How much is reliable agree in historical test - EURUSD M1 chart

I create an EA to work in EURUSD currency and realize many tests with historical tests, the base used to tests was from jan/2010 to dec/2014.

In the tests I observes any stranges results, for example the results more recents, 2014 the strategy was good, otherwise, in 2010 was not so good. I am confused if I should trust in this results, because its possible has any gaps in historical base that could compromise the results, and I know that the real tests always is different.

What are yours suggestions? I should trust in the results?

Thanks!