|
|
18. Nd7 Qe6
|
Better than 18...Qd6 19. Nxf8 Rxf8 20. Qxe7 Qxe7 21. Rxe7, losing the rook. This only loses the exchange and a pawn. |

|
|
|
19. Nxf8 Rxf8 20. Qxe6 fxe6 21. Rxe6 Bb5
|
?? I don't know what I was thinking here. Re8 would have been fine. |

|
|
|
22. Rxe7 a5 23. Re5 c4 24. dxc4 Bxc4 25. Rfe1 Bxa2
|
When down in material, trade pawns not pieces. |

|
|
|
26. Rxa5 Bd5 27. Re7 Bf7 28. Raa7
|
If the bishop moves, Rxg7 Kh8 Rxh7 Kg8 Rag7#
If the rook moves (other than threatening mate, I later noticed), Rxf7 |

|
|
|
28... c5 29. c4
|
Stopping the c pawn, so it can't reach bishop protection. |

|
|
|
29... h6
|
Trying to open an escape for the king. |

|
|
|
30. Rac7
|
At first I thought, "no way to stop Rxc5 and the promotion of the pawn," then I saw a chance. |

|
|
|
30... Rd8
|
Threatens Rd1 , followed by Rxe1#. This also allows me to move the bishop, e.g. 31. h3 Bxc4 32. Rxg7 Kf8 and the checking cannot continue. |

|
|
|
31. h3 Bxc4 32. Rxc5
|
Now all the action is on the kingside. |

|
|
|
32... Bd3 33. Rcc7
|
Making the seventh rank threat again. |

|
|
|
33... g5
|
Avoiding pawn loss. |

|
|
|
34. g3 Rb8
|
Distracting him with some checks? This also allows Bf5 to threaten Bxh3, as the king can head to d8 if checked. |

|
|
|
35. f4 Bf5
|
Not gxf4 gxf4, giving white a passed pawn. Trying to force white to move a pawn (Kg8 or Kh8 is met with Rb2 ). If g4, white has a backwards pawn. |

|
|
|
36. fxg5 hxg5
|
Now h4 leaves white with just a h pawn, which I think gives me the best chance of drawing. |

|
|
|
37. Re5
|
If Bxh3, Rxg5 leaves black with a passed pawn. |

|
|
|
37... Bxh3
|
Rf8 fails to g4, losing the g pawn anyways. Rb1 doesn't accomplish anything; the king can move Kf2-e1-d2-c1, and the rook has to move back to the 8th rank to avoid checkmate, so the bishop is lost. |

|
|
|
38. Rxg5+ Kf8
|
Not Kh8 Rh5 |

|
|
|
39. Rh5
|
Better would be g4, trapping the bishop. |

|
|
|
39... Be6
|
Protecting against Rh8#. |

|
|
|
40. g4 Kg8
|
Moving the king to prevent Rh8, allowing the bishop to move perhaps to f7. An immediate Bf7 would have lost to Rf5.
This does allow the fork Rg5 if the bishop moves to the g file. |

|
|
|