Accuracy of TickData Testing

[Fiverr]

Does anyone have experience with Tickdata Testing on multiple computers? I am getting two different results on the same tickdata from two different computers. See attached.

[tickstory]

Hi Fiverr,

Looking at the results it appears you are using CSV2FXT to generate your back-test databases. We don't support this script since Tickstory already has a feature to output HST/FXT data, but I suggest doing the following to trouble-shoot:

1) Copy the HST/FXT files across the two machines. Do you still get different results? If so, something odd is happening. Otherwise;
2) Copy the CSV data to the other machine and re-generate your HST/FXT files with CSV2FXT. If there are differences in the back-test, then make sure your CSV2FXT parameters machine between machines. Create a .set file on the originating machine and load it on the other.
3) If there are still differences, check your CSV2FXT versions are the same.
4) Open the FXT file in Tickstory->MT4 FXT Editor and compare the two sets of information to make sure they match on both machines.

I would recommend that you use Tickstory's 'Export To MT4' function if there is not a good reason to do so. You can save your settings in an .mt4config file which can be shared between machines (similar to the .set file) and should ensure all the important broker information is the same across the two machines.

Hope this helps.

[Fiverr]

Thanks for your prompt reply. I will give it a go.

[Fiverr]

I believe that one set of tick data FXT came from Tickstory while the other came from TickDataDownloader. Which version of FXT is correct? Shouldn't they be similar since they both came from Dukascopy?

[tickstory]

Hi Fiverr,

If the two extracts were produced from CSV files, there could potentially be the following differences:

1) The CSV tick data from SQL Downloader vs Tickstory is different (this should not be the case as you mention since both are sourced from Dukascopy).
2) The import process into CSV2FXT has been incorrect. Check that each file was loaded correctly into CSV2FXT and that the fields were correctly detected.
3) If there were any changes to the CSV2FXT parameters then this will cause back-test differences. Step 4) mentioned above will help identify if there are any such differences in the "Metatrader Info" information.

Hope this helps.

[Fiverr]

Attached are backtesting results from 2008 to 2014 with 99% modeling of TickData vs CSV2FXT. As you can see the general equity shapes are similar; however, the profits are different.

[tickstory]

Hi Fiverr,

I assume that you have managed to resolve the major differences by making sure the two FXT headers are the same? I note that in your back-test report there are significant differences in the number of bars and modelling ticks between the CSV2FXT & Tickstory datasets. If you are working to get the back-tests the same, I would suggest reducing the export sample and then analysing the trading results of your EA under both export samples.

Regards.

[danny_pip]

Hi Fiver,

I 've also had the experience that there are differencies.
I've also found out the TickDataDownloader is not always downloading perfectly all files; sometimes data are missing.

Regards,

Danny

[Fiverr]

Attached is my settings of CSV to FXT script. Can someone provide me the default TickStory setting so that I am comparing apple to apple?