Studying theory: engine vs popular moves

I was studying theory for the caro-kann using 365chess.com in conjunction with my engine. Eventually I ran into a variation (one that I played before), where there were two moves, are more popular one , and a less popular one (scores statistically about the same). However, my engine scores the less popular one a full .48 better than the other move. Here's the specific position incase you're wondering

So really, what I'm asking. When I study theory, should I go with what the engine wants if it scores it significantly better?

If you can't explain why the engine thinks it's better, I'd say go for the position you feel more comfortable in.

I doubt that that difference in evaluation will last to be honest.

Also, computers are notoriously inaccurate when analyzing the position in the opening stages. 

You can not trust computers in the opening.

rubydave engines dont understand openings. Im not hundred percent sure how theory involves but i believe, people checked it already with stronger computers, even though that doesnt necessarily makes it correct.

as Scott says having a feel for a position is much more important. If you have no plan, it doesnt help you to have the best position.

lol.  Don't trust the engine so early. 

Let it think a long time for each of the next 10 moves, backtrack and go again (it uses already calculated lines as reference) and watch the difference drop to nothing.

Even then it wont give you a good evaluation (engines are made for middle games).  But it will help it be a little more realistic by moving it towards the part of the game that yields to calculation more easily.

I would agree to go with the position you feel more comfortable in, but make sure to look at both positions a few moves in to see where they're going.

The Caro-Kann is what I'm currently focusing on as well. I will say in a current online game (which is far past this and I won't discuss anything else about it) I went with Bf5 because the natural plan with the Caro-Kann is to get the light-squared bishop outside the pawn chain before playing e6.

However, you would still likely end up with it outside the pawn chain after Nf6 since either white captures, you take with the g pawn and then white plays nf3 at which point you can play Bf5 or Bg4. Or white retreats the knight and you head down some lines that are easy for black.

I might go with Nf6 next time, seems like a fun approach although I generally expect white to play the advance against me in the first place.

Nf6 is the Korchnoi move right?

Granted engines aren't as accurate in the opening as others have said, however in this case that point is somewhat irrelevant because the Nf6 move does tend to score higher for black in master games over Bf5 so it is a solid choice.

Yeah, Nf6 is a line.

Bf5 is actually better scoring for black. Black may have 0.9% more wins with Nf6, but 9.5% more losses.

They call it the Bronstein-Larsen..I didn't know Korchnoi played the CK.

If you leave the engine on for any length of time it will not say that Nf6 is 0.48 better than Bf5, unless you have a weak engine.

Heh, true.  My houdini after 30 seconds like Bf5 better by 0.05 pawns.

He has played it, obviously not his most famous opening.

Dunno where I got that from though, but seems he did play it in his WC match vs Karpov.

There's no trusting engines in openings or endgames, and there's no trusting humans in middlegames.

Actually wait, you were right...5...gf is the B-L, and 5...ef actually is called the Korchnoi Variation.

In the 365chess database it gives Nf6 2.9% higher chance to win for black over Bf5 and 3.7% higher chance to win for white. So it is much closer there than here and a difference I would certainly go for. Maybe if someone has one of those massive chessbase databases with 15 million games or so they can say what the difference is there as well.

I will generally always go with something that gives me a better chance to win even if it gives my opponent even slightly more of a chance to win as I find draws boring (even as black).

I'm sure I'd feel differently if playing in tournaments, but when it comes to regular play I'd rather go for it and lose then go down a path that tends to lead to a draw.

Ah cool, I was going to remark that it could be the move afterwards.

but if they are not trustworthy in the endgame, how can their middlegame calculation be correct??

i know its a layman question, but i really dont know.

even if it were good in openings, stop using the computer for openings, where is the point?? it tells you a  good move but you dont understand it.