Also, black can't put his Queen on 7th or 8th row, as White's Qh8+ will win. So, no more attacks on the White king from the side, and the back is covered by pawns.

And Black can't put his Queen on the diagonal a2-g8 because of White's Qf6+ and Qe6+

In addition, White can kick the Black king out at will. It's really over:

64.Kg6,Qb1+
(It doesn't matter much what Black plays, White put pawns on f5 and g5 in any case)

65.f5,Qd3 (nothing to do, really)
66.g5,Qc2
67.Qg7+,Ke8
68.Qf7+,Ke8
69.Kg7

click for larger view

<JohnBoy: <ChessYouGood> - black should NOT have resigned earlier. White was having a hard time finishing the job, and black appropriately demanded a proper finish.>

<furrer: I agree [with] John Boy, and don’t we all remember an endgame very much the same as this one?

Topalov vs Anand, 2005 >

<JohnBoy> is absolutely right. Queen endgames are very, very tricky and almost always worth playing out. In the game cited by <furrer>, which ultimately was drawn in 97 moves, both of the elite players involved went astray. Specifically, the position after 73. … Qxg5+:

click for larger view

is theoretically drawn, but Anand soon went wrong with 76. … Qg5+?. (The only drawing move here was 76. … Qh3+.) The position at this point became a theoretical win and remained so until Topalov erred with 90. h7? (when either 90. Qh5+ or 90. Qg4 were the only theoretically winning moves). From that point to the end, the game was a theoretical draw.

Thus, in the game cited by <furrer>, which was played at a regular time control in the 2005 World Championship tournament, the two players currently [i.e., as of September 2009] rated #1 and #2 in the world each committed a half-point blunder in the ending; and, just to illustrate how tricky these endings can be, although such a blunder from any elite player would be extremely unlikely, at either move 94 or 96, if Toplaov had tried to avoid a repetition by playing 94. Kh6?? [or 96. Kh6??], Anand would have won on-the-spot by checking on the 6th rank.

click for larger view

is a theoretical win. There were several points after that where White either missed the fastest win or actually allowed the position to become a theoretical draw. For example, 68. Kf6? threw away half-a-point (it was the “only drawing move”; in the position after 67. … Qc2, any other move that did not hang the White Queen was a theoretical win). Black quickly went wrong, however, with 69. … Qc3+ (when 69. … Qc7 was the only drawing move).

A couple of dozen moves later, 93. Kf6? was again an “only” drawing move. Unfortunately for Black, he went wrong with 96. … Qb4+ (instead of 96. … Kg7!, a good “only drawing move”). In reply, White failed to find 97. Ke3! (only winning move), but after 98. Kc4, Black did not find any of three drawing moves (98. ... Qc8+ or 98. ... Qe4+ or 98. ... Kg7), and after 98. … Qc6+?, the position was winning for White and remained so to the end of the game (even though White did not always play the fastest winning move).

It should finally be noted, however, that as late as move 108, in this position:

click for larger view

the plausible 108. Qg4 would have allowed a draw with 108. … Qf2+ (among other moves). After 108. Kh5!, Black’s next two moves hastened his loss by walking into a forced mate on the light squares.

If 58...Qc2+, then 59. f5 Qc7 (forced) 60. Qe6+ Kh8 (<60...Kf8 61. f6> Qc2+ 62. Qf5) 61. f6 Qh7+ 62. Kg5 wins.>

48.Qb4+ angling to exchange queens by checking (or pinning) from f4 or h4. e.g.

I agree. After I posted I found <58. Kg6 Qc2+ 59. f5 Qc7 60. Qe6+ Kh8 61. f6 Qh7+ 62. Kg5> Qg6+!, and white cannot take the ♕ because of stalemate. (There are other stalemate threats in alternate lines as well.)

click for larger view

If 46...Ke7 47.Qb7+ followed by h7

If 46...Kf7 47.Qb7+ followed by h7

I never thought playing a queen endgame with two pawns up would be so difficult to win... Pferd, RandomVisitor, and Peligroso Patzer, your analysis are all correct.

I don't want to blame the time trouble, since it existed both for me and my opponent, but I don't remember myself practicing 3-4 move forced wins in queen endgames ever. Usually, players put great amount of training to solving complex combination quizzes, but when it comes to put similar training to increase endgame technique, many are lazy, including myself : - )

So I missed 48.Qb4 and 46.Qa8+!... Rest was really long... I knew it had to be a theoretical draw when my king started to walk around the middle of the board, but somehow it managed to turn back to the edge of the board and hide.

I uploaded a bunch of new games recently, and some of them included minor notation mistakes so I had to get in touch with the site admins for corrections, and then, Daniel was kind enough to make one my games as the "game of the day" : -)

I suggest you to checkout my simul games against Kasparov and Karpov. I managed to get a full point from these two world champions, and no, I wasn't playing Kasparov's moves with white against Karpov with black, and vice-versa... Both games are pretty interesting, I'll update some analysis later on... In summary, in both games, I managed to hold on after disadvantageous openings.

Kasparov simul game
Kasparov vs T Karatekin, 1997

Karpov simul game
Karpov vs T Karatekin, 1997

Smitten, Patrick Hummel was a first or 2nd year student at Caltech at the time of the game.

Carlsen is the highest rated player in the world during that game. Gashimov is also among top 10 I believe. He was also one of the highest rated blitz players on ICC.

And their blitz encounter results in a queen endgame.

White is a pawn up, 3 pawns vs 3 pawns on king side, white has the c pawn extra.

Well, white could not force win under time trouble. Instead, he attempted a similar king march, the one I attempted against Hummel, and it led to a forced perpetual check. After 67. c7??, Black has perpetual check, and it should be draw. (Carlsen must have lost on time) Instead, 67.Qb6 wins.

This was a blitz game, I am sure under regular timing controls such mistake would not happen, but it shows the automatic response of a top GM. Just walking the pawn forward with queen support is not enough to win. Somehow, the queen must also take part in the defense of the perpetual check threat, while still helping the pawn advance.

End game technique is very automatic, you make a move mostly based on your end-game instincts, and under current timing controls, when the endgame stage is reached, usually not much time is left to calculate as well.

So, I don't think even the very best practice queen endgame technique that much. I had asked my Ukrainian teammate in Turkish league Moiseenko (ranked top 50 in the world), and he also admitted queen endgames being difficult even at their level, showing a few of his own games.

I am sure there are GMs out there who studied queen endgames in and out, but for the most players, queen endgame technique is lower on their to-train list.