My Best Chess Ending

Peter P. Wakker - Luc Compagnie (1-0)

POSITION AFTER BLACK'S 54TH MOVE:
White Kf1, Rb7, Rd7, pawns f2, g3, h2;
Black Kg8, Rf8, Bh6, pawns f7, g6, h5.
White to play.

The game had been adjourned in this position. The people present thought that the position was a dead draw. I like this ending because white follows a clear schematic plan and it is surprising that white is, probably, winning. As a preparation, we first consider three known background positions.

BACKGROUND POSITION 1.
White Kg6, rook on f-file.
Black Kg8/h8 and B dark-squared.
Black loses no matter on which dark square his bishop is placed, because the white rook can always combine attacks on the bishop with threats for mate. For example, with the black bishop on g3, white plays 1.Rf3  Bh2.  2.Rf2  Bg3.  3.Rg2  Bd6 (3...Bh4  5.Kh5+, 3...Bf4  5.Kf5+; this is why the bishop has to leave the kingside) 4.Rd2  Be7.  5.Ra2  Bd6.  6.Ra8  Bf8.   7.Rb8 etc.

BACKGROUND POSITION 2.
White Kg4, Rb7, pawn g2;
Black Kg8, Bd2, pawn g6.
This position is drawn, with black builing a fortress. The black king remains on g8 or g7, depending on the white rook. The black pawn remains on g6. The black bishop moves carefully, switching between the diagonals c1-h6 and a1-h8 to prevent the white king from occupying g5 or f6. Note the crucial role of pawn g6. In such positions, black should have its pawns on a color opposite to the bishop so that the pawns, jointly with the bishop, block the white king. Moving pawns to dark squares usually loses quickly because white can invade.
    A version of this background position arose, at the time of this writing, in a game Timman - Morozevitsj, analyzed by John Henderson, Corus 2002 tournament, Wijk aan Zee, The Netherlands, January 17, 2002, with a white h- instead of g-pawn. This gave white the extra plan of putting the king on e6, the rook on the g-gile, and then trying to attack black's pinned g6-pawn with the white h-pawn. Black can defend by a well-timed sortie of king and bishop against the white h-pawn. The game ended in a draw.
   [Added on October 11, 2004: Another version of this background position arose in the 5^th game of the worldchampionship match between Kramnik and Leko, October 2, 2004. Kramnik could have built a fortress and obtained a draw as described here, but missed this chance and lost. Here is an analysis.]

BACKGROUND POSITION 3.
White Ke6, Rd3, pawns f4, g3, h4;
Black Kg8, Bb2, pawns f5, g6, h5.
White wins because he can dislodge pawn g6 through a double pawn sacrifice, g3-g4 and, after taking on g4, pushing the f- or h-pawn. Then pawn g6 disappears and the white king can enter. The lines are complex and are analyzed in Averbakh, Yuri (1985, “Comprehensive Chess Endings, Vol. 2: Bishop against Knight Endings; Rook against Minor Piece Endings.” Pergamon, Oxford. Translated from Russian into English by Kenneth P. Neat; pointed out to me by Paul van der Sterren, letter of June 7, 1990). I did not know about this possibility during the game, and feared a black f7-f5.

In the game against Compagnie, the white plan consisted of four stages.

STAGE 1. Trade the g-pawn against the black h-pawn.
STAGE 2. Trade the f-pawn against the black g-pawn.
STAGE 3. Move the h-pawn to h6.
STAGE 4. Trade the h-pawn against the f-pawn in such a way that Background Position 1 arises.

Black can try to resist at each stage, but white can always push through because the active rooks properly restrain the black bishop. The third stage is the most difficult one. An alternative plan after Stage 1 could be a “frontal” attack with both white rooks on the g-line, pressing on g6, and at a proper moment, with the black king forced to be on the g file after which g6 is pinned, f4-f5 and h4-h5. Given that my plan would probably win, I did not investigate alternative plans.

CONTINUATION OF THE GAME. Moves leading directly to Background Position 1, or rotated versions thereof, are usually not mentioned explicitly. Lines where black puts pawns on dark squares are easily refuted, and are not analyzed here.

STAGE 1. 55.Kg2  Kg7.
I thought it would be unlikely that black would play lines with f7-f5, and dedicated most of the time available to the analysis of other lines. In the time dedicated to f5, not knowing Background Position 3, I did not find a clear plan for white. It was, hence, a relief that f5 was not played in the actual game. Immediately 55...f5 is, by the way, refuted by 56.Rh7, but 55...Bg7.  56.Kh3  Be5 and ...f5 can be. Probably I would have played f4 which may prevent f5 after all.

56.Kh3  Kf6.  57.g4  hg4+.  58.Kg4

STAGE 2. 58...Ke6.  59.Re7+  Kf6.  60.Rec7  Kg7.
60...Ke6.  61.Rc6+  Ke5 (Kd5.  62.Rf6).  62.f4+  Bf4 (otherwise Rf6).   63.Rb5+.

61.f4  Kg8.
61...Kf6.  62.Rc6+  Kg7  63.f5  gf5+ as in the game.

62.Rd7
62.f5 can also be, 62...Ra8 seems to be active but after 63.Rf7  Ra4+.  64.Kf3  Rf4+.  65.Kg3  Rf5.  66.Rf5 or 65...gf5.  66.Rg7+!  Kf8.  67.Rg6 white still wins.

62...Kg7.  63.f5  gf5+.
63...Kg8 or Be3 then 64.Rd6 and black has to take on f5.

64.Kf5

STAGE 3. 64...Kg8.  65.Rbc7
The white rooks are put on the c- and d-file to restrain the black bishop on the c1-h6 diagonal. The reason will become clear in the variation 68...Be3, analyzed after the game.

65...Kg7.  66.h3  Kg8.  67.h4  Kg7.  68.h5  Kg8.
Black decides to put his bishop on the “long” diagonal, i.e. a1-h8. Thus, the white king cannot occupy f6. Now, however, white can easily play h5-h6. Alternatively, black can keep his bishop on c1-h6, in which case 68...Be3 is the natural move. Then the white king will occupy f6. This line is of a different nature, and more complex, than the game, and is analyzed at the end of the game.

69.Kf6  Bg7+.  70.Kg5  Be5.
The active rooks will prevent checks on the c1-h6 diagonal.

71.Rb7  Kg7.  72.h6+

STAGE 4. 72...Kg8.  73.Rb5
73.h7+?  Kh7.  74.Rf7+  Rf7?  75.Rf7+  Kg8.  76.Kg6 does not work because of 74...Kg8! with a draw.

73...Bc3. 
73...Re8. 74.Ra5, the bishop has to move, and Rf5 follows. 74.Rf5  Bb2.  75.h7+  Kg7.
75...Kh7.  76.R7f7+! enforcing an exchange of rooks, and the winning Background Position 1 arises.

76.R7f7+  Rf7.  77.h8Q+  Kh8.  78.Rf7
Black resigns because Background Position 1 arises.

We finally consider 68...Be3 instead of 68...Kg8. The lines are numbered as I.A.1.a etc. The position is: white Kf5, Rc7, Rd7, and pawn h5, black Kg7, Rf8, Bh6-e3, pawn f7.

69.Rc6!
Black is in zugzwang, and has to give up square f6, as follows from lines I-IV hereafter. This position shows why the white rooks are where they are: the bishop has no good square on c1-h6.

I. 69...Bh6?  70.Rg6+  Kh7.  71.Kf6.
II. 69...Bf2.  70.h6+  Kh7.  71.Kg5 Transferring a rook to the f-file and trading the h-pawn against the f-pawn, as in the game.
III. 69...Rh8.  70.Rg6+  Kf8.  71.Rd8+;
IV. 69...Ra8.  70.Rg6+  Kf8.  71.Rf6;
These lines show that the black king has to move, thus allowing white's king to occupy f6.

V. 69...Kg8.  70.Kf6
Rook moves on the sixth rank are less useful than after 69...Kh7 in line VI, where their use will not be very clear either. The white threat in this position becomes clear in line V.A.
V.A. 70...Bg1.  71.Rc7
A forced winning manoeuvre, useful in other lines, so no need to analyze other moves.

V.A.1. 71...Bb6.  72.Rb7  Bd8+.  73.Kf5
V.A.1.a. 73...Kg7.  74.Rb8  Be7.
After other bishop-moves white plays h6-h7 etc.
75.h6+ and Rb6+ or Rf8+.
V.A.1.b. 73...Bh4.  74.Kg4
and h6, Kh5, and the rook is transferred to the f-file, after which the white h-pawn is traded against the black f-pawn, as in the game.
V.A.2. 71...Kh7.  72.Rf7+  Rf7.  73.Rf7+  Kh6.  74.Kf5
(black cannot take on h5) and Kg4. We next turn to other moves than 71...Bg1.
V.B. 70...Bf2.  71.Rc4
(against Bh4+)
V.B.1. 71...Be3.  72.Rc7
as in line A.
V.B.2. 71...Kh7.
To have one tempo more after Rf7+.
72.Rf4  Bg3.
72...Be1 then it can't go to a3-f8 and, hence, 73.Ke7 and soon Rd5; also 73.Rf7+  Rf7.  74.Kf7  Kh6.  75.Rf5  Bh4.  76.Ke6  Bg5.  77.Rf1  Kh5.  78.Kf5 wins.
73.Rg4  Bf2.  74.Rf7+
Or 74.Rd5.
74...Rf7.  75.Kf7  Kh6.  76.Rg2  Bh4.  77.Ke6  Kh5.  78.Kf5
V.C. 70...Bf4.  71.Rc4
Is better for white than line V.B.

VI. 69...Kh7.  70.Kf6
After 70.Rf6  Kg7.  71.h6+  Kh7! I did not find a clear win, neither after 70.Re6  Bc1.  71.Rb6  Be3.  72.Rc6  Kg7.  73.h6+  Kh7! There seem to be many subtle escapes for black. 70.Rf6  Kg7.  71.Rg6+ and Rg2 with sacrifices of the h-pawn and Rd7 to h-file to mate, I did not analyze thoroughly. In several places, lines with a white rook transferred to the g- or h-file may be worthwhile, but I did not analyze them because no need. Only in the following line this plan will be used.
70...Kh6.
Counter-attack on white's h-pawn; 70...Kg8.  71.R6c7  (71.Rc3  Re8) is like line V.A.
71.Rd5
Against Bg5+ and f7-f6; after 71.Ke7+  Kg7 (threatening Bg5+ and f6)  72.Rd5  Ra8, or 72.h6+  Kg8, or 72.Rf6  Bc1, I see no clear win.
71...Kh7.
With the rooks restraining the bishop, black must remain passive.
72.Rd3
When Bg5+ is no more possible; 72.Rc7  Kh6.  73.Rf7  Bg5+
72...Bf2.  73.Rh3
Prevents Bh4+.
With the black t on h7 and the white rook on h3, white wins by Rc7 if no check on a1-h8 is possible, and by Ke7 if no check is possible on a3-f8 (e.g. after 72...Kh6), in each case meeting black's Kg7/8 with pushing the h-pawn. For example,
- 73...Kh6. 74.Ke7+ Kg7. 75.h6+ Kg8. 76.h7+ Kg7. 77.h8Q+ Rh8. 78.Rh8 Kh8. 79.Kf7 and a rotated version of Background Position 1 has arisen.
Or
- 73...Be1. 74.Rc7  Kg8.  75.h6 and f7 will fall whereas white's h-pawn is safe.
White wins in each case.