a3 move, why?

I've been analyzing a game and have following position:

Hello dear!! I think a3 is not bad this move is have Idea first avoid double pawn in c file and actually make a way to push the e pawn in my point of view.

Generally, a and h are used for a few things...  making a hole for an "outside" bishop (can't retreat due to a blocking pawn in your formation), preventing knight or bishop incursions, luft (if castled etc, of course), and sometimes to use a pawn assault on a castled king.

Here, I don't see it either... I see no effect on C from moving A.  ?

Yeah... I guess it's just deep analysis of position that's beyond my skill level. Once computer tells me - a3 is best, I can analyze and see benefits, but if somebody just showed me position, there is no way I would've picked that as best move.

That's basically why I'm asking question - but it seems I won't get definite answer. Thanks for your insight!

I think I figured it out.

White wants to play e4; that's the most logical move in the position.  There follows e4 dxe4, fxe4 e5!, and Black is doing quite well, as the computer will tell you.  If White plays a3 first, then Black doesn't have anything better than e6.  White then plays e4 like above, which reaches the exact same position but with the extra a3 tempo for White.

The tempo a3 is quite useful, as it controls b4, which stops Bb4 and Na6-b4 ideas.  Also, it allows for the b4 advance, which controls c5.  After the d-pawn exchanges for Black's e-pawn, Black had ideas of Bc5 (potentially with Na6 first), putting the Bishop on a very annoying diagonal.

I believe this is why the computer favours a3; it's an extra tempo.  At the same time, I think this is an example of where you don't completely trust the computer.  Engines are quite fallible in the opening, which this clearly is.  White can Bd3 and have a completely fine game, or play the plan of Bd2, Q moves and then 0-0-0, something used by aggressive players in the QG.  White can also play cxd5, transposing into an exchange Slav where Black has clearly misplaced his Bishop.  Computers can't think in plans, which makes them relatively weaker in such positions.

My main takeaway from this position, honestly, is that an early f3 by White in the QG is pretty problematic.  White can't stop the advance e5 from Black.  Black also has the freeing c5 in the cards, and whether White takes it or allows Black to take, then the e3-pawn is weak.  After an eventual dxe4 fxe4, the weak e5-square seems far more important than the half-open f-file. If White had the Nf3 instead, he just has a great position.

The engine rarely favors a3. In some positions thanks to a3 your opponent can't play his most active line, but out of several moves considered good by the engine rarely a3 happens to be the best, moreover an engine is very difficult to play against because it's tactically challenging, it has nothing to do with calm positional maneuvering.

Assuming the position is from a slav, then Bd7 is a rare sideline and according to the database it was played 5 times by GMs. After Bd3 e6 was played and now I have no idea why I would like my bishop on d7, as it is the natural square for my knight if white doesn't exchange on d5.

Bf5, Be6, Bc8 and Bh5 (has very bad results, however, white always had higher rating)  are more common moves. Bf5 seems most natural as black tries to keep his bishop outside the pawn chain as he will play e6 at some point.

Why the engine on this site likes a3 is a good question. Maybe it plans to play c5 and b4. Playing c5 immediately allows for b6 b4 a5.  Having played a3 already would allow to go Rb1.

Btw, I would not go e4 as proposed by SmithyQ just because the pawn structure after dxe4 fxe4 e5 is ugly and I don't see what I get in return for giving black the outpost on e5 and isolating my e4 pawn.

That makes a lot of sense.   My takeaway from the position is that white moved too many pawns in the opening -- it may be a known opening and line and all, but I would not enjoy playing it.   I get it now, but its deeper than I can visualize unless its a really, really good day.

Yeah, I think that's it.

In the line 1.a3 e6 2.e4 dxe 3.fxe e5

The followup 4.dxe Ng4 5.Nf3 Bc5 6.b4 and then either piece to f2 is definitely an engine line (the engine likes white, but who would play this way). But the OP was just asking about the engine's recommendation, and SmithyQ doesn't recommend it.

Although instead of capturing on e5 if white plays for 4.Nf3 exd 5.Qxd4 I wouldn't mind taking white... that said, like most humans, I'd probably go for c5 like you said.

I can't offer any personal insight since players better than me have contributed already.

But after 1 minute this is what evals Stockfish shows me, it likes Qb3 the most:

You are probably confused (unless what everyone has shed some light on the situation) because your analysis is incomplete. This is not the correct way to use computer analysis. Your computer suggests a3, but that doesn't help you. There are lots of questions you should ask before asking the question of why a move is good. Definitely, don't go down the computer line after playing a3. Here are just a few questions you should ask.

1: What move do I want to play. What if I play that move?

Often, computers gives moves that prevent a threat you didn't see. If you don't see the threat, the move seems like a move. Play something like Bd3, and see how black responds )and what the computer's evaluation is.) I find that about half the time, this sheds immediate light on the computer move.

2: Play the plan or goal of the position.

Sometimes, seemingly weird computers are moves that are making preparation you didn't know you needed to make to play the goal of the position. In the position on the board, you have two pieces defending e5 and black has two pieces defending e5. That means you can push e5 without losing the pawn. Push e5, what happens then? See how black responds to this move. If black doesn't have an immediate threat that the given computer move parries, then this approach usually sheds light on why the computer chose its move. It saw a tactical resource black had that you didn't see that prevented and idea you wanted to play in the position.

3: Play a nothing move.

This happens a lot in endgame: you play a move, and you find out that move was "inaccurate" and the computer suggests a move that is better by +1.0. I have spent a lot of time analyzing moves like this only to discover a simple truth: The position was winning for me, so every move moves. It's not a matter of this move or that move, the computer simply prefers the move you didn't play in the game. It is valuable to look at this analysis so that you can get new tricks for your arsenal, but I have played many many endgames where I never game away the win but the computer says I made lots of "inaccuracies" because I played a +3.0 instead of a +4.0. Sometimes, the computer has a good point, but often its just preference and not worth your time to analysis the position.

4: Play the given computer move (a3) then play a natural move as black. Sometimes, a seemingly strange computer move is just making a threat that you don't see. After you play the black pieces, you might find that a3 threatens something very nasty (or something as simply as e4, as some other people are saying.) It's important to see what the natural response is for black against the computer given move if you don't see any plan that its preparing for or any threat that its parrying. This can shed light on the move.

There are other possibilities, but this covers most of them. Make sure to be more inquisitive than just looking at the number 1 move and going down the line. In general, the computer mainline is the most useless line because it includes none of the tactics or ideas that cause the line to make sense, it avoids all of them. If there are tactics you don't see, you're going to have to run into them to understand why the computer makes the moves it makes.

Most of all, and this goes for any move, if you see a computer suggest a move that seems like nonsense, you should try to verbalize why you think it is nonsense and then come up with moves for black (or white) that punish the computer given move. By doing this, you will understand the method behind the computer madness of moves like this.

Yeah - the primary reason why I posted is to get understanding of WHY. And I am truly thankful that you took time to post your response as it gives concrete steps on how to think and get to why. Lots of people post theory and you posted practice. So, again, thanks - I am sure your answer will not only benefit me, but anybody who stumbles upon this topic and is looking for deeper understanding.

One thing you'll notice happening sometimes, when exploring a line with an engine, is that after a few (or sometimes many) moves, the eval will drop back to being closer to zero (after it suggested a weird looking move was really good).

Then you can play a normal looking move instead. Sometimes there's a tactical problem that makes it bad. Other times your move (the normal looking move) was better than whatever the engine liked at first.

Of course the longer you let them think, the less this will happen... but definitely don't trust a weird move you haven't explored. Also it's good to set them to show a few moves. I like the top 3. That way you can quickly see e.g. if #2 is only a few 100ths of a pawn difference, but is a normal looking move, you can just ignore the #1 move and continue analysis using the normal move.

And of course the better you get, the more you'll have a sense for when to ignore the engine. One thing it will never tell you is e.g. practical chances.

It looks like it should be black to move?  But besides that...a3 certainly looks playable, I just dont see that as my first choice.  But beyond that, here is what i think a3 does:

After an eventual Bd3-dc4-Bc4, white can play b4-b5 playing a minority attack on the queenside.  Either opening the c-file, or giving black a backward c-pawn.  

a3 is good in many black openings esp. sicilian. Preventing queenside ideas for white early on.