Backtests run only since 01.01.2013

[render86Participant May 9, 2014 at 10:52 am Post count: 51]

#18095

I have repeatedly downloaded the historical data via Tickstory for EURUSD and GBPUSD since 01.01.2006 but the backtests keep running only since 01.01.2013. Why?

[imported_tickstoryParticipant May 9, 2014 at 12:32 pm Post count: 1479]

#20144

It is likely you have come across a limitation with MT4 which only allows FXT files to be a maximum of 4gb.
In such as case, MT4 back-tester will use only the data within this size boundary and hence why you can’t back-test the entire date range.

Please see this post for further info and workarounds:

viewtopic.php?f=4&t=176

Thanks.

[render86Participant May 9, 2014 at 8:03 pm Post count: 51]

#20145

@tickstory wrote:

It is likely you have come across a limitation with MT4 which only allows FXT files to be a maximum of 4gb.
In such as case, MT4 back-tester will use only the data within this size boundary and hence why you can’t back-test the entire date range.

Please see this post for further info and workarounds:

viewtopic.php?f=4&t=176

Thanks.

How many years of the historical data does 4GB correspond to (with the “Filter duplicate ticks” box checked)? I’ve tried re-downloaing the data limited to the past 5 years but it still appears to be too big – although it used to be fine before.

[imported_tickstoryParticipant May 9, 2014 at 10:11 pm Post count: 1479]

#20146

This will depend on the particular symbol and the accompanying tick activity. As a general rules, 2-3 years is usually under the limit. Outside of this, if you want to find the absolute maximum you will need to experiment to find the boundaries for the symbol you’re exporting.

[render86Participant May 10, 2014 at 12:24 am Post count: 51]

#20147

@tickstory wrote:

This will depend on the particular symbol and the accompanying tick activity. As a general rules, 2-3 years is usually under the limit. Outside of this, if you want to find the absolute maximum you will need to experiment to find the boundaries for the symbol you’re exporting.

I know what the actual problem might be. I had to re-select my MT4 in Tickstory following a change in the MT4 folders path after an upgrade. However, Tickstory does not appear to work with MT4 build 646. Can you confirm this?

[imported_tickstoryParticipant May 10, 2014 at 10:02 pm Post count: 1479]

#20148

We have tested the Tickstory launcher on Build 646 and it appears to be working correctly. We have no other known cause of the issue you have described aside from the FXT file being over 4gb. If you have selected the correct Data Folder in Tickstory and MT4 (and set your Launcher settings to have the ‘/portable” parameter if appropriate), then this shouldnt be the cause of any issues between versions.

[render86Participant May 13, 2014 at 5:23 pm Post count: 51]

#20149

@tickstory wrote:

We have tested the Tickstory launcher on Build 646 and it appears to be working correctly. We have no other known cause of the issue you have described aside from the FXT file being over 4gb. If you have selected the correct Data Folder in Tickstory and MT4 (and set your Launcher settings to have the ‘/portable” parameter if appropriate), then this shouldnt be the cause of any issues between versions.

I have reduced the downloaded historical data amount by one year and the backtest now runs until the end – yet I keep getting “n/a” instead of “99.90%” for the modeling quality on the backtest report. I have tried re-downloading the data but the problem persists. I backtest on EURUSD M5 for the period of 01.01.2007-08.05.2014.

[imported_tickstoryParticipant May 14, 2014 at 1:47 am Post count: 1479]

#20150

Hi Render86,

The only known issues that we know of that causes the “n/a” issue is when the MT4 has not been launched via Tickstory or that the exported data is not being used by MT4 correctly. As such, you can check the following:

– That you have launched MT4 via the Tickstory ‘Launch’ function.
– That the exported data is in the correct data folder.
– If your MT4 symbol is different to that of the Tickstory named symbol (eg. EURUSDm in MT4 but EURUSD in Tickstory), that you have correctly used the ‘Maps To’ field.

If you have already checked these things, then the only other suggestion is to make a small 1-month export and see if that works. It could be that there are still unexpected issues when MT4 internally tries to process large files. If this is the case, you’ll need to reduce your exported range further until it gives the correct modelling quality (or keep working as-is with the potential of other issues occurring).

Regards.

[render86Participant May 14, 2014 at 1:52 am Post count: 51]

#20151

@tickstory wrote:

Hi Render86,

The only known issues that we know of that causes the “n/a” issue is when the MT4 has not been launched via Tickstory or that the exported data is not being used by MT4 correctly. As such, you can check the following:

– That you have launched MT4 via the Tickstory ‘Launch’ function.
– That the exported data is in the correct data folder.
– If your MT4 symbol is different to that of the Tickstory named symbol (eg. EURUSDm in MT4 but EURUSD in Tickstory), that you have correctly used the ‘Maps To’ field.

If you have already checked these things, then the only other suggestion is to make a small 1-month export and see if that works. It could be that there are still unexpected issues when MT4 internally tries to process large files. If this is the case, you’ll need to reduce your exported range further until it gives the correct modelling quality (or keep working as-is with the potential of other issues occurring).

Regards.

I’ve checked all of the above and all holds true. By how many years should I limit the exported data to get a modeling quality of 99.90%? I have run a backtest for GBPUSD M5 and the modeling quality was 99.90% for the same period of 01.01.2007-08.05.2014, unlike with EURUSD M5 where I keep getting the “n/a”.

[imported_tickstoryParticipant May 14, 2014 at 4:30 am Post count: 1479]

#20152

You will need to experiment with ranges for each instrument, bearing in mind that each will have it’s own level of tick activity and therefore resultant file size. Keep reducing the date range for EURUSD by say 6 months until you find the working limit.

[render86Participant May 15, 2014 at 12:49 am Post count: 51]

#20153

@tickstory wrote:

You will need to experiment with ranges for each instrument, bearing in mind that each will have it’s own level of tick activity and therefore resultant file size. Keep reducing the date range for EURUSD by say 6 months until you find the working limit.

I’ve reduced the period for the downloaded data by one year for EURUSD to 01.01.2008-08.05.2014 and started getting the 99.90% modeling quality on the backtest reports. However, when I did a backtest on EURUSD, then on GBPUSD, and then went back to EURUSD to run another one with slightly different EA settings, I could only run it since 01.01.2013.

[imported_tickstoryParticipant May 15, 2014 at 1:56 am Post count: 1479]

#20154

Does it work again if you restart MT4? If so, it sounds like something odd is happening within MT4. I’m guessing that your FXT files are under 4gb? You might need to try reducing the file size further to get some consistency.

[render86Participant May 15, 2014 at 2:05 am Post count: 51]

#20155

@tickstory wrote:

Does it work again if you restart MT4? If so, it sounds like something odd is happening within MT4. I’m guessing that your FXT files are under 4gb? You might need to try reducing the file size further to get some consistency.

No, it’s the same issue even after restarting MT4. Only the lengthy re-downloading of the data helps but only until I run a backtest on a different currency pair – then the problem repeats. The files are under 4gb. I can opt for an even smaller period to have them smaller but that undermines the idea of having a single backtest report over as many years as possible.

[imported_tickstoryParticipant May 15, 2014 at 5:39 am Post count: 1479]

#20156

Understand your predicament, however unfortunately we can only work within the limitations of MT4. We’re already stretching them by allowing back-testing with tick data, so there will be other compromises. I would recommend at least finding the maximum date boundaries and make sure this issue is resolved first (you will need to give some slack as tick data volume will obviously vary from day to day, month to month). It’s seems odd that a restart doesn’t fix things and implies that something else could have changed in-between your back-tests.

[render86Participant May 15, 2014 at 4:19 pm Post count: 51]

#20157

@tickstory wrote:

It’s seems odd that a restart doesn’t fix things and implies that something else could have changed in-between your back-tests.

The only thing that changed was the download of historical data for another currency pair. Does it somehow replace the previously downloaded data for the other pair?

[Author]

Posts