Which data to use?

Hi guys!

Im pretty new working with data and have some questions about.

Data differs from one provider to another, they are never exact. With that in mind, the results generating strategies using one or another, will differ in results so… which one should i use?

Lets keep it simple. I generate an strategy using Dukascopy EURUSD M1 data available on SQX.
The strategy is profitable but when i retest it using my exported broker data, the result differs by a lot, making it unprofitable.

Then… should i base my strategies on a quality data as Dukascopy (or whichever) or should i use my broker’s data as is the data my broker has been using?

My fear is building strategies using certain data and the strategies not working that well when using live with my broker.

If the answer is “Build the data using the broker’s data you will be using later” (which has sense for me) does it means that if someday i change broker, should i adapt my whole portfolio to the new data?

Sorry if im asking something really dumb, im pretty new and still trying to figure out how that works hehehe.

Thanks a lot !

The best you can do is use Larger timeframes (H4, D1, W1) because no matter wich database you use, normally you will have noise on the data on small timeframes.

Its possibile to visualize this if you inspect the data on the “Data manager” section > tools > View and Analyze

 

 

Borja

 

I have same problem. Customer servce has help em to know the Darwinex data because I thought were europe saving day zone.

 

But really my Darwinex data is USA trade saving Daylight time (+7 , +8).

Even if you create strategies with higher timeframes, the candels created by dukascopy data  will be different than broker. Backtesting in mt5 only works exactly if I export the dukascopy data to create a custom symbol in MT5 and backtest it. Then works exactly well.

SQX customer service told me to retest on data exported by your broker.

Then I am building this process again:

I Build using symbol cloned to  USA New York Daysaving data.

In multimarket retest  I add  to restest the symbol with my broker data imported from mt5 broker server. in  1 hour bar to get more period.

All strategies survivals will pass  a retest using symbol created by  tickdata Darwinex the SQX has in the data module.

At the moment the first strategies built but not retestes, using uct + 7 saving did not match in mt5 backtesting with SQX, neither utc+7.

I am waiting to the end of robustness test to see if works. Perhaps the symbols need to be retestd in broker data, not only creatd, is what customer service told me.

 

Anyway, even without the coincidence all the backtesting on mt5 are profitable, but I am looking for the confidence on same results.

Finally it seems I found the solution.

 

I build the strategies using dukascopy data for a IS period. Let´s say 2006 – 2020.11.30

And use higher timeframe only.

 

In bulding strategy I retest in cross check two robustness test:

 

the dukascopy data in 1 minute

And in the others markets option I reste my broker data in  1 minute bars.

Due to the fact my broker is Darwinex, I use the high quality Darwinex tick data from sqx, but modified to 1minute bar and to the timezone of the broker utc + 2dst.

I tried to do using the data exported from my broker in MT5, but some periods of bad quality data made issues and problems.

 

Now I have backtested some of strategies and works fine in mt5 strategies tester.

 

The problem is the building is really slow, about 60 strategies per hour, other set up allowed me even 2000 strategies per hour in my laptop; so I need to be patient, because I like to generate thousands of strategies to be retested.

But they are already created with some robustness in the other hand.