Time for Real Back Tests !

New comment
 
This thread is dedicated to back testing. If anyone has concrete methodology, please post it here.

This is what i know:

To get high quality test results, i.e. model quality of 90%, then 1M data have to be imported into metatrader. Alpari-idc, has a data bank with MT4 compatible data. First download the data then import it using the history center in MetaTrader. This way you will get 90% model quality and youcan set the date from around June 2004 till today.

Questions remain, if you write an EA to work on 30M, do you set the period on strategy tester to 30M (this way you get 90% model quality) or is it better to use the 1M period. The results are dramatically different.

If anyone has concrete instructions, would appreciate a post.

Thanks
Aymen
 
If I interpret your statement correctly, the difference is 'cause the expert is running on 1m timeframe as against 30m timeframe and as such is bound to pvoide completely different results.

Did I miss something?
 
AymenSaket:
 Alpari-idc, has a data bank with MT4 compatible data. First download the data then import it using the history center in MetaTrader. This way you will get 90% model quality and youcan set the date from around June 2004 till today.
I tried to import the .hst file with EURUSD M1 using the history center with the North Finance MetaTrader but it ignores, do nothing, when i click at the import/ok button.
No data appear in the list. Can you describe in detail what you did so you can successfully import the data from Alpari-idc
 
I'm facing a similar issue. I'm trying to test an expert, but the data that is being given to the expert seems to differ depending on the timeframe I use. I set the model to every tick, so I didn't expect different time frames to differ in basic data, just in how often bars were closed and the charting, etc. Instead I find that highs and lows for periods of the longer timeframes aren't quite right.

For instance, I have data imported for EURUSD that shows Feb 1 at 8:36 as getting up to 1.2158 in M1, but in M30 the high for the equivalent bar (08:30) is 1.2153. When I examine the journal for the two time periods I get stuff like this:
For M30: 2006.02.01 09:00 NS - IgOr Strategy: Period 1 Order BUYSTOP at 1.21680000, SL: 1.21480000, TP: 1.22780000, Ask: 1.21460000
For M1: 2006.02.01 09:00 NS - IgOr Strategy: Period 1 Order BUYSTOP at 1.21669999, SL: 1.21469999, TP: 1.22769999, Ask: 1.21360001

The ask is 10 pips off, and I know it didn't move that much in that single minute. I can also tell from the journal that it's firing events on more than just every 30 minutes, so I'm pretty sure that both journal lines are actually from minute 09:00.

From this I assume that either the larger time periods are doing some fudging, or I'm not adequately understanding what it's doing. Can anyone shed some light?
 
AymenSaket wrote >>
This thread is dedicated to back testing. If anyone has concrete methodology, please post it here.

This is what i know:

To get high quality test results, i.e. model quality of 90%, then 1M data have to be imported into metatrader. Alpari-idc, has a data bank with MT4 compatible data. First download the data then import it using the history center in MetaTrader. This way you will get 90% model quality and youcan set the date from around June 2004 till today.

Questions remain, if you write an EA to work on 30M, do you set the period on strategy tester to 30M (this way you get 90% model quality) or is it better to use the 1M period. The results are dramatically different.

If anyone has concrete instructions, would appreciate a post.

Thanks
Aymen

Hi Aymen you are one step ahead of me.

I have downloaded history selected my EA ran it and all i get is a mesage indicating the trades were rejected.

Can figure out the problem.

Chris

 
chrisfromoz:

Hi Aymen you are one step ahead of me.

I have downloaded history selected my EA ran it and all i get is a mesage indicating the trades were rejected.

Can figure out the problem.

Chris

It's not alpari data, it's MetaQuotes. But if you download one currency pair's history at a time, close history center, re-open history center, click download again, click re-calculate, then close history center, and run strategy tester on the pair, you get 90% modeling quality.


Don

New comment
Reason: