A Mednis Error

Edmar Mednis, Practical Rook Endings (1982) is proving its value as I am using positions from it with a student who wisely requested a focus on rook endgames. Last week, I spent one early morning studying Mednis’s analysis of Ljubojevic — Gligoric, 1979. Ljubomir Ljubojevic won game nine of their match when Svetozar Gligoric missed a critical defensive resource. However, this resource was possible because of an inaccurate move by Ljubojevic.

After finding the analysis instructive, I set up the game to play against Stockfish. The engine showed me that had Ljubojevic played the recommended move, there was another defensive resource available to Gligoric. This resource refutes Mednis’s claim that Ljubojevic missed a win when he pushed his g-pawn before playing 68. c4!

Mednis’s analysis begins after 66…Rg4.

White to move

The game concluded 67.g7 Kb7 68.c4 Rg2 69.Kc3 and Gligoric resigned.

According to Mednis, White should have played 67.c4, when White wins after 67…Rg3+ 68.Kc2 Kb7 69.g7 Ka7 (there is a typo in the book giving Kc7, which the prose makes clear is not the intended line, as it loses immediately to 70.Ra8) 70.Kd2.

Black to move

From Mednis's Analysis

The idea is that White’s king will move towards the rook, up the f-file, and then over to capture Black’s c-pawn. All of this is explained clearly and accurately by Mednis. White’s c-pawn then destroys Black’s defensive fortress with the king safe from skewers by remaining on a7 and b7.

However, Stockfish, which relies on deeper calculation rather than endgame principles, showed that Black can play 68…Kd7!! Playing this position out leads to a drawn pawn ending.

My game against Stockfish from the diagram at the top of this post went thus:

67.c4 Rg3+ 68.Kc2 Kd7! 69.g7 Ke7 (69...Ke6 is also drawing)

White to move

70.Ra8 

I might have tried 70.Kd2 Kf7 71.Rf8+ Kxg7 72.Rf5, but exploring this line later showed that all roads lead to a draw. Still, Black must either understand the needs of the position or calculate with the sort of depth employed by Stockfish.

70...Rxg7 71.Ra7+ Kf6 72.Rxg7 Kxg7 73.Kb3

Black to move

73...Kf6 74.Ka4 Ke5 75.Ka5 Kd6 76.Kb5

Black to move

All of my students learn early on how to draw this position with Black.

76...Kd7 77.Kxc5 Kc7=

Having played against the computer many of the positions from Mednis, Practical Rook Endings, I have observed that his analysis usually holds up to computer checking. In this particular ending, however, he did not look deep enough into a possible pawn ending had Gligoric opted to ignore the usually prudent movement of the king to the crucial a7 and b7 squares.

While ruminating over this analysis during my morning walk with my dog, it occurred to me that analysis presented by Mednis may not have originated with him. A quick check of the endgames published in Chess Informant confirmed my hunch. All of Mednis's analysis appeared in the endgame section of Informant 27. There it is credited to two of the founding members of CI's editorial team: Borislav Milic and Aleksandar Bozic.

Knowledge

When does a player refuse a draw offer in a dead drawn position? If time is a factor, such a refusal could make sense. Often a draw offer is refused because a player does not know the position is a draw, or suspects that the opponent does not have the requisite knowledge to hold the position.

I had Black in this position this morning.

White to move

49.b7??

After this error, the game is a dead draw. White should have played 49.g4, or started moving the king towards the b-pawn. I offered a draw after a dozen moves, having reached this position.

White to move

Instead of accepting the draw, my opponent played another 20+ moves, eventually setting a trap with 84.Rh8?? (White's king was on e4). I could take the pawn, stepping into a skewer. Or, I could take the free rook. After I took the rook, White resigned.

I have played similar endgames before in all sorts of time controls (see one example at "Winning" [2016]). I am guided in the knowledge that my king must remain on the seventh rank and the g- or h-file. With the pawn on g6, the king cannot move. I recall reading about this technique in a book that included a discussion of the resulting skewer tactic if the defending king strays.

However, looking through my endgame books, I could not find the remembered passage. Even so several books contain examples that are close enough that an attentive reader can easily derive the relevant knowledge.

The Books

Nikolay Minev, A Practical Guide to Rook Endgames (2004) shows a stalemate trap when the stronger side has a useful f-pawn, but prematurely sets up the skewer (21-22). From Khiut -- Alalin, USSR 1952.

White to move

1.Kf4 Kf7 2.Rh8??

White sets up the skewer.

2...Rxa7 3.Rh7+ Kf6 4.Rxa7 stalemate.

Yuri Averbakh, Chess Endings: Essential Knowledge (1996) shows an interesting drawing idea from Johann Berger (67).

White to move

1.Kf7

Black cannot get to the seventh rank fast enough, but can avoid checks using the Black king as a shield.

1...Kf5 2.Ke7 Ke5 3.Kd7 Kd5 4.Kc7 Kc5 5.Kb7 Rb1+ forcing Black's king back to the c-file.

Although it was fixed in my memory that I learned the technique employed this morning from Averbakh, it is not in Chess Endings: Essential Knowledge. I did find the idea expressed clearly in Edmar Mednis, Practical Rook Endings (1982), but I've known the technique far longer than I've owned this book. Mednis explains, "the stronger side wants to avoid the following two potential problems: immobilizing his Rook and depriving his King of shelter" (22). Both problems exist in the illustrative diagram. My opponent created the first with 49.b7. Pushing the g-pawn forward introduced the second, but there was no way to dislodge my rook from the c-file.

White to move

In his illustrative diagram, Mednis explains both Black's need to keep the king on g7 or h7, and the rook remains on the c-file, leaving only to check White's king when it gets near its pawn.

Two books that I have had for several years and have spent some time reading explain the ideas, too. One of the critically important blue diagrams in Dvoretsky's Endgame Manual (2003) shows the winning idea missed in Minev's example when the stronger side has an f-pawn (152).

White to move

White wins with 1.f6+ because taking the pawn leads to 2.Rf6+ followed by a8Q, while moving in front allows White to set up the skewer with 2.Rh8. Dvoretsky points out that a pawn on the g- or h-file, however, does not present problems for Black. Although Dvoretsky's description of the skewer does not match my recollection, it may be the book from which I learned this idea.

Jeremy Silman, Silman's Complete Endgame Course (2007) offers three pages of analysis with three diagrams with the white pawn on a7, and three more pages and diagrams with the pawn on a6. These are in the endgames for Class A. I recall that I read about that far within days of buying the book when it first came out. His "A key tactical idea" underneath the diagram below comes close to what I recall studying. Black attempted a "queenside trek" (230).

Black to move

So, I may have learned the idea from Dvoretsky, and certainly encountered it in Silman. It may also be in some other endgame books on my shelf. The simple idea appears in many books. My opponent either lacked this knowledge, or suspected that I did. In the end, he set up a skewer threat that was shocking enough I could have fallen for it on impulse. However, I took a few seconds to assess and grabbed the free rook.