Single games of chess can routinely have a significant impact on opening theory, of course, but it should be quite uncommon for a single game to significantly advance endgame theory, since that's a phase of the game where we can and do have some absolute knowledge that will never change, and since the chance combination of a situation that is ripe for something fundamentally new to be discovered along with a player capable of doing the discovering in the course of a game is a rare thing.


I am looking for pointers to historical instances where the course of a game's play revealed new theoretical endgame knowledge that was unknown before that game was played, e.g. an unknown winning method in a certain type of endgame, or an unexpected drawing resource that had escaped notice until that point.

  • Interesting question. I think this is going to be a hard one because the rules governing the movement of the rooks and minor pieces are old. I believe there are studies from the late medieval period for the Rook that are still valid today that predate even Lucena and he lived during the first half of the 14th century. I think Al Adli did some endgame studies that are still valid in the 9th century? Aug 14, 2012 at 19:14
  • @Ed Dean: Curious, why you would look for such examples? Also, based on what little I know of chess history, patterns in game play were not directly related to the performance of those patterns in some cases; meaning some patterns in game play were just cultural memes.
    – blunders
    Aug 14, 2012 at 20:59
  • 1
    @blunders, my interest here springs purely from an appreciation of fine chess and history. It's easy to find games that are classics because they feature a brilliant attack, say, or a display of superb endgame technique, etc. But when this question struck me, I couldn't think of any particular instances of what I'm asking for (though I think there probably are some), so I'm hoping someone can enlighten me.
    – ETD
    Aug 14, 2012 at 21:25
  • @RobertKaucher, your point is a good one, and the fact that many parts of endgame theory are quite old does mean that answers to this question will almost certainly involve relatively minor advances in endgame theory. But while we won't get, say, that first ancient game where someone shocked the world by demonstrating that K+R vs. K is indeed a win, I'm hoping there will be some modern instances where less fundamental, but still reasonably widely applicable theory was forged during a game :-)
    – ETD
    Aug 14, 2012 at 21:49
  • It depends how basic is the endgame theory you think about. My impression is that as far as "simple" (read: few pieces) endgames go, studies have done a much greater contribution than actual play. Aug 16, 2012 at 22:18

5 Answers 5


In a game at the 1954 Olympiad in Amsterdam, Botvinnik had the white pieces against Nikolay Minev of Bulgaria, and reached the following position after 57. Qxe6:

[FEN "8/8/4Q3/k6K/8/6P1/8/q7 w - - 0 1"]
[StartFlipped "0"]

In Half a Century of Chess, Botvinnik writes,

Ten years prior to this game I had such an ending against G. Ravinsky ... There I did not understand the specific character of the ending and I tried, as in a rook ending, to keep my king on the eighth rank to support the promotion of the pawn at g8 and, this made the win most difficult.

That earlier game against Ravinsky had been analyzed by Keres, but still little had been properly understood about this ending. For instance, in a note after 61. ... Ka4 in the Minev game, Botvinnik points out, "Minev willingly occupies a4 with his king, as this was recommended by Keres in a well-known article where he analysed the ending of my game with Ravinsky." Instead, it turns out that the best place for the black king in this situation would have been the corner a1 square. There was an adjournment at move 73, and Botvinnik notes,

Although I did not find the correct plan before the adjournment, nevertheless I instinctively avoided moving my king to the eighth rank. After thorough analysis I finally found the right way of playing this ending. ... The winning plan consists in placing the white king on the same rank (or file) as the black king, or on the adjacent one. In this case White has a good chance of sheltering his king from checks. Once the method is found, of course, it looks simple.

Finally, after White's 91st move, Botvinnik's plan had reached fruition:

[FEN "8/6P1/8/2KQ4/k7/8/7q/8 w - - 0 1"]
[StartFlipped "0"]

Black has as many as three checks, but they all lead to the exchange of queens. Here the game ended, Black resigned, but the first page of genuine theory was opened on the ending "queen and pawn against queen".

For yet another source commenting on the ending to this particular game, here is Giddins in his The Greatest Ever Chess Endgames:

Relatively little was known about this type of endgame at the time this game was played, and despite a few inaccuracies in the play, Botvinnik first demonstrated the winning method for such an ending in this game. ... Without the aid of either tablebases or any substantial theoretical practice, apart from his own game against Ravinsky years before, Botvinnik effectively "solved" this ending in his adjournment analysis, identifying all the key elements of the winning method in such positions.


I do not think it's really a new technique discovery, still, this game has certainly inspired some people to fight harder in seemingly hard to win endgame positions: Martin Ortueta vs Jose Sanz Aguado. After 28 ... dxc4 we enter an endgame where 3 pawns poorly structured manage to dominate a rook and a knight.

[FEN ""]
[Event "Madrid"]
[Site "Madrid"]
[Date "1933.??.??"]
[Round "?"]
[White "Martin Ortueta"]
[Black "Jose Sanz Aguado"]
[Result "0-1"]
[WhiteElo "?"]
[BlackElo "?"]
[ECO "C00"]
[Opening "French"]
[Variation "King's Indian attack"]
[EventDate "?"]
[PlyCount "70"]

1. e4 e6 2. d3 d5 3. Nc3 Nf6 4. e5 Nfd7 5. f4 Bb4 6. Bd2 O-O 7. Nf3 f6 8.
d4 c5 9. Nb5 fxe5 10. dxe5 Rxf4 11. c3 Re4+ 12. Be2 Ba5 13. O-O Nxe5 14.
Nxe5 Rxe5 15. Bf4 Rf5 16. Bd3 Rf6 17. Qc2 h6 18. Be5 Nd7 19. Bxf6 Nxf6 20.
Rxf6 Qxf6 21. Rf1 Qe7 22. Bh7+ Kh8 23. Qg6 Bd7 24. Rf7 Qg5 25. Qxg5 hxg5
26. Rxd7 Kxh7 27. Rxb7 Bb6 28. c4 dxc4 29. Nc3 Rd8 30. h3 Rd2 31. Na4 Rxb2
32. Nxb2 c3 33. Rxb6 c4 34. Rb4 a5 35. Nxc4 c2 0-1

Now it might be historical because Petrosian mentioned this endgame as having made a great impression on him.

Related link : http://timkr.home.xs4all.nl/chess/rxb2.htm

  • Wow! Thank you, Mog, and welcome to chess.SE. I did not know that game, and I am very perplexed by the additional existence of the game Tylkowski-Wojciechowski as mentioned in the related link you gave, since it ends with the exact same combination after a very different lead-up. I absolutely love stuff like this. (That said, I think what you say at the beginning of your post is right, and that while this is a truly spectacular endgame combination, it's not really an advancement of endgame theory.)
    – ETD
    Aug 17, 2012 at 16:46
  • 3
    Caruana recently used a very similar idea in his game against Nisipeanu, and tweeted "That was my Ortueta-Sanz"! chess.com/news/…
    – Maxwell86
    Jul 14, 2015 at 7:07

By Capablanca's own admission, one of the games that advanced his understanding of the endgame (and probably that of others) was his game against Richard Teichmann in 1913:

[FEN ""]
[Event "Berlin exh"]
[Site "Berlin"]
[Date "1913.11.20"]
[EventDate "?"]
[Round "2"]
[Result "1-0"]
[White "Jose Raul Capablanca"]
[Black "Richard Teichmann"]
[ECO "D63"]
[WhiteElo "?"]
[BlackElo "?"]
[PlyCount "75"]

1.d4 d5 2.Nf3 Nf6 3.c4 e6 4.Bg5 Be7 5.Nc3 Nbd7 6.e3 O-O 7.Rc1
b6 8.cxd5 exd5 9.Bb5 Bb7 10.O-O a6 11.Ba4 Rc8 12.Qe2 c5
13.dxc5 Nxc5 14.Rfd1 Nxa4 15.Nxa4 b5 16.Rxc8 Qxc8 17.Nc3 Qc4
18.Nd4 Qxe2 19.Ncxe2 Rc8 20.Nf5 Kf8 21.Nxe7 Kxe7 22.Nd4 g6
23.f3 h6 24.Bxh6 Nd7 25.h4 Nc5 26.Bf4 Ne6 27.Nxe6 Kxe6 28.Rd2
Rh8 29.Rc2 Rc8 30.Rxc8 Bxc8 31.Kf2 d4 32.exd4 Kd5 33.Ke3 Be6
34.Kd3 Kc6 35.a3 Bc4+ 36.Ke3 Be6 37.Bh6 Kd5 38.Bg7 1-0

It was a situation where Black's isolated queen pawn, and the resulting "holes" that it left in his position for White's knights, more than cancelled out the advantage of Black's Bishop pair. Later, these same holes (for White's king) enabled him to win with opposite colored Bishops and only one extra pawn.

In another game against Marshall, Capablanca sacrificed a pawn to get a rook on the "seventh" rank (Marshall's second), which was at least enough compensation. Hence Marshall had to "play for a draw" by exchanging all his other pieces, even with a pawn ahead. When he failed to do so and "played to win," Marshall lost:

[FEN ""]
[Event "New York"]
[Site "New York, NY USA"]
[Date "1918.11.01"]
[EventDate "1918.10.23"]
[Round "8"]
[Result "0-1"]
[White "Frank James Marshall"]
[Black "Jose Raul Capablanca"]
[ECO "D64"]
[WhiteElo "?"]
[BlackElo "?"]
[PlyCount "78"]

1.d4 d5 2.Nf3 Nf6 3.c4 e6 4.Nc3 Nbd7 5.Bg5 Be7 6.e3 O-O 7.Rc1
c6 8.Qc2 dxc4 9.Bxc4 Nd5 10.Bxe7 Qxe7 11.O-O Nxc3 12.Qxc3 b6
13.e4 Bb7 14.Rfe1 Rfd8 15.d5 Nc5 16.dxe6 Nxe6 17.Bxe6 Qxe6
18.Nd4 Qe5 19.Nxc6 Qxc3 20.Rxc3 Rd2 21.Rb1 Re8 22.e5 g5 23.h4
gxh4 24.Re1 Re6 25.Rec1 Kg7 26.b4 b5 27.a3 Rg6 28.Kf1 Ra2
29.Kg1 h3 30.g3 a6 31.e6 Rxe6 32.g4 Rh6 33.f3 Rd6 34.Ne7 Rdd2
35.Nf5+ Kf6 36.Nh4 Kg5 37.Nf5 Rg2+ 38.Kf1 h2 39.f4+ Kxf4 0-1
  • Tom, I added a link to what I think must be the Marshall-Capablanca game you had in mind. I'm letting you know just in case you actually intended a different, similar game.
    – ETD
    Aug 17, 2012 at 17:10

A famous game is Timman - Velimirovic. According to the endgame theory of that time, the ending was indeed winning, but the shortest win would exceed the 50-moves rule, so it seemed that the game would end in a draw. After the third(!) adjournment of the game, Timman and his seconds found a shorter win than the textbook, within 50 moves, and Timman went on to win the game. There is an interesting article, written by GM Ree, which describes the atmosphere of the time in which this game was played.

However, for the endgame theory of today, this game might be less important, as we have tablebases and no more adjourned games. According to the tablebase, black's move 68....Kf8 was a mistake.

[FEN ""]
[Event "Rio de Janeiro Interzonal"]
[Site "Rio de Janeiro BRA"]
[Date "1979.10.02"]
[EventDate "1979.09.23"]
[Round "8"]
[Result "1-0"]
[White "Jan Timman"]
[Black "Dragoljub Velimirovic"]
[ECO "D30"]
[WhiteElo "?"]
[BlackElo "?"]
[PlyCount "205"]

1.d4 d5 2.c4 e6 3.Nf3 c5 4.cxd5 exd5 5.g3 Nc6 6.Bg2 cxd4
7.Nxd4 Bc5 8.Nxc6 bxc6 9.Qc2 Qe7 10.O-O Bd7 11.Nc3 Nf6 12.Bg5
O-O 13.Bxf6 gxf6 14.Rac1 Bb6 15.e3 h5 16.h4 Kh8 17.Ne2 Bg4
18.Nf4 Rac8 19.Bh3 Bxh3 20.Nxh3 Qe5 21.Nf4 Rg8 22.Qe2 Rxg3+
23.fxg3 Bxe3+ 24.Kg2 Bxf4 25.Qxe5 Bxe5 26.b3 Kg7 27.Rfd1 Rc7
28.Rc2 Bd6 29.b4 Bxb4 30.Rxd5 Kg6 31.Rd4 a5 32.g4 c5 33.gxh5+
Kxh5 34.Rd5+ Kh6 35.Rc4 a4 36.Kf3 a3 37.Rd6 Re7 38.Rxf6+ Kg7
39.Rf5 Rd7 40.Rg5+ Kf8 41.Rg2 Rd5 42.Ke4 Rh5 43.Kf3 Ke7 44.Kg4
Rh7 45.Rf2 Rg7+ 46.Kf5 Rh7 47.Rf3 Kf8 48.Rh3 Rh5+ 49.Ke4 Ke7
50.Kf4 Kf6 51.Kg4 Kg6 52.Rf4 f5 53.Kf3 Kf7 54.Ke2 Ke6 55.Rc4
Kd5 56.Rf4 Ke5 57.Rc4 Kd5 58.Kd3 Ba5 59.Rh1 Bd8 60.Rf1 Be7
61.Rcf4 Bxh4 62.Rxf5+ Rxf5 63.Rxf5+ Ke6 64.Rxc5 Bf6 65.Rc6+
Ke7 66.Ke4 Bb2 67.Kd5 Kf7 68.Re6 Kf8 69.Ke4 Kf7 70.Kf5 Kf8
71.Kg6 Bc3 72.Ra6 Bb2 73.Ra7 Ke8 74.Kf5 Kf8 75.Ke6 Kg8 76.Rf7
Bc3 77.Rf3 Bb2 78.Ke7 Kh7 79.Rg3 Kh6 80.Kd6 Kh5 81.Kc5 Kh4
82.Rg8 Be5 83.Kd5 Bb2 84.Kc4 Bf6 85.Rg6 Bg5 86.Kd5 Bc1 87.Ke4
Bb2 88.Kf5 Kh5 89.Rd6 Kh4 90.Rd3 Bc1 91.Rc3 Bb2 92.Re3 Bc1
93.Re1 Bd2 94.Rh1+ Kg3 95.Rd1 Bb4 96.Rd3+ Kf2 97.Ke4 Ke2
98.Kd4 Bc5+ 99.Kc4 Be7 100.Rh3 Bd6 101.Kb3 Bf8 102.Rh8 Bd6
103.Ra8 1-0

Another game that comes to mind is Fischer - Taimanov, where white exploited the advantage of bishop vs. knight. I'm not sure whether this game is the very first one in this kind of endgame to actually advance the endgame theory, but it definetely is a textbook example.

[FEN ""]
[Event "Fischer - Taimanov Candidates Quarterfinal"]
[Site "Vancouver, Canada"]
[Date "1971.05.25"]
[EventDate "1971.05.16"]
[Round "4"]
[Result "1-0"]
[White "Robert James Fischer"]
[Black "Mark Taimanov"]
[ECO "B47"]
[WhiteElo "?"]
[BlackElo "?"]
[PlyCount "141"]

1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Qc7 5. Nc3 e6 6. g3 a6
7. Bg2 Nf6 8. O-O Nxd4 9. Qxd4 Bc5 10. Bf4 d6 11. Qd2 h6
12. Rad1 e5 13. Be3 Bg4 14. Bxc5 dxc5 15. f3 Be6 16. f4 Rd8
17. Nd5 Bxd5 18. exd5 e4 19. Rfe1 Rxd5 20. Rxe4+ Kd8 21. Qe2
Rxd1+ 22. Qxd1+ Qd7 23. Qxd7+ Kxd7 24. Re5 b6 25. Bf1 a5
26. Bc4 Rf8 27. Kg2 Kd6 28. Kf3 Nd7 29. Re3 Nb8 30. Rd3+ Kc7
31. c3 Nc6 32. Re3 Kd6 33. a4 Ne7 34. h3 Nc6 35. h4 h5
36. Rd3+ Kc7 37. Rd5 f5 38. Rd2 Rf6 39. Re2 Kd7 40. Re3 g6
41. Bb5 Rd6 42. Ke2 Kd8 43. Rd3 Kc7 44. Rxd6 Kxd6 45. Kd3 Ne7
46. Be8 Kd5 47. Bf7+ Kd6 48. Kc4 Kc6 49. Be8+ Kb7 50. Kb5 Nc8
51. Bc6+ Kc7 52. Bd5 Ne7 53. Bf7 Kb7 54. Bb3 Ka7 55. Bd1 Kb7
56. Bf3+ Kc7 57. Ka6 Nc8 58. Bd5 Ne7 59. Bc4 Nc6 60. Bf7 Ne7
61. Be8 Kd8 62. Bxg6 Nxg6 63. Kxb6 Kd7 64. Kxc5 Ne7 65. b4
axb4 66. cxb4 Nc8 67. a5 Nd6 68. b5 Ne4+ 69. Kb6 Kc8 70. Kc6
Kb8 71. b6 1-0

Edit: I found a very nice collection of famous endgames, subdivided in categories. It comes as no surprise that this collection also contains some of the examples that were given in several answers.


I'll have to look for the reference, it was in one of my chess books, and I gave my library away. But the game involved two GMs or GM-level players from back in the day, where one of them figured out that a K+N+N can mate K+P, if the P can move and prevent stalemate.

  • Tony, perhaps you have in mind the game Znosko-Borovsky - Seitz, Nice 1931? (chessgames.com/perl/chessgame?gid=1151993) This is the oldest win of a KNN v. KP I can find. Any game that will count as an answer here would need to predate Troitzky's thorough analysis of this ending, which appeared as early as 1937 (see chess.stackexchange.com/a/696/167). Seitz's 1931 win fits that bill, but it also seems likely that it was already known at that time that KNN v. KP could sometimes be won. So far I can't find a source that asserts/denies that Seitz was breaking new ground.
    – ETD
    Aug 16, 2012 at 17:46
  • 1
    Well, I had a book by Znosko-Borovsky that I recall reading. However, what I remember more than the game was an annotation stating that the K+N+N win was something new.
    – Tony Ennis
    Aug 16, 2012 at 23:25
  • 1
    Reviewing the game, that's it. 62. Bxf5 was a joke move, Z-B played it thinking the game was drawn. He said that Seitz looked at him sadly and said, "Yes, that used to be so."
    – Tony Ennis
    Aug 17, 2012 at 0:26
  • Any idea which Z-B book it was? How to Play Chess Endings seemed like a good guess, but I looked and the game's not in there. (Oddly, a game Z-B - Seitz, Nice 1930 is in the book, which features a materially balanced knight+pawns ending.)
    – ETD
    Aug 17, 2012 at 16:54
  • The book is gone now, but this looks like it: amazon.com/How-Play-Chess-Openings-Dover/dp/0486227952/…
    – Tony Ennis
    Aug 17, 2012 at 21:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.