent results in the test from July 1, but in the test from April 1 all failed.

First, one can easily make millions on a known history.
Secondly, not all of the Expert Advisors can be tested on this history. First, one has to convince Metaquotes to create a multicurrency tester :)
 
Better писал (а):
Yurixx wrote (a):

I think that if you add to this the requirement that the expert survives the test in the previous say 9 months, it will definitely weed out the reckless experts and avoid the need to invent artificial restrictions.

Survival testing can be made more objective, if one does not test all 9 months at a time, but three tests per quarter.
Expert sashken'a showed excellent results in the test from July 1, but in the test from April 1 all failed.

First, one can easily make millions on a known history.
Secondly, not all of the Expert Advisors can be tested on this history. First, you have to convince Metaquotes to create a multicurrency tester :)

If your Expert Advisor survives each of three consecutive quarters without readjustment, it is 99% free of adventurous ideas and other nonsense. And therefore worthy of taking part in the championship. After all, it's only about selection, not about awarding prizes. :-))

And the fact that not all of the Expert Advisors can be tested with the built-in tester is indeed a problem. But the problem is technical, which means that we can find an acceptable solution. The developers will get their hands on this. They have the time.
 
Yurixx писал (а):

If your Expert Advisor survives each of three consecutive quarters without readjustment, it is 99% free of adventurous ideas and other nonsense. And therefore worthy of taking part in the championship. After all, it's only about selection, not about awarding prizes. :-))


Now imagine that you are Stringo and checking an EA before the Championship. I haven't provided you with MQL-code of the Expert Advisor and you are not allowed to decompile it.

Here is my Expert Advisor's algorithm:

If (Today=='03.01.2006') BuyMultiEur();
 If (Today=='24.01.2006') SellMultiEur();
 If (Today=='14.04.2006') BuyMultiEur();
 If (Today=='15.05.2006') SellMultiEur();
 If (Today=='10.07.2006') SellMultiEur();
 If (Today=='19.07.2006') BuyMultiEur();
 If (Today=='Beginning of Championship') DoSellMoney();


How can you tell if he is doing something stupid? :)
 
Better писал (а):

Here's my expert's algorithm:

If (Today=='03.01.2006') BuyMultiEur();
 If (Today=='24.01.2006') SellMultiEur();
 If (Today=='14.04.2006') BuyMultiEur();
 If (Today=='15.05.2006') SellMultiEur();
 If (Today=='10.07.2006') SellMultiEur();
 If (Today=='19.07.2006') BuyMultiEur();
 If (Today=='10.07.2006') DoSellMoney();


How will you determine if he's doing stupid? :)
Well, there's no way to tell without decompiling ;o)))! Well, the man will just get a public disgrace at the end of the championship, if the results diverge in the same way as here : https://www.mql5.com/ru/users/foil#comments.
Who needs it, when there is always a much more attractive way to just adjust EA's parameters to known quotes, which won't be a secret for anyone? I think the second way will be mainly used by everyone. However, no matter how hard you make the rules, you won't be able to avoid exceptions - in any case, someone will send EAs in the Zonker's style, for example.
12345678910
New comment