|
< Earlier Kibitzing · PAGE 5 OF 6 ·
Later Kibitzing> |
Jan-23-19
 | | keypusher: <ChessHigherCat: SF says +0.69 but I'm not sure what that means because theoretically +1.00 = one pawn up, which is generally a draw, so +0.77 would mean that white doesn't even have the full pawn advantage. Any SF pundits out there?> I'm not an SF pundit, but I've spent a lot of time lately looking at SF10 evaluations of AlphaZero-SF games, and yes I'm aware that's problematical. SF has an evaluation function that includes material, but in my experience one pawn ahead only rarely seems get you a +1 in evaluation score -- often the eval is lower, sometimes it's higher. The only thing you can say with confidence about a +0.69 evaluation in an endgame is that the engine hasn't found either a clear winning procedure or a clear drawing procedure within whatever search depth it's using. If you're running a reasonably high-ply search in a simplified ending my sense is that the engine is doing an enormous amount of calculation trying to get to an end state that's clearly won or clearly drawn. It's not just giving white a point for the pawn and then going from there. |
|
| Jan-23-19 | | ChessHigherCat: <keypusher> Thanks a lot! |
|
| Jan-23-19 | | SChesshevsky: I think that this was more of a brutal game for Anand and really nice win for Carlsen than computer evals would show. Seems Carlsen came out nicely with 5. Nc3 throwing it to Anand as to what kind of Four Knights, Ruy or Italian structures he might want to try to get and forcing him to spend some time in the tank. Apparently after going for a kind of decent Italian with 5...Bc4, Carlsen presents something like a Deferred Exchange Ruy. Usually pretty solid with a continous slight edge and probably lines that Anand hasn't seen that often and will have to think about. I thought that Anand's pass after 9. Be3 was telling and suggested that Anand may have thought he was caught in some prep which may have been true. The combination of possibly relative unfamiliar territory, possibly being caught in prep, and Carlsen pushing and pushing probably doesn't show up in the computer evals much but it's huge pressure over the board. By move 30 Anand's on the defensive by 40 he's a pawn down. Not a great pawn and Black has some compensation, so it's not much on the eval probably, and though Carlsen has to be careful, Anand likely has to be more careful. By move 54 with rooks off, now it's seemingly Carlsen with only winning chances though arguably slight on the board. But Carlsen's got three big advantages: Anand's probably been thinking much harder over the last 50 moves or so, Anand knows one bad slip is probably fatal, and Magnus has ultimate confidence in his own endgame technique. Again, things that might not show up in evals but dominant over the board. Best thing about this game I thought was the video at the end. Place has been cleared out for who knows how long with only Carlsen and Anand left playing and post game with Anand and Carlsen doing a little discussion at the board it appears even an arbiter wants to give them a closing time bum's rush so that he can get out of there too. Maybe a brutal game for everyone except for the victor. |
|
| Jan-23-19 | | ajile: I was looking at 76..b3 but then the instant 77.Nd4+ ruins Black instantly.
click for larger view |
|
| Jan-23-19 | | Eyal: 
click for larger viewSeveral people already mentioned that 70...b5? (instead of bxa5!) was the losing move, but it should be noted that 70...bxa5 works for a <very> concrete reason, which requires precise calculation: 71.Kd5 Nf4+! (only move) 72.Kxc5 Ne2! (only move) 73.Nd6(!) Nc1! (only move - after 73...Nxc3? 74.Nb5+ Nxb5 75.Kxb5 the pawn endgame is lost) - and now the only way for White to preserve the b pawn is 74.Nb5+ Kb7 75.Nd4 - but this allows Black to defend his remaining pawn with 75...Ka6. That's a good example of how Carlsen wins "drawn" positions by posing for his opponents problems they find hard to solve, especially in the 6th or 7th hour of play. |
|
| Jan-23-19 | | Jaburu: I think that 52...Re4 maintaining the rook, is draw. |
|
Jan-23-19
| | keypusher: <Eyal> Thanks. <(only move - after 73...Nxc3? 74.Nb5+ Nxb5 75.Kxb5 the pawn endgame is lost)> When I was looking at it myself, I wondered if Anand saw that 73...Nxc3 didn't work and that's why he rejected 70....bxa5. If I understood sesse correctly, in the game continuation 72.Ke5, 73.Nd5+, 74.b4, 75.Nxc3, 76.Ne2 were all "only" moves. Some of those are obvious, but some aren't, 74.b4 for example. Tough game. |
|
| Jan-23-19 | | ChessHigherCat: I appreciated the <Eyaluation> too. 70...? is Sunday puzzle material if you calculate it out to the draw. |
|
| Jan-24-19 | | dhotts: Carlsen seems to be a clairvoyant when it comes to end games with few pieces riddled with a multitude of subtly...I think he figured out a few years ago that these positions presented "real time" thinkers with perplexing problems i.e. one poor move results in a loss 10-15 moves later...With many pieces on a chess board, seeing 10-15 moves ahead is not that difficult for a GM, but in the simplicity of end games 10-15 moves is a hazy environment, except for the clairvoyant Carlsen, the Mozart of Chess. |
|
| Jan-24-19 | | raju17: A push to the pawn to b3 wins for black as white can not stop pawn advance. Else black can devour the white pawn. It takes 2 moves for white to reach Nb3 but black can cross this point directly by playing b3 |
|
| Jan-24-19 | | raju17: 76...b3 77. Nc3 b2 78. Kd4 b4 79. Nb1 Kb5 |
|
| Jan-24-19 | | rogge: 76...b3 77. Nd4+, and b3 is gone. |
|
| Jan-24-19 | | MrMelad: Impressive |
|
| Jan-24-19 | | Ulhumbrus: If Anand missed the way to draw after 70...bxa4 and one is going to speculate on the reasons why, one may guess more than one possible explanation. Three examples of guesses are (1)fatigue after several hours of play (2) an oversight which 70...b4 was based on, and this could have been caused partly by the first reason and (3) Anand needs more practice at seeing or recognizing quickly the resources employed by Black in arriving at the draw after 70...axb4. Anand's own view can be assumed the reliable one. |
|
| Jan-24-19 | | paavoh: A very enjoyable game with high caliber maneuvering. |
|
| Jan-24-19 | | starry2013: If Caruana was hoping he'd at least temporarily gain rating top spot in this period it looks like he'll be disappointed. |
|
| Jan-24-19 | | The Kings Domain: Shades of 2013-2014. Reminds me of Morozevich's comment on the Norwegian's style of play: "He doesn't play the best move, he plays the right move". |
|
| Jan-24-19 | | ChessHigherCat: That reminds of me the reporter who asks Capablanca's opponent how many moves he can see ahead: "11. I can see 11 moves ahead". Then he asks Capa the same question: "Just one, but always the right one." |
|
| Jan-24-19 | | John Abraham: Carlsen's domination of Anand is getting to Kasparovian levels. |
|
| Jan-24-19 | | BOSTER: <Eayl>.Another Q. Why did Anand move his knight far from his king playing move 67... Ng6+ not Nc6+? |
|
| Jan-24-19 | | BOSTER: Supposed to be <Eyal>. |
|
| Jan-25-19 | | Eyal: <Why did Anand move his knight far from his king playing move 67... Ng6+ not Nc6+?> 67...Nc6+ takes away freedom of maneuvering from the knight - White should be winning after 68.Ke6, e.g. 68...Nd8+ (moving with the king allows White Kd6) 69.Kd5 and now a5 is coming next move, whatever Black does. Let's say 69...Nb7 70.a5! bxa5 71.Nb2! followed by Na4 (prepared by c4 if necessary) and capture on c5 with a winning position. So 67...Ng6+ was an only move (allowing Black to defend against 68.Kd5 with Nf4+) - another illustration that at least by this stage, the ending is not so easy for Black to defend. |
|
| Jan-25-19 | | BOSTER: Thanks <Eyal>.But computer evaluation after 67...Nc6+ is equal. |
|
| Jan-25-19 | | Eyal: <BOSTER> Not really. I mean, when you say "computer evaluation" it always depends on what computer and which ply-depth of analysis - perhaps you looked at some relatively shallow evaluation, but if you let a decent engine run long (that is, deep) enough it will show you the win. |
|
Jan-25-19
| | keypusher: <BOSTER>
If you run SF on this website for 15 minutes the eval after 67....Nc6+ is +5.11 (54 ply). I don't pretend to understand all the moves, but it follows the variation <Eyal> gave, and White captures the c-pawn and eventually the a-pawn with a simple win. |
|
 |
|
|
< Earlier Kibitzing · PAGE 5 OF 6 ·
Later Kibitzing> |
Kibitzing
icon.