Annotated chess game

This was a surprise for me; at first blush this seems to move the queen out of the battlezone, for which reason I hadn't given it more than a passing glance. Instead I had looked at likely lines after 21.Qc4, where I considered 21...Qd6 to be my best reply, on which I would expect 22.Ba3. Now the sharp 22...Rac8 23.Rb3 Rxc4 24.Bxd6 Rd4 25.Rxc3 Rxd5 26.Bxe5 Rexe5 27.Bc4 Rd2 gives White the advantage, but I thought I would be able to improve that line a bit if it came to that. Alternatively I could go for 21...Bh2+, conveniently tucking the bishop out of the way (!) after 22.Kh1, and now either 22...Qe5 23.Re1 Rac8 24.Qd4 with unclear play or 22...Re4 23.Qxc7 Bxc7 24.Rxe4 Bxe4 with equality (eg.: 25.Rd1 Rd8 26.Bc4 c2 27.Rd4 Bf5 28.g4 Be5 29.Rd2 Be4+ 30.Kg1 Bc3 31.Re2 Bxd5 32.Rxc2 Bxc4 33.Rxc3 Bxb5 =)

I'm not very sure how Black can respond if White plays 21.Qc4. But to me this looks like trapping White's queen. A plan for Black in the event of 21.Qc4 would be to avoid queens exchange and then place a rook on c8.

The head aches. I know very well this kind of games. An you play with a +210... you sure spent a good time on this one!  . --- By the way, white can consider Rc4, followed by g4, and if black puts a rook on c8, Rc6. After this, black can play Be4, further restricting white's queen movement (QxB, Bh2+, KxB, RxQ).

That's complete!

This seemed to me the best option. If Qf3 really was a mistake, then it would be because the queen has been removed from active operations on the Q-side. My queen, along with the powerfully sweeping bishops, had a chance to prove decisive on that flank - if I acted now. Incidentally the GK analysis supports this view, giving Black an almost ½ pt advantage despite being a pawn down in material. I spent quite a lot of time looking at alternatives. But 21...Bh2+ is less good here as it is difficult to see how Black can continue after 22.Kh1 Qe5 23.Bd3. If 23...Bxd3 then 24.Qxd3 and 24...Qe2 25.Qxe2 Rxe2 26.Kxh2 is clearly lost for Black, while I might as well resign after 24...Rad8 25.g3 Rxd5 26.Qf3 c2 27.Kxh2 Rd1 28.Rxd1 cxd1=Q 29.Qxd1. For a while I thought 21...Qc5 would be best, but then I saw 22.Ba3! which gives White control of the a3-f8 diagonal, as well as the important d1 and d6 squares.

22.a3 is the simplest reply here.

Rb3 won't do because of Bc2.

A bit passive square for the rook.

I'd like to be able to use a decent engine. This is far too complicated for amateur review like mine!  . But I can't see why black can't cash the pawn inmediately - even if the black's bishop pairs whorts more than a pawn, for sure!

I gave my 64bit Stockfish half an hour to chew on this question, and it came up with the following lines. In brief, it seems that I should actually have played 23...Rac8 with the idea of pushing the c-pawn and using that as a lever to neutralize White's budding K-side initiative and break down the defence of his b- and d-pawns. I give the lines in full here, but please note that chess engines do get progressively less accurate the further you move into one of their lines: 23...Rac8 24.Bd3 Bd7 25.Rh4 f5 26.Bf4 Qxa3 27.Bxd6 Qxd6 28.Rc1 c2 29.Bf1 Rc5 30.Rb4 Rec8 31.Rb2 Kf7 32.Qb3 Qxd5 33.Qxd5+ Rxd5 34.Rbxc2 Rxc2 35.Rxc2 Rd1 36.Rb2 f4 37.f3 Kf6 38.Kf2 Be6 39.Be2 Rd4 40.b6 axb6 41.Rxb6 Ke5 42.Rb2 Rd7 with complete equality. Interestingly this line diverts significantly from the KG suggestion. As an alternative it also suggests (after 23...Rac8 24.Bd3 Bd7): 25.Bf4 Qxa3 26.Bxd6 Qxd6 27.Rc1 c2 28.Re4 Rxe4 29.Qxe4 Bf5 30.Qe2 Bxd3 31.Qxd3 Qb4 32.d6 Qb2 33.Qd2 Qxb5 34.d7 Rd8 35.Rxc2 Qxd7 36.Qxd7 Rxd7 37.Ra2 Kg7 38.f4 Kf6 39.Kf2 Kf5 40.Kf3 Rb7 41.Ra5+ Ke6 42.g3 and Black has a clear advantage. The GK analysis is inaccurate, btw. as 25.a4 c2 26.Rh4 is refuted by 26...Qc3! with Black advantage: (Stockfish analysis) 27.Re4 Rxe4 28.Bxe4 Qxf3 29.Bxf3 Rc4 30.Be2 Rxa4 31.Bd3 Ra2 32.Be3.

This is one of the subtle positions where one does not readily understand why an engine likes or dislikes certain move!

Actually, I didn't suspect that White would go for the exchange sacrifice until this move was played; I'd thought he would try to break through on the h-file... but taking my bishop would wreak serious holes in my K-side defences and I spent some time looking at the wreckage and figuring out how to handle the ensuing chaos. The rook would be gone, of course, leaving the queen and bishops as his only active agents on the K-side. With both of them operative, I was dead. But could he keep them both? Turns out he couldn't; that advanced c-pawn saves the game.

So better to take it when the rook was not on c4. Maybe this is the reason the GK engine shows Bc5 as inferior. True is that now white has g4, and if the bishop retires, Rxf7, followed by Bh6 or Bc5... and white recovers its breath with edge.

This is actually a defensive move aimed at eliminating the c1 bishop. Though there is a real sting too: White can hardly afford to ignore the threat of the advancing c-pawn. Attempting to strengthen the K-side with 24...Re7 fails due to 25.d6! and Black has (due to the threat against a8) nothing better than to offer the exchange by 25...Rxe2.

White might have kept on to the bishop with some advantage though. On 25.Be3 we would likely see something like 25...c2 26.Rxf5 Rxe3 27.Qxe3 c1=Q 28.Rxc1 Bxc1 29.Qxc1 gxf5 and Black's open K-side position means he will have difficulties keeping the passed d-pawn in check.

It glances g4 as an alternate try.

Interesting complications arise after 25...Rxe2!? However, after 26.Rxf7 Re7 (26...Rae8 27.Bh6!) 27.Rxe7 Bxe7 Black is again faced with 28.d6!

You can clearly see why Bc5 was a mistake: it gave white a free tempo to build an attak on the kingside. Black's key was not to leave this tempo to white. Now the result is uncertain (to me!).

Why didn't you consider Qg3+

This is the point. Black has eliminated one of the bishops, leaving White unable to force a conclusion on the K-side. On looking at this position again, shortly after posting my move, I thought I had missed the best line however, as 26...Bf8 might have been better. The very simple refutation of that move, 27.Bd3!, escaped me at the time...

What about Bd3 at once? How can black defend Qxh7+?

I was looking closely at the alternative, 27...Qb4. But after 28.Bd3 Qd4 29.Qxh7+ Kf8 30.Qh6+ Qg7 31.Qxg7+ Kxg7 White can enter a probably-won endgame with 32.Rxc3 Red8 33.Bc4 Rac8 34.Rc2 Rc5 35.Bd3 Rcxd5 36.Be2.

In such few moves, the situation is completely reversed. Black is struggling now. Giving back the pawn at such low cost means the strategic defeat. Now, let's see if white can materialize a win.

This was the logical consequence of not doing 23...Rac8. Black didn't properly support the c-pawn and it thus lost the weight it needed in order to force White to divert resources to blocking it. Leaving White a free hand on the K-side eventually forced Black to sacrifice the c-pawn in order to save the game.

"Objectively", it is clear that White is better here. But computer analysis does not tell the whole story. If one is in her/his comfort zone, this may already surpass a +1/+2 advantage.

I was feeling comfortable here, as Qg6 guaranteed me either a certainly-drawn endgame or (should White refuse the proferred exchange) a double-edged position with good counterplay. The line preferred by GK, 28...Gh6, didn't enter into my calculations at all, and even with GK's recommendation I wouldn't take it. It leaves Black with no shelter against White's roaming queen and massive problems keeping those pawns back. After 29.Bh5 Re1+ 30.Kh2 f6 31.f4! (Stockfish' suggestion) Black's position is truly awful.

Precisely. This was the situation I was aiming for when I planned the bishop exchange with 26...Bxc1. One of my rooks is already behind the passed pawn, the other will swing out in front of the pawns shortly, and together they will work to pinch them off. If those pawns were connected, or the white king could go to aid in their defence and make trouble for the rook on e2, it would be another matter. But the King is going to be locked up before it gets a chance to do anything, and the queen will be unable to steal tempi (by checking my king) except for certain, very restricted files and diagonals. The main danger of situations like this - a queen which can "magically" teleport itself to virtually any field on the board - has now been eliminated.