What data do we have to trust?

Luca De Andrea  
Good morning, colleagues, the title of my topic is deliberately provocative. 

The question I want to submit is this: I’m backtesting an EA on MT5 with real tick data provided by my broker IC Markets. I checked and the real tick history of the broker IC Markets is only one year. If I extend the backtest beyond the year, I run errors in the Journal and automatically in the previous period the Every Tick model is used.

So I created a custom symbol and imported the data downloaded with Tickstory, coming from Dukascopy. Here you can see the two backtests, both performed on real tick data. The one done with my broker’s data, IC Markets, is much better.

My question now is, why this difference? Which data is best trusted?

I’ll be grateful for your advice. 



BACKTEST WITH EVERY TICK BASED ON REAL TICK IC MARKETS




BACKTEST WITH EVERY TICK BASED ON REAL TICK - TICKSTORY, DUKASCOPY



CodeFx  
You will most definitely get different results using same EA with different brokers during backtesting, demo and live trading. This is because commission and swap charges differ from one broker to another.
Luca De Andrea  
CodeFx #:
You will most definitely get different results using same EA with different brokers during backtesting, demo and live trading. This is because commission and swap charges differ from one broker to another.

Thanks CodeFx, I think I can consider valid the backtest on my broker’s real tick data only as long as they are provided (one year). On a system that works at low timeframes the real tick is crucial. 

In order to do longer backtests, I think i'm gonna use the data provided by Dukascopy 

Reason: