Automated Trading Championship 2007: common mistakes in experts - page 7

New comment
 

Since the end of August, almost every day I receive an automatic e-mail message saying that the expert has successfully passed the test and has been accepted for participation.

I would like to ask about the meaning of constant automatic checking of the same file, which I have not changed for a month? Can the server not re-check the EA only if a new version of a file is downloaded, especially if the current version of the EA was checked successfully?

I'm not really bothered by it, but would just like to understand the meaning of repeated checks of the same file with a notification sent to the author?

 
A week before EA testing started, I deleted my EA because I decided to tweak the code a bit due to the results of the demo, then (when testing started) I got a message that my EA was successfully tested and accepted - Errors -0, etc.
I have a question, what EA did you test if I deleted it?
And now I keep getting messages:

Checking of your EA is completed. Test report for Expert:
-----------------------
2007.09.08 19:31 loading
invalid configuration
0 min 0 seconds
Errors: 1

Although I have not added upgraded Expert code yet. Please take note. Thank you.
 
Renat:
folver:
i.e. twice I've been emailed that the exponent has passed the test successfully and I'm registered... and on the third one it says that I have an error in the code... why should I change the code of the expert I uploaded it to the profile once?
Judging by the results, the Expert Advisor has caught a margin call. I cannot say anything yet. Tomorrow morning the next checks will be completed and we will check the results.

My Expert Advisor is optimized waiting for the test results ... I got better results with lower quality modelling. ... At 90% the Expert Advisor really got a margin call, but it is hard to understand how it was able to miss two previous attempts with MK
 
folver:
Renat:
folver:
i.e. twice I've been emailed that the exponent has passed the test successfully and I'm registered... and on the third one it says I have an error in the code ... why should I change the code of the Expert Advisor I uploaded it to the profile once?
Judging by the results, the Expert Advisor has caught a margin call. I cannot say anything yet. Tomorrow morning the next checks will be completed and we will check the result.

My Expert Advisor is optimized waiting for the test results ... I got better results with lower quality modelling. ... at 90% the Expert Advisor has indeed caught a margin call, but it is hard to understand how it has always been loss-making twice before.
Apparently, the Expert Advisor has always been loss-making. However, we started showing a margin call as an error (why should we accept an Expert Advisor that is guaranteed to be unprofitable) a few days ago so the first two tests were "without errors".

From now on the excessive amount of generated logs will be considered as an error. If more than 64 megabytes of logs are generated during 8 months of the standard testing, an Expert Advisor will not be allowed to participate. This measure is forced, because currently some experts generate hundreds of megabytes of useless log messages. As we are publishing Expert Advisors' logs on the Championship website in real time, such huge amount of logs is absolutely unacceptable. Last year we disqualified a participant whose Expert Advisor generated about 30 Gbytes of logs during a 24-hour period.
 
solandr:

It doesn't really bother me in principle, but I would just like to understand the point of repeatedly checking the same file and sending a notification to the author?

We fix scripts, add new types of diagnostics and error controls almost every day. That is why we periodically run full rechecks. Only modified EAs are rechecked in the usual mode.
 
Renat:

From today, excessive volume of generated logs will be counted as an error. If more than 64 megabytes of logs are generated during 8 months of standard testing, the expert will not be allowed to participate. This measure is forced, because currently some experts generate hundreds of megabytes of useless log messages. As we are publishing Expert Advisors' logs on the Championship website in real time, such huge amount of logs is absolutely unacceptable. Last year we disqualified a participant whose Expert Advisor generated about 30 Gbytes of logs during a 24-hour period.
My report shows 249 MB of logs for 8 months. And there are no useless messages, the only thing that is shown there is a modification log and order opening and closing during trailing.
Is there an option in the terminal (or in the code) to disable output of "useless messages"?
And I think most EAs that use trailing stop at this limitation. Again, redoing it... Could this have been mentioned at least a month ago?
 
abadan:
According to my report the log for 8 months is 249 Mb. And there are no useless messages, the only thing that is displayed there is the modification log and opening and closing orders when trailing.
Suppose the message on modification takes 250 bytes (this is even too much).
Thus, the Expert Advisor modifies orders more than a million times in 8 months!
That is more than 6000 modifications per day or 250 modifications per hour.


These are all "useful" messages!
 
komposter:
abadan:
According to my report the log for 8 months is 249 Mb. And there are no useless messages, the only thing that is displayed there is the modification log and opening and closing of orders when trailing.


Suppose the message on modification takes 250 bytes (this is even too much).
Thus, the Expert Advisor modifies orders more than a million times in 8 months!
That is more than 6000 modifications per day or 250 modifications per hour.


These are all "useful" messages!



Sorry, a message just popped up when checking the EA, so I got confused. Of course it wasn't 249MB, it was 249kB. I almost scared myself.
 

A small question - if there are no errors during tests and the Expert Advisor is admitted, but during the Championship it produces errors, will the Expert Advisor be disqualified?

I have noticed (in particular during preparation of my EA) that comparison of non-normalized values, for example, in trailing stop conditions may give a signal for OrderModify although the price has not actually changed, as a result the EA executes OrderModify() with the same values and therefore error 1.

I have not noticed this before, and the EA with such code, for example, did not produce errors at the last championship, but now it gives error 1: 

                  if(OrderStopLoss()<(Bid-TrailingPoints*Point)) {
                     OrderModify(.......);
And debug prints gave me in the log "1.97550000 < 1.97550000"... I don't know why, but this way out of this situation, for example:
                  if( NormalizeDouble((Bid-TrailingPoints*Point)-OrderStopLoss(),nDigits)>0) {
                     OrderModify(.......);
and by the way I saw good advice somewhere, I don't remember whose (by Rosh's way), similar comparison operations perform as comparison of normalized result of value difference with necessary error. I hope it will help someone to correct errors.
 
It can be done simply:
bool  modifyResult;                    
oldSL=OrderStopLoss();
newSL=Bid-TrailingStop*Point;
if (MathAbs(newSL-oldSL)>Point) 
{
modifyResult=OrderModify(OrderTicket(),OrderOpenPrice(),NewSL,OrderTakeProfit(),0,Yellow);
if (!modifyResult) Print("TrailingStop:Ошибка модифации ",GetLastError(),"  oldSL=",oldSL,"   newSL=",newSL);
}

Trailing will only take place if the old value differs from the new value by at least one point.
1234567891011
New comment