Great EA in backtest! - page 66

 
xxDavidxSxx:
where does 1.89d come from? Is it a version of 88 thats been modified buy one of us or the developers of CT?

1.89b/c/d are succesors of 1.89 originally posted by fikko in post 569

after clean install and fresh historical data imported i'm still getting very good results both v1.85 and v1.89. i don't know what's wrong. working to solve the problem.

 

https://www.mql5.com/en/forum/174806

this is an idea that I think would be good to have one of the competent developers we have on this thread take a look at and run with.

I believe in the concept and would like to see it added to the CT version we are using to help limit the negative impact of news and the hyper market volitility conditions they create.

The idea is simply that since these announcement dates and times are known we can build our program to accomodate around them automatically.

Simply going on standby and closing all open orders just prior to these events and reopening a hour or two afterwards when things have past the most explosive and inherently unmanageable times like yesterday's announcement.

I would like to see someone put together an #include file that can be easily used to input dates and times to do this. It will make it better for everyone when this is available.

Dave has mentioned repeatedly how news is difficult. Well why not remove that difficulty with a handy add on to the EA?

 
kalamari:
1.89b/c/d are succesors of 1.89 originally posted by fikko in post 569 after clean install and fresh historical data imported i'm still getting very good results both v1.85 and v1.89. i don't know what's wrong. working to solve the problem.

I am using 1.89d and 1.85f

One thing I have found significant recently to reduce drawdown is to change

StopLossIndex = 2.5 to StopLossIndex = 5

running with this at 5 has reduced the drawdown significantly.

attached is my statement from my live account this week....

the first 6 trades were from the EA, trade 7 was a manual trade that I took 2 pips on. The EA made the next 3 trades the first at .5 and then I reduced the maxlots=.3 to preserve what it had just won. turns out it was a good thing because the buy @ .3 on 2006.10.6.12:52 was the EA's response to the announcement which I didn't see coming. I made one manual trade @ .1 lots just after the announcement and then the EA made one @.3 lots after I did. I made one more experimental trade at .03 before I quit for the week. I was wrong on it.

had the EA had the auto news 'go on standby' feature I just mentioned in my last post it would not have taken that -10.50.

 
kalamari:
i've opened offline stored data - it looks ok. backtester caches data for testing purposes. i've also attached a chart preared by bt. wtf?

that right image is showing you the tick by tick representation that the tester generates.

 

i've opened offline stored data - it looks ok.

backtester caches data for testing purposes. i've also attached a chart prepared by bt. wtf?

is it how the data is prepared for test or is it sth wrong?

Files:
offline.gif  25 kb
backtester.gif  16 kb
 
Aaragorn:
https://www.mql5.com/en/forum/174806

this is an idea that I think would be good to have one of the competent developers we have on this thread take a look at and run with.

I believe in the concept and would like to see it added to the CT version we are using to help limit the negative impact of news and the hyper market volitility conditions they create.

The idea is simply that since these announcement dates and times are known we can build our program to accomodate around them automatically.

Simply going on standby and closing all open orders just prior to these events and reopening a hour or two afterwards when things have past the most explosive and inherently unmanageable times like yesterday's announcement.

I would like to see someone put together an #include file that can be easily used to input dates and times to do this. It will make it better for everyone when this is available.

Dave has mentioned repeatedly how news is difficult. Well why not remove that difficulty with a handy add on to the EA?

thats a great idea Or even better..if it could detect excessive volitility. Like in a trade or 2. So it don't keep placing 5-10 trades durring too much volitity.

Because the basic news times are wel known...there are 4(2 durring london, and 2 durring U.S.). Ther are alot of other times that news can come out but not always. Like after US session and befor london session. And scattered through out london and u.s. too.

4 basic news times...all at EST...4,5 am and 8:30,10 am. But there are more at 2 am 7 am 9 am 12 pm,4:30pm, 5pm, 7and 7:30 pm, 9 pm.

I am manually setting no news times each and every day on the one I run on real account. And some days there are no news at all. some weeks no news at all. We cann't just set it to exclude all the possible news times because it would only trade a few hrs a day.

If we could enter specific times with dates at the beginning of each week would help. But for back testing it remains a problem.

My back tests have the 4 major times taken out.

 

Notice the drawdown is under 25% and that is with a risk setting of .25 not .025!

This is the difference changing the StopLossIndex = 5

 
Aaragorn:
that right image is showing you the tick by tick representation that the tester generates.

thx Aaragorn.

i'm runnig another test for 1.85 and after 24 months 14k pips were made. i still don't know what's wrong with my test. kat's and nikkeifx results are very different.

 
xxDavidxSxx:
thats a great idea Or even better..if it could detect excessive volitility. Like in a trade or 2. So it don't keep placing 5-10 trades durring too much volitity.

Because the basic news times are wel known...there are 4(2 durring london, and 2 durring U.S.). Ther are alot of other times that news can come out but not always. Like after US session and befor london session. And scattered through out london and u.s. too.

4 basic news times...all at EST...4,5 am and 8:30,10 am. But there are more at 2 am 7 am 9 am 12 pm,4:30pm, 5pm, 7and 7:30 pm, 9 pm.

I am manually setting no news times each and every day on the one I run on real account. And some days there are no news at all. some weeks no news at all. We cann't just set it to exclude all the possible news times because it would only trade a few hrs a day.

If we could enter specific times with dates at the beginning of each week would help. But for back testing it remains a problem.

My back tests have the 4 major times taken out.

Neither my friend that started that thread nor I have quite enough programming savvy to pull it off ourselves. Making a working include file presents me with some programming issues I don't fully understand how to get it to compile and manage the variables and such. But it would be great if someone that sees the value in the idea would take up the task and make something with easily used external inputs for the dates and times as well how long to sleep when they are triggered.

 
kalamari:
thx Aaragorn. i'm runnig another test for 1.85 and after 24 months 14k pips were made. i still don't know what's wrong with my test. kat's and nikkeifx results are very different.

I appreciate yours and everyone's efforts on this thread. One thing I believe wholeheartedly is that we can do more together than any one of us can do alone.