MT4 Strategy Tester - some information needed

Hey Folks,

I'm just experimenting a bit with some basic experts and historic data - just to get a feel for it.
Have data from 2001 until now and test with the full time frame (in case this makes a difference for some questions below).

I have a couple of questions where I can't find an answer anywhere:

 

1. Can I export the results of a back test somehow (to xls/csv etc)?
2. Why is my Tester not showing any Information about modeling quality (see pic). Did I miss a setting?
3. What does the color bar mean (green/pink/grey)?

Thanks.
MikeFT

1. Right click anywhere on the report page and click save as copy.
2. You most likely need more data. Press F2 to see your history center.
3. The following colors are used in the color diagram:

    

   1. Lime - modeling on minutes, marked with 7 in the picture below.
   2. Deeper green colors show modeling on large timeframes, from M5 to H4.
   3. Pink color - pure fractal modeling without data of a smaller timeframe, marked with 2 in the picture.
   4. Gray color - modeling limitation by date, marked with 1 in the picture.

http://articles.mql4.com/c/articles/...ng_quality.png

Great picture

Thanks a lot for spending the time therefor.

Greetings.
MikeFT

I just forgot:
What about this (see picture) - any idea?

See #2

Sorry - I must have over read it. Guess I've been too excited about the picture...

Anyways: I have 10 years of historical 1M Data. The graph is from when I was running Year 1-10.

Also I'm not sure if it is working right. Maybe it is, but I'm just confused (please see pictures).

This Strategy tester really is not my friend - not yet.

You need to use the period converter script.

Why?
I thought if I have M1 that's all the Strategy Tester needs - at least tickwise.
Do I need to manually create all other timeframes too, before it can perform right?

And something else: Do I need to load historical data of the proper Symbol everytime before I let the Tester run ( File->OpenOffline)?
I'm just wondering because it's already been into the history before...

If I think about my picture from post #4 - it says something about 7million ticks.

If I did the math right:
A trading year has approx. 250 days, meaning 3.6million trading minutes.
Seven million tick means approx. 2 ticks per minute. I guess this is all right, even if real minutes may have way more (maybe not on average, but some).
This testing data and if I understood it right, in the best case it never has more than 4 ticks per minute (O,H,L,C).
So everything should be all right...

I still don't understand, why the tester always says n/a for the Modeling quality...
The graph from post #4 is just bright green, so the quality of the data is perfect and 10 years a whole bunch of data...

Read everything including the links mentioned in the article: http://www.easyexpertforex.com/backt...imization.html

I think you are using the Control Points option. If you chose All Ticks mode, you should see the numeric value. 50% is quite high in my experience, I don't know if it is very good, but I suspect not.

Bars in test1097Ticks modelled25847Modelling quality50.00%

That's been helpful. Thanks.

No, I'm using every tick.

Still wondering why it says n/a for the Modeling quality...

run backtest on 1M. Does it still say n/a?

Yeah, you're right.
If I set the period to M1, it shows a value.

But why is that?
The choosen period has nothing to do with the available ticks, does it? As far as I understodd M1 (or whatever lowest time frame is available) will be used for testing.
So whatever I choose for period is only importand if my EA would need a specific period to work right OR if I want to use the visual mode on a specific time frame... Right?

Also: Why would I need to convert a M1 to a higher time frame for the tester? Isn't this done autimatically while testing and using a higher Period?

Thanks.
MikeFT

Why is that?

Because there is data in the history center for M1

The chosen period has nothing to do with the available ticks, does it?

It does. Metatrader doesn't store tick data. It generates tick data as it is needed. I don't fully understand how the interpolation algorithm works, but it's rarely 100% accurate and is dependent on the timeframe.

As far as I understodd M1 (or whatever lowest time frame is available) will be used for testing.

Wrong. Lower timeframes are used to verify but not to generate testing data in the backtester. That's why you see 'mismatched data errors' when backtesting. The highest value in 1M bars between 00:00 and 01:00 does not match the corresponding high value in a 00:00 1H bar. Further, lower timeframes are used to determine if the high was hit before the low, or how many times the open price was crossed.


So whatever I choose for period is only importand if my EA would need a specific period to work right OR if I want to use the visual mode on a specific time frame... Right?

Could you clarify? The point of using any period when backtesting is for time-dependent indicators to be calculated properly.

Also: Why would I need to convert a M1 to a higher time frame for the tester? Isn't this done autimatically while testing and using a higher Period?

Because Metatrader doesn't do it for you. Metatrader merely verifies the data using existing bars data from different timeframes.

I mean if an EA is coded to use data from the current time frame used in a chart and I know it works best on H1, then I set the period in the tester to H1.
But if the EA is coded to use H1 always, no matter what the time frame of the chart window is I should be able to use whichever period I want with the tester and the results should remain the same.

This is an interesting point here. I just checked this.
Used Historic M1, generated daily and it really is as you said - but why?
If the highest M1 candle is 1.5000 during a specific day, the corresponding D1 candle should have the same value. Where is the difference coming from?
This is really confusing me.

My test results have two different High values.
What I did: I made some code to give me the previous days high:
and print it in MY Testers log file
Then I did let the Tester run on M1, stopped after a day and draw some lines.
After that I opened the generated historical EURUSD_D1 and checked the daily candle.
The high values are different (see picture). In this case we're talking about more than 8 pibs. How can that be?

Did you download the daily data using metatrader? Or did you import it?

Most likely, you used two different datasources.

Just get all your data for 1M from one datasource and then use the period converter to convert to the other timeframes.

@Ronald Raygun:
Please note: I made a mistake in the previous post. Updated it already. The daily high and the output from my code do match. Just not M1 and Daily.

The data is from forextester.

I deleted all files in the hitory folder already, imported M1 and created 1440.