That's a good site. But even on there, the opening Na3 only has 30 games compared to e4's 1.6 million. And yet, Na3 has the highest percentage of white winning. With so few games however its hard to know if that is chance or because it is the best move. Getting an engine to play it a few thousand times will give a better percentage.
Chess games database idea
Even strong engines usually use opening books, because engines are bad at openings, so your idea won't produce very useful data.
With the chess games database (game explorer) on chess.com that allows you to search through openings, it's a bit of a shame that the unconverntial openings only have a couple of games to browse through, even though they are often, such as na3 (with 5 games on database), apparently succesful (for na3, white won 4 of 5 games).
To have a better database of these openings, would it not be possible to get a few dozen different top end chess engines such as Houdini, Rykba and the like, and make them play each other thousands of times with these unconvential openings, to have a better database for people to browse through?
Is this possible, or is there some sort of blockade against having computer games shown?