12 May 2022

Surprisingly Strong Move

In mid-April, I played a move that was better than I thought. I played it in desperation because I was frustrated after reaching what looked like a drawn ending from what seemed like a clear advantage. It was a ten minute game and it sent me to my endgame books as well as extensive engine analysis.

Pawns were exchanged on move 33 and we shuffled our kings and rooks for another 20 moves before I pushed a pawn.

White to move
53.d5!

I expected 53...exd5 54.Kxf5 and maybe I can do something with my f-pawn. But I also thought that would still likely lead to a draw.

53...Kxd5

I was sure this move was an error. It equalizes material, but now Black's king gets cut off.

54.Rd2+ Kc6

I easily won Black's pawns. The rooks came off the board on move 77 and my opponent resigned two moves later.

To my surprise, the engines evaluate 53...Kxd5 as no worse than 53...exd5. That sent me into the endgame books to study rook endings with two pawns to one, three pawns to two, and four pawns to three. Such positions are often drawn.

Catching my eye as suitable for practice was a tragicomedy from Hebden -- Wood, Hastings 1994/1995.

Black to move
Wood's move here resembles what I thought to be an error in my game.

1...Kxf4??

Mark Dvoretsky writes, "A terrible error. The black king will be cut off along the f-file now, and the g7-pawn will be inevitably lost" (Dvoretsky's Endgame Manual, 5th ed. [2020], 189).

Dvoretsky gives two ways to draw. I opted for the harder one against Stockfish on the iPad.

1...g6 2.Rf7+ Ke4 3.Kg4 Ra1 4.Re7+ Kd5

White to move
My king is cut off from the pawns, but White's king has no points for penetration. Nine moves later, Stockfish moved the rook from the e-file, allowing my king to return. Another dozen moves and I offered a chance for the engine to force the exchange of rooks. My confidence that I could draw the pawn ending was grounded upon extensive practice with a pawn ending from Dvoretsky (see "Opposition and Outflanking").

White to move
Stockfish did not opt for 22.Re2+ Rxe2 23.Kxe2 Ke6, which I knew to be drawn. Ten moves more of rook shuffling, and Stockfish pushed a pawn.

Black to move
A Philidor position seems likely, but instead the game concluded with a rook exchange that left lone kings. It was good practice. My play was not error free, but one can also back up against a machine.

Reuben Fine, Basic Chess Endings (1941) offers positions where the weaker side's king is cut off from the pawns. No. 390b is presented without a diagram. No game is referenced, although the pawn structure is close to No. 390 (Saemisch -- Spelmann, Teplitz-Schoenau 1922) with colors reversed. Fine claims that White wins No. 390b (Saemisch was able to hold a draw because his king was with the pawns), according to Fine. Stockfish does not agree.

Black to move
1...Rb7

Stockfish gives 1...Kd6 2.f5 Ra3+ 3.Kf4 Kd7 and better for White by less than a pawn, usually an indication that the game is headed towards a draw.

2.f5 Ra7 3.g5 Kd6 4.h4

Stockfish recommends 4.Re8

4...Ra4 5.Re1

Black to move
5...Ra1

A common idea in such positions, but wrong according to Stockfish.

5...Re3+ 6.Kg4 Kd7 and the engine sees Black holding the position. White's Re8 is not possible.

6.Re8

White will give up the f-pawn for both of Black's pawns after Rg8.








No comments:

Post a Comment