|
< Earlier Kibitzing · PAGE 5 OF 5 ·
Later Kibitzing> |
Jul-23-19
 | | Richard Taylor: <cunctatorg: One of the most interesting from A to Z (and dramatic) games of chess literature!!
Korchnoi's resourcefulness and creativity here are so remarkable as his lack of vision (a certain lack of combination vision or loss of faith?) at move 55; a draw then...> No. The term for Korchnoi is as always:
TIME TROUBLE
He was an addict of the above drug. |
|
Jul-23-19
| | Richard Taylor: <Howard: What I fail to understand is why the computer gives a two-pawn advantage to Korchnoi even though the position at that time is an absolute dead draw ! > A lot of engines do that. You have to be wary, and wait to see if the evaluation changes. If it stays 2 pawns down it is probably a theoretical draw. Even though it is a theoretical draw it is still true that the player with a piece has an advantage. And, by the way, in some lines either player can force a win. The engine takes into account material and I presume the potential to win. Something like that. |
|
Dec-05-20
 | | HeMateMe: wonderful, hard fought draw. great rivalry! |
|
| Dec-05-20 | | areknames: The last 15 moves are superfluous but yes, what a titanic struggle (and a nice pun)! 55.Be4+??? is of course one of the worst blunders in WC history. |
|
| Dec-05-20 | | Ironmanth: Utterly ballsy king walk! Great game; some silliness though for the last many moves. How many hours for them to play this; anyone?! Thanks, chessgames. Y'all please stay safe out there. |
|
| Dec-05-20 | | areknames: <How many hours for them to play this; anyone?!> Not sure about the hours but it was adjourned twice at move 42 and 92 and played on 27, 28 and 30 July. Game 6 which was a short draw was played on 29 July. |
|
| Dec-05-20 | | Ironmanth: Thanks, areknames! Now that I think about it, I recall this game a bit hazily. They played that match in my junior year at UVA. |
|
| Dec-05-20 | | Bachibe: whats wrong with 52. bXf6? |
|
Dec-05-20
 | | AylerKupp: <<Howard> What I fail to understand is why the computer gives a two-pawn advantage to Korchnoi even though the position at that time is an absolute dead draw !> I'm not sure what position you're referring to. If it's the position after 54...Kd5
click for larger viewThen Stockfish 12 finds the mate-in-7 via 55.Bf7+ Kc6 56.Qe6+ Kb7 (as has been mentioned, if 56...Kb5 then 57.Qc4 Ka4 58.Qa6#)) 57.Qxe7+ Ka6 58.Bc4+ b5 59.Qd6+ Kb7 60.Qc7+ Ka8 61.Qc6# (or, of course, 61.Qb8#) in less than 1 sec. If you'll post the position that you are referring to I'll run some engine analyses and see what they say. But, regardless of the starting position, engines don't just count material, they evaluate various positional factors like king safety, pieces being attacked, control of space, mobility, minor piece outposts, whether rooks have open files and bishops open diagonals, rank of a passed pawn, and many, many other factors. And clearly if it finds a forced mate for one side it doesn't matter what material advantage the other side has. That was before the advent of neural network-driven engines like AlphaZero and LeelaC0. With engines using NNUE (Stockfish 12, Komodo Dragon) the engines' evaluation function is implemented using a small neural network, only 4 layers including the input and output layers.. Otherwise Stockfish 12, for example, is a standard classic engine using iterative deepening and minimax to determine its principal variation(s). And I suspect that the same applies to Komodo Dragon. In that case, I have no idea what factors are included in the evaluation function and what weights are assigned to these factors. But I suspect that, given the ratings advantage that Stockfish+NNUE has over Stockfish w/o NNUE and Komodo Dragon over regular Komodo, that many, many more factors other than material are considered in their evaluation functions. Then again, properly evaluating fortress positions has been a problem for classic chess engine for a long time since its detection might be beyond their search horizon. It may be that chess engines using MCTS (AlphaZero, LeelaC0, and even that hybrid, Komodo MCTS) have a better chance of detecting fortress-like positions and evaluate the position properly since they don't necessarily rely on an evaluation function but conduct simulated games or their equivalent to determine the probability that the result from a given position is a win, draw, or loss. I am less sure about engines using NNUE for their evaluation function since I don't know if those NNUEs are sufficiently deep to detect fortress positions. If they don't then since Stockfish 12 is otherwise a classic chess engine which uses iterative deepening and minimax to evaluate its principal variation, it may not perform any better in this respect than if it weren't using NNUE. One thought is that since <properly trained> neural networks can be very good at pattern recognition and matching, maybe some enterprising fellow will develop and train a neural network to recognize chess fortress positions. Then, just like with tablebases but possibly assisted by GPUs like LeelaC0 to speed it up, the chess engine could probe this neural network to return a draw/not draw indicator of a detected fortress position. |
|
Dec-05-20
| | AylerKupp: <<beatgiant> Because it's simply counting material and it doesn't have a tablebase to look up the draw?> Not if it's the position after 54...Kf5. For one thing, the Syzygy tablebases that Stockfish and most of the top engines use do not have distance-to-mate information so Stockfish or any other engine would know that there is a forced mate in the position but it would not know the necessary moves or even the number of moves required. So, if you ever see any engine using the Syzygy tablebases and they report a mate-in-N, that result comes from that engine expanding their search depth sufficiently to find the mate, not from the result provided from the Syzygy tablebases. And, any rate, I disabled Stockfish's use of tablebases from the position after 54...Kf5 and Stockfish had no problem finding the mate-in-7 in less than 1 sec. Of course, if an engine knows that a position is a Syzygy tablebase win, even not necessarily a mate, it could focus on those lines to the exclusion of others and it would help the engine find the moves that lead to that mate. |
|
Dec-05-20
| | AylerKupp: In case anyone wonders or cares, after 68.Kxh6 the position is a theoretical tablebase draw.
click for larger viewI tried to use the FinalGen tablebase generator to analyze the position after 60.Kxf3 since its main criteria (kings, pawns, and at most 1 piece for each side) had then been theoretically satisfied.
click for larger viewBut FinalGen indicated that it would need 30,972 hours (about 3.5 years!) on my slow computer and require 172 TB of uncompressed disk storage, neither of which I have. Nor do I have, or ever expect to have, a UPS of sufficient capacity to ensure that the analysis could be completed if one or more power outages should happen. It kind of demonstrates the difference between "theoretical" and "practical" analysis. |
|
Dec-05-20
 | | beatgiant: <AylerKupp>
To clarify, from move 86 through move 106, the cheap stockfish engine linked on the site here gives evals of +2 to White. It drops to 0.0 after White captures the b-pawn on move 107. |
|
| Dec-05-20 | | Saniyat24: Can't Korchnoi give up his bishop on move 108, and take Karpov's last pawn, and win the game? |
|
Dec-05-20
 | | OhioChessFan: <Saniyat>, with best play, a Rook Pawn plus Bishop can't Queen if the Queening square is a different color square than the Bishop operates on. The Black King just sits in the corner, and White can't get him out. That's called a Wrong Color Bishop and is an important component of endgame play. Here's one place that discusses the matter:
https://thechessworld.com/articles/... |
|
Dec-05-20
| | beatgiant: <Saniyat24>
108. Kxa4 Kxd6 109. Kb5 Kc7 and Black's king will reach the corner in time, with a draw. |
|
| Dec-05-20 | | Saniyat24: thanks... OhioChessFan and beatgiant...! |
|
| Dec-05-20 | | RandomVisitor: Looks like white had a win at move 50:
click for larger viewStockfish_20112916_x64_modern:
NNUE evaluation using nn-62ef826d1a6d.nnue enabled <59/86 2:33:19 +5.62 50.Qg2 Qxg2 51.Rxg2 Rh1+ 52.Rh2> Rxh2+ 53.Bxh2 Kg8 54.b5 Nd7 55.Bg6 Nf8 56.Bc2 Nd7 57.Bb3+ Kf8 58.Be5 Nxe5 59.dxe5 Nf5+ 60.Kg4 Ne3+ |
|
Dec-05-20
| | HeMateMe: So korchnoi actually missed a mate in 7? I bet he'd like to have this one back...i read somewhere that either korchnoi or karpov missed a mate in one in a bad time pressure game, i think it was in benjamin's book <American grandmaster >. |
|
| Dec-06-20 | | areknames: <whats wrong with 52. bXf6?> 52...Qf4+ would follow and White's advantage is now much smaller. |
|
Dec-06-20
| | AylerKupp: <<beatgiant> To clarify, from move 86 through move 106, the cheap stockfish engine linked on the site here gives evals of +2 to White.> Thanks for the clarification. But, given that Stockfish has always been free, including the latest version, I don't know how it could get any cheaper than that! Unless, of course, they <pay> you to use a particular version. But that's unlikely to happen. It may be that there's something wrong with the set up for the "cheap" Stockfish linked on this site, but I couldn't manage to duplicate it. After 85...Ke4, given that White's bishop does not control the a-pawn's queening square, the position after 85...Ke4 is a draw if Black's king gets to the a8-b7 corner. But, it's possible for Black to lose if he makes seemingly accurate moves. For example, after 85...Ke4 86.Kd6 Kd4 87.Kc6 (going after Black's q-side pawns)
click for larger viewAccording to the Nalimov tablebases (http://www.k4it.de/index.php?topic=...) Black <loses> after the seemingly obvious 87...Kc4 to protect his Pb5. White can then play 88.Bd6 (88.Bf8 as he played in the game delays the result by one move since White's bishop <must> be on d6 in order to prevent Black's king from reaching the b8 square and the a8-b7 corner and draw). Then White can capture the Pb5 and, eventually the Pa4 while still keeping Black's king away from the a8-b7 corner. So even "simple" KBP vs. KPP endings are not so simple if the bishop of the side with the KBP does not control its pawn queening square. In a way this game reminds me of the potential position reachable in the Spassky vs Fischer, 1972 gam, which we are both very familiar with, where Black could have drawn in spite of White's extra bishop if he had been able to reach the a8-b7 corner. |
|
Dec-06-20
| | AylerKupp: <<OhioChessFan> with best play, a Rook Pawn plus Bishop can't Queen if the Queening square is a different color square than the Bishop operates on.> That's true, but <only> if the side with the lone king can be prevented from reaching the corner of the pawn's queening square. There are always exceptions. Consider for example the following position:
click for larger viewWhite will win regardless of whether it's White's or Black's move since there is no way for Black's king to prevent White's a-pawn from queening. But, of course, it gets more complicated if Black had some k-side pawns, and it that case it would depend on how far advanced the pawns were. In the following position White wins as he pleases by advancing the a-pawn but of course he can lose if he does something stupid like playing 1...Be7.
click for larger viewAnd advance the pawns one square and White can only win if he plays 1.Be1
click for larger viewAdvance the g-pawn one square and the best that White can do is draw by 1.Be1
click for larger viewAnd advance the h-pawn one square and White will lose since the Bishop can't prevent both sides from queening.
click for larger viewSo, in endgames, it's necessary to calculate, particularly if the position results in a race to queen. It would be very frustrating to follow general "rules" and find yourself in a losing position! |
|
Dec-06-20
| | OhioChessFan: <<OhioChessFan> with best play, a Rook Pawn plus Bishop can't Queen if the Queening square is a different color square than the Bishop operates on.> <AK: That's true, but <only> if the side with the lone king can be prevented from reaching the corner of the pawn's queening square. There are always exceptions. Consider for example the following position:
click for larger viewWhite will win regardless of whether it's White's or Black's move since there is no way for Black's king to prevent White's a-pawn from queening> I appreciate the clarification for the sake of the original questioner. Yes, the King must be able to get to the corner. I'll point out that you could remove the Bishop from that position and White wins with or without the move. I've never considered whether the Bishop makes any difference in Black's ability to draw. My first guess is no. |
|
| Sep-24-21 | | Chesgambit: Too many blunders by both sides |
|
| Dec-03-21 | | stridergene: The Game 6 of Carlsen Nepomniachtchi 2021 World Championship is more than 100 moves and counting. |
|
| Dec-03-21 | | stridergene: Game 6 of Carlsen Nepomniachtchi officially broke the record for the longest game in a classical world championship. The game lasted 136 moves |
|
 |
|
|
< Earlier Kibitzing · PAGE 5 OF 5 ·
Later Kibitzing> |
Kibitzing
