Analysing a single move in a very interesting position in the Closed Sicilian

Stop and think!

I was playing around with various positions that appeared in the Masters Lichess database when I got to the following position. Go ahead, try to determine what is the best move for White. This is not a puzzle, this is a completely equal position, this is not some mate in N or trick move. It's just pondering at the beauty of chess.

1. e4 c5 2. Nc3 Nc6 3. Bb5 Nd4 4. Nf3 e6 5. O-O a6 6. Bd3 Nc6 7. b3 d6 8. Bb2 Nf6 9. Re1 Qc7

This position has been reached 16 times in the Masters database, one game from Mio Perunovic, with 272 games in the complete game database. Moves followed were Master level moves so there is no imbalance.

Spoilers incoming

OK, so in this position Big Mio played 10. Bf1, which is the most played move by Masters and regular players alike and what Stockfish recommends until ply 20. But then Stockfish changes its mind. The top two moves are always moves of the knight on c3.

So with this hint, what move would you make with the knight on c3?

Another hint: the second top move, the one that only briefly gets to be top move, is Ne2. Perfectly good move, but not best.

After depth 36 a move that was never on the radar comes up. By depth 38 it's considered the top move.

Another hint, the move that SF thinks is the best from depth 20 to 37, with small hiccups for Ne2, is Nb1. Yes, retreating the knight is the best move at those depths!

The answer is... drum roll!... Nd5, also attacking the Black queen.

Analysis

First thing that comes to mind is: wouldn't the knight be captured by the pawn? But then you smugly realize the immediate recapture will attack the knight on c6 with check! Only to see the knight can go to e7 and block the check while attacking the undefended d5 pawn a second time.

What is going on here? If we run the LiChess Tools /piecevalue command, that I just described in the previous blog post, on depth 38, just to make sure we're talking about the same thing, I would not be writing this post because that would mean 32 analyses up to depth 38, which would take an eternity and turn my laptop to hot slag. But even at depth 26 some things are pretty apparent:

The least valuable White piece, according to SF, is the knight on c3. The most important Black pawn is the one on e6. Even without any other considerations, exchanging one for another is not as bad.

But also look at the consequences of that move:

• the rook on the open e-file becomes a monster
• the Black knight on e7 can't move, meaning the f8 bishop can't move and short castle is blocked.
• the Black knight on f6 is attacked by the b2 bishop, so g6 cannot be played, therefore the bishop on f8 can't move that way
• after c4, defending d5, the White queen eyes c2 and a4 (assuming b4 is played
• the d6 pawn will remain backward probably forever

Indeed, the b4 move is the best White move after Black plays 12...Bg4 - only move that maintains equality.

And again, it's a deep line, as at first SF says Black is better, then eval goes to 0, then slightly increased for White after depth 28. However, that is not the purpose of this post. All I want to focus on is why 10.Nd5 is the best move.

What are the consequences of a centralized king who can't castle short or use a bishop and a rook? The only option for Black is to castle long or move the king. Indeed, SF will castle long. and White will immediately attack on the queen-side.

The second best option for White other than 13.b4 is 13.h3. In fact, even with 13.b4 in, SF will play h3 eventually, however 13.h3 is not a good move. While it preserves a strong looking pawn structure and eliminates one of Black's most advanced pieces, it takes the queen away from the queen side and the long castle is now a lot safer. b4 will get played, but the Queen will have lost tempi and has to reroute via g3, eyeing the d6 backward pawn.

And yes, I know that it's a difference of less than a pawn, but we are talking the difference between the chance of a win for White and total draw. BTW, none of the human games did play b4 when they should have and they all ended up with Black winning.

As expected, the b2 bishop is a monster as well as the rook on e1. The a and b pawns you almost want to get rid of, which will happen when you push b4 and maybe even a3. Meanwhile Black would like to push pawns on the king side and the queen is a bit overloaded.

There are other features from LiChess Tools that can help discerning what SF "thinks" and help us find a plan of action, not just a single move we don't understand.

One of them is to highlight the same move in multiple lines: this helps with determining thematic moves that would get executed sooner or later. Another is the button to extend the number of total computer lines to up to 10 (this doesn't work for external engines, only local ones). And, of course, you can set your desired computer engine depth where the evaluation will stop.

So after configuring all that, let's see what SF16 thinks White should do:

As you can see, the only serious option for Black is Bg4, to which White will in a vast majority of scenarios respond with b4, many times followed by a3.

In other words, something like this:

Let's perform the same experiment for Black, after 12...Bg4:

Black wants to castle long, exchange the f3 knight if possible, move the knights around. Look at that rook valued pawn on g7. Of course, its value is given by the hanging knight if it were missing, but it's more than that, surely. Look at that d6 pawn, holding the position like Atlas himself.

I don't know about you, but I would not want to be Black knowing what I know.

Conclusion

Computer engines have this reputation of inscrutable gods: they give you a number to evaluate your position and maybe a good move if they deem you worthy. With some simple ideas Stockfish's oracular tenets can be decoded and useful information extracted about plans, elements of the position and what the enemy is trying to achieve.

In this position in the Closed Sicilian, a beautiful move changes the entire structure of the board. And it happens in a completely equal theoretical position, played by chess professionals as the mainline. We used Lichess and LiChess Tools to discern why that single move is the best move in the position.

Here is more of the line: 1. e4 c5 2. Nc3 Nc6 3. Bb5 Nd4 4. Nf3 e6 5. O-O a6 6. Bd3 Nc6 7. b3 d6 8. Bb2 Nf6 9. Re1 Qc7 10. Nd5 exd5 11. exd5+ Ne7 12. c4 Bg4 13. b4 (13. h3 Bh5 14. b4 cxb4 15. a3 b3 16. Rb1 Nd7 17. Bd4) 13... cxb4 14. Qa4+ Nd7 15. Nd4 O-O-O 16. Qxb4 Nc5

I hope you get something out of this. Time to let my computer rest now.