borisbadenoff: Cyphelium: <borisbadenoff> The problem with 12.- Nf6 is that after 13. exf7+ Kxf7 14. Nf3 the g6 pawn is likely to be lost anyway in view of 15. Ne5+.>Well I see no problem which could lead to losing the g-pawn quickly. 14. .. Ne4. And Ne5+ fails to Bxe5
<19. e1 is interesting. One alternative is 19.- e5, which forces white to sacrifice the exchange with 20. xe5 xe5. White seems to have compensation after 21. d3 (idea h5 and f1) 21.- f6 22. h5+ f7, but if it's enough to win I don't know.>
Well 19. .. Ne5 isn't so a good idea. You overlooked the idea of
19. Re1 Ne5 20. Rxe5 Bxe5 21. Nxd5!! (21. .. exd5 is not possible because of Qf7#) so 21. .. Rf8 22. Qxf8+! Kxg8 23. Nxb6 (axb6 not possible because of Bxh6+ Kg8 Rd8#) 23. .. Kg7 24. Nxa8 Kxg6 and white is a rook up.
Even more I like the line which my engine pointed out to me.
21. Nxd5 Rf8 22. Qxf8+ Kxf8 23. Bxh6 Bg7 24. Rf1+ Kg8 25. Nxe7+ Kh8 26. Bxg7+ Kxg7 27. g5 Qe3+ 28. Kb1 Qg1 29. Rxg1 Black has to sacrifies the queen by itself to stop white from Rf7+ Kh8 Rh7#
<I don't see why 21. Rxd4 is more playable than 22. Rxd4. After 21.- Qxd4 it's the same position, the bishop is just on a5 instead of b4. Then if 22. Ne2, a transposition to the game continuation is likely.>
Well after 21. Rxd4 Qxd4 its the same position with the bishop on a5 and a mate in 3!!
22. Qxe7+ Kg8 23. Qe8+ Kg7 24. Nh5#
So 21. .. Qxa3 22. bxa3 Nxg6 23. Re1
Nxf4 24. Rxf4 And black is pawn up while in the other variation black keeps his one rook up advantage.
<The last fatal error might have been 27. Qg5+. White could have tried 27. g5 instead.>
Well but black would never play Ng4. But instead something like Qe3+ or Qf3 with the idea Rf8