Moderators: hgm, Rebel, chrisw
Fascinating.Dan Andersson wrote:After 40. Kb3 playing out the pawn endgame after 40. .. Rxd7 is pretty forcing as well as illuminatingThe superficial quip makes me wink impishly.
MvH Dan Andersson
Answering my own question: apparently not. Unlike 40.Kb3? Rxd7 41.Rxd7 which probably is a draw, 40.Kc3 Rxd7 41.Rxd7 appears to be winning.Houdini wrote:Does the same problem exist in the line with Kc3 given by Kim with eval +2.42?
Code: Select all
38/67 1:56 +3.34 42.Kd4 Ke6 43.f4 b6 44.c5 b5 45.g3 Ke7
46.Ke5 a4 47.Kd5 Kd8 48.c6 b4 49.Kc4 b3
50.axb3 axb3 51.Kxb3 Kc7 52.Kc3 Kxc6
53.Kc4 Kd6 54.Kd4 Ke6 55.Kc5 (787.736.387) 6781
TB:534.934 
.
Here is a better line for white which wins:Jack Lad wrote:How about 34.d7 Rcc2 35.Rc4 Rxc4 36.bxc4 h5 37.Kc1 Rd4 38.Kc2 Kf6 39.Rh7 Kg6!
However I feel that there is a good chance there is a forced win for white somewhere in this line.
Correct, your analysis is consistent with what I posted above:Jack Lad wrote:Here is a better line for white which wins:Jack Lad wrote:How about 34.d7 Rcc2 35.Rc4 Rxc4 36.bxc4 h5 37.Kc1 Rd4 38.Kc2 Kf6 39.Rh7 Kg6!
However I feel that there is a good chance there is a forced win for white somewhere in this line.
1. d7 Rcc2 2. Rc4 Rxc4 3. bxc4 h5 4. Kc1 Rd4 5. Kc2 Kf6 6. Rh7 Kg6 7. Kc3
Rd1 8. d8=R Rxd8 9. Rxb7 Rc8 10. Rb2 Kf6 11. Re2 h4 12. Kd4 a4 13. c5 a3
14. Rc2 Rd8+ 15. Ke3 Ke7 16. c6 Re8 17. c7 Kd7+ 18. Kf4 Kc8 19. Kxf5 Re1
20. Kg4 Rb1 21. Rc3 Rb4+ 22. f4 Rb2 23. Rxa3 Rxg2+ 24. Kxh4 1-0
Code: Select all
37/83 74:54 +0.85 34.d7 Rcc2 35.Rc4 Rxc4 36.bxc4 h5 37.Kc1 Rd4
38.Kc2 Kf6 39.Rh7 Kg6 40.Kc3 Rd1 41.d8Q Rxd8
42.Rxb7 Rd1 43.Rb2 Kf6 44.c5 Ke6
45.Kc4 Kd7 46.a4 Kc7 47.Kb5 h4
48.Re2 Rb1+ 49.Kxa5 Kc6 50.Re6+ Kxc5
51.Rf6 Rg1 52.Rxf5+ Kc6 (27.478.478.730) 6113
IMO this isnt correct by Lad or Houdini. After 7.Kc3 (move 40 in the game) Black doent play Rd1 but Rxd7 so that White doesnt get the P on b7. The line isnt mine but from CB.Houdini wrote:Correct, your analysis is consistent with what I posted above:Jack Lad wrote:Here is a better line for white which wins:Jack Lad wrote:How about 34.d7 Rcc2 35.Rc4 Rxc4 36.bxc4 h5 37.Kc1 Rd4 38.Kc2 Kf6 39.Rh7 Kg6!
However I feel that there is a good chance there is a forced win for white somewhere in this line.
1. d7 Rcc2 2. Rc4 Rxc4 3. bxc4 h5 4. Kc1 Rd4 5. Kc2 Kf6 6. Rh7 Kg6 7. Kc3
Rd1 8. d8=R Rxd8 9. Rxb7 Rc8 10. Rb2 Kf6 11. Re2 h4 12. Kd4 a4 13. c5 a3
14. Rc2 Rd8+ 15. Ke3 Ke7 16. c6 Re8 17. c7 Kd7+ 18. Kf4 Kc8 19. Kxf5 Re1
20. Kg4 Rb1 21. Rc3 Rb4+ 22. f4 Rb2 23. Rxa3 Rxg2+ 24. Kxh4 1-0In this variation White always appears to end up with 2 pawns bonus, Black cannot defend his a-pawn.Code: Select all
37/83 74:54 +0.85 34.d7 Rcc2 35.Rc4 Rxc4 36.bxc4 h5 37.Kc1 Rd4 38.Kc2 Kf6 39.Rh7 Kg6 40.Kc3 Rd1 41.d8Q Rxd8 42.Rxb7 Rd1 43.Rb2 Kf6 44.c5 Ke6 45.Kc4 Kd7 46.a4 Kc7 47.Kb5 h4 48.Re2 Rb1+ 49.Kxa5 Kc6 50.Re6+ Kxc5 51.Rf6 Rg1 52.Rxf5+ Kc6 (27.478.478.730) 6113
Don't confuse the 39...Kg6 and 39...Ke6 lines.Rolf wrote:IMO this isnt correct by Lad or Houdini. After 7.Kc3 (move 40 in the game) Black doent play Rd1 but Rxd7 so that White doesnt get the P on b7. The line isnt mine but from CB.
Are you a computerchess expert? Kind of sad if you begin to prove even the best human patzers wrong with their incompetent errors. But that reminds me of similar "analyses" where the position waschanging by chance so that a new judgement came into sight. It took quite a while to discover the incorrectess in the impossible position.Houdini wrote:Don't confuse the 39...Kg6 and 39...Ke6 lines.Rolf wrote:IMO this isnt correct by Lad or Houdini. After 7.Kc3 (move 40 in the game) Black doent play Rd1 but Rxd7 so that White doesnt get the P on b7. The line isnt mine but from CB.
After 34.d7 Rcc2 35.Rc4 Rxc4 36.bxc4 h5 37.Kc1 Rd4 38.Kc2 Kf6 39.Rh7 Kg6 40.Kc3 Black cannot play Rxd7 for obvious reasons.
After 34.d7 Rcc2 35.Rc4 Rxc4 36.bxc4 h5 37.Kc1 Rd4 38.Kc2 Kf6 39.Rh7 Ke6 40.Kc3 Rxd7 we have the long line that was discussed in previous posts - white swaps rooks and wins the pawn ending.
Anand surely must feel some disappointment about not having chosen 34.d7, a natural move he didn't play because of a small tactical oversight (as also explained in earlier posts).
In the commentary to the fourth game today, Jan timman was of the opinion that there was not a forced winning line, at least he did not see it in his own analysis. But not sure he reads Talkchess comments on a regular basis, just talking about his own analysis and other comments he found on the net. But he thinks Gelfand would have been able to find the line he talked about that gives a draw, with the time remaining. There was no analysis board visible so I'm not quite clear exactly what line he had looked at, Dirk Jan mentioned they could not show it right at that moment but that the experts would probably be able to figure it out. I stopped my own computer from the start position, the eval went up at first but the last two iterations clearly went lower again. Luckily there where not yet the very strange scores (Eelco de Groot wrote:At depth 28 the eval has gone up a little. It could be, with mainly Rooks involved a situation where you can only make small progress, I don't know. Most of the players here will be stronger than me and certainly Robert he is a 2100+ Fide rating. But regarding the output, combined with the usual variation in scores, I don't think I can see it just letting the program add one more iteration. But I will try at least to go to depth 29, may take a few hours though, because this is with just 256 Mb hash, 75 % of the Q6700 available. If you go too deep, > 1000 minutes I think the reliability of the hash contents is not so good anymore, so the usefullness of 31+ iterations is a bit doubtful in itself. With very good position learning learning or maybe separate storage of PV nodes in their own transposition table maybe that could be improved. That is just a hypothesis though.
28/72 87:18 +1.32++ 34.d7 Rcc2 35.Rc4 Rb2+ 36.Kc1 Rxa2
37.Rc7 f4 38.Rxb7 (14.442.826.367) 2756
28/72 118:45 +1.40 34.d7 Rcc2 35.Rc4 (19.804.729.682) 2779
Code: Select all
1r3rk1/2q1bp1p/p2p2p1/1p1Qp1Pn/4B3/2P1BP2/PP5P/1K1R2R1 w - -
Engine: Rainbow Serpent 2.2.4 Build 055 (256 MB)
by Tord Romstad, Marco Costalba and Joona
31 1712:09-0.11++ 22.a3 (198.620.587.856) 1933
30 1712:09+M0-- 22.Rg2 Rfd8 23.h4 (198.620.587.856) 1933
30 1712:09-0.27 22.Bc2 Qb7 23.Qxb7 Rxb7 24.Bb3 Rc7
25.Bd5 (198.620.587.856) 1933
best move: a2-a3 time: 1971:32.126 min n/s: 1.933.424 nodes: 198.620.587.856