Queens Gambit Declined - Problem like Win

Here is a game I play in the final round of the Pillsbury Memorial tournament in MA. My opponent neglected his dark squares around his King and allows a beautiful problem like manuever to force the win.

Looks like 18...b4?? is the losing move.  Black had to defend against the Qd2-Qh6-Qg7 mate threat with 18...Nb6 threatening 19...Nd7 and trying to exchange the knight for the White Bishop.

True, but there are better ways to do it than retreating the Knight to where it came from. For instance, 15. ... f5 .

Obviously 18. ... b4 is the losing move because it came right before my tactical shot. He really got himself in trouble with g6, weakening his dark squares far too much.

I haven't really analyzed it with any real depth. I was just very pleased with myself for finding Nb3-Qd2 over the board.

I don't think Qd8 was a good move for black; gift of the Queen. Move 21.Qg5 for white was better. Black would have moved his queen. Where does not matter. White mates in 2 from here.

Actually, Qd8 is forced. Any other move is met by Qh6 and mate on the next move. Black needs to remove the Bishop on f6 to avoid mate and the only way this can be accomplished is by trading for the Queen. I expected my opponent to resign after 20. Qd2!! but he had me play it out. 

If White were to play 21. Qg5, this is actually a mistake because of 21. ... b2! and now Black gets two very powerful connected passed pawns on the b and c files. The way I played it wasn't exactly the best ( 21. e5! ) but it gets the job done.

Do you understand the tactics involved here?

Btw...Fritz finds Nb3 but it actually takes it a bit of time before it considers it. Pretty neat to see the machine think about all the "obvious" moves before it realizes that Nb3 is the winning move.

Nice QGD

what about Nxc4 instead of Nb3? lol

 19. Nxc4?! fails to 19. ... dxc4 and now the Black Queen can reach h5. The only way for White to prevent this is by playing 20. e5 but now 20. ... h5 is able to meet any mate threat with Kh7.

This may be ok for White, but 19. Nb3!! is immediately crushing and far more accurate.

if dxe4 instead of pawn takes knight?

Black's 17...Be6? was a blunder - he should have played 17...f6. It literally sealed his fate, and ensured his loss.

18...b4?? was the final blunder which made his position hopeless - apparently black had no clue about your beautiful impending attack - 18...Qa6 19.Nxc4 dxe4! 20.Qd2 (20.Ne5 e3!) Bf5 might still have saved the immediate loss, if not the game.

Excellent attack strategy - a class game!

 The Queen is attacked, so black must respond to that threat.

 I think my analysis below will show that 18. ... Qa6 is also losing, but there are definitely some alternatives for black that won't lose. I won't go into them now because it is fun to demonstrate why Qa6 isn't good. (Atleast I think it does; I haven't used an engine to check any analysis, I did this all real quick at work.) Atleast Qa6 coupled with the plan to trade the Bishop off on f6 is losing. I can't see any other plans after Qa6.

A sample line may go : 19. e5 Nb6 20. Nb3! (the point being the same that if cxb3 then Qd2 ends the game) 20. ... Nd7 21. Nc5! and if 21. ... Nxc5 then 22. Qd2 should lead to a forced win and if the Queen retreats, say 21. ... Qc8 then 22. Nxd7 Qxd7 23. Qd2 is lights out again. See the viewer below.

NOTE: I haven't checked any of this analysis with an engine and for all I know there is a refutation to 20. Nb3! in the sample line. If there is, I'd like to know what it is because I cannot see it.

Done. After 18. ... b4?? 19. Nb3!, black is forced to trade his queen for the bishop and knight. However, the alternative 18. ... Qb6 avoids losing any material with just a slightly inferior position (-0.20). If 19. Nxc4! dxc4 20. Qd2 Bf5! 21. e5 Nc5 22. dxc5 Qxc5+ 23. Rf2 Rfc8 24. Qh6 Qf8 25. Qxf8 Kxf8 and material is equal with Rybka considering white slightly better (+0.30).

Therefore I don't think 17. ... Be6 was a blunder, or even a mistake, and in fact it is a pretty solid move. But it might be considered an innacuracy (Nb6 may have been stronger), and it certainly made black's life difficult since he was finished if he didn't find 18. ... Qb6!.

EDIT: My bad, you wanted to analyze the variation. After 18. ... Qa2, black is ok, but not great, since although if 19. Nb3 Bf5!, and black is fine, if 19. e5! Bf5 20. Ne4 h6 21. Qd2 Kh7 and black is ok but in a bad position (around -0.80).

18. ... Qa2? 19. e5! Nb6?! is not a big mistake, but it is not great, because of 20. Nb3 or 20. Ne4, both of which lead to similar lines and leave white with a positional advantage. If 20. Nb3 (Ne4) Rfb8 21. Qd2 Qc8, black can be ok, but her majors are cramped in the corner (it's about -0.40). if 20. ... Nd7? 21. Nc5!, then Qa3 is the only response. 21. ... Qc8?? simply drops the knight on d7 (black cannot recapture due to the mating threat), and 21. ... Bf5?? loses big time to 22. Qd2. So 21. ... Qa3 22. Qd2 Nxf6 23. exf6 Kh8, and black can hang on, but with a huge positional deficit (about -1.50).

 Thank you! I noticed the ...Qa3 resource when I got home so that the Black queen can go to f8 after Rfb8 but other than that I think I had the right ideas for the position.

The nature of these lines show how sharp the position actually was, although it looked not that dangerous.