Magnus Carlsen wins longest world chess championship game of all time
chess.com
chess.com
I prefer to watch analysis of games from YouTubers like Agadmator or Jerry from Chessnetwork and discover who the winner was by watching the game.
I wonder if these long games are not particularly helpful. Most chess final games end in draws as players are reluctant to create an imbalance and instead play it safe. There tends to be more exciting games played in games with a shorter time limit.
I'm not advocating bullet but perhaps they could try 30 minute matches instead?
I get your sentiment but this is really exciting news in the chess community. A win after so many years!!
Unfortunately, they're probably optimizing for views, and Magnus has a lot of name recognition.
For few of previous matches by Magnus, she found moves that even Magnus had missed.
Skip to 12:48 for the start of the commentary, 32:15 for the start of the game.
Although Lichess isn't perfect and it's kinda annoying. One of the bad things about Lichess is how unrealistically inflated their ratings are; 2100 on Lichess is equivalent to about 1800 on Chess.com. Chess.com is much more similar to the FIDE rating and other real-world systems. Lichess is just super inflated for some reason and it makes it kinda hard to compare your progress to other metrics.
There's also https://chessgoals.com/rating-comparison if you prefer a table
- they use different rating systems (Elo for FIDE, Glicko-2 for LiChess and afair, Glicko/Glicko-2 for chess.com)
- ELO / Glicko rating is calculated for a _player pool_. Which means it will never correspond between FIDE-registered players and chess websites because they don't have the same player base.
It has nothing to do with being “inflated” or “unrealistic”. The difference is “by design”
150 Elo on levels below 2700 FIDE is almost night and day difference.
Chess.com ratings show nothing about theoretical FIDE ratings. Comparison tables have been built with data mining and lucky guesses, there's no quality difference between Chess.com <-> FIDE and LiChess <-> FIDE.
It's just a fact that chess.com ratings are closer to 'real' ratings than Lichess are. It's not perfect but that doesn't mean it's a zero-information statement either.
Of course these things are correlated, but e.g. there have been times in the world championship when one of the players thought about a position for 20+ minutes before making a move. That's hard to do, and requires very strong visualization skills.
Lichess is overall better in my opinion, better & slicker product, free analysis, opening explorer etc. Chess.com only leads in video content imo.
With rating systems, the absolute scale doesn't matter, all that matters is the predictive power of the system. Every time I have benchmarked, glicko does a better job in predicting win/loss/draw probabilities than ELO ratings used by FIDE.
If you want to know your rating relative to other players on the site, you can check the distributions here [2]. If you want to compare them to FIDE ELO, or chess.com, you can use this site although I am not sure about exactly what methodology they use [3].
I wouldn't agree that lichess ratings are inflated or unrealistic relative to FIDE Elo, or any other rating system any more than kilometers are unrealistic and inflated relative to miles. They are simply different measurement systems.
[1] http://glicko.net/glicko/glicko2.pdf
I wonder if there’s some way other than time constraints to encourage higher variance moves.
It would have been 4 years if not for COVID.
And what it basically means is that there were decisive games 4 years ago, and then 2 years later (the World Chess Championship is played every 2 years), in the 2018 match between Caruana and Magnus there were no decisive games.
So essentially there was 1 match with all draws during the standard chess part of the tournament. And that match was between Fabi who is regarded as one of the players with some of the greatest prep and play in standard chess, and Magnus, who is really one of the greatest players, if not the greatest, of all time.
I love Jerry, and I do watch him occasionally. But https://www.youtube.com/watch?v=EhJl3BKA-n8&ab_channel=Gotha... is well done.
(If you do want to watch Gotham for this particular match, https://www.youtube.com/watch?v=VoY5RSCi5pE&ab_channel=Gotha... starts the series, I think.)
He also comments frequently, and keeps his analysis understandable to my amateurish level while having better insight than Levy. Unfortunately his comment for the world championship is in the Russian channel (although he's from the US).
Great Youtube reviews and explanation of strategy & gameplay.
Other time controls is already a thing. Personally a huge rapid fan. I think Fisher was right. The problem is just prep and theory. If we still have classical time but chess960 so there isn't 20 moves of theory memorized.
I'm of the opinion that there's nothing to gain from fans having engines open during a game, despite working in sports analytics and ostensibly being on the side of the machines. Obviously engines have a lot to say about the modern game, and the match so far had clearly been dominated by excellent prep by both teams. But the excellent commentary on Levitov Chess World by the peerless Peter Svidler, Vladimir Kramnik and Evgenij Miroshnichenko was firmly of the opinion that we had a game on our hands throughout. They were correct.
I looked into some stats from the most recent TCEC superfinal just to get a feel for the error bars in engine evaluations. I think it's clear that even up to about +1.5, a draw is still on the cards between the superhuman engines, let alone humans. About a quarter of the superfinal games were decisive, despite us treating them as oracles. They're far more accurate in their draw predictions - in games that reached a 0.0 evaluation, only 3 ended with a win.
But in this game, even when we hit the tablebase results, where the result is a foregone conclusion for the machines, they still don't reflect the outcome of a human game, especially when the time controls start to get more stressful.
Anyway, what a wonderful, reassuring game. The players will be exhausted and it sets up a fascinating weekend of play. I couldn't be happier.
Isn't this game a great argument for engines. Using a threshold of 2 to be well above your 1.5 this game had a total of 4 moments where one of the players had a very good chance for a win, and one of those moments was for Nepo:
Having the engine show you that is a great complement to hearing the commentary from top super GMs. Why would you throw away that part of the coverage?
>But in this game, even when we hit the tablebase results, where the result is a foregone conclusion for the machines, they still don't reflect the outcome of a human game, especially when the time controls start to get more stressful.
While true I think the solution is to double down and extract metrics from the engine search that reflect that. Not only the evaluation of the position but also how narrow the path is for white and black, to better understand how even if it's a draw from the machine's point of view, how hard of a draw it will be for the human to hold.
So yes, I agree with your second point, the engines/metrics/stats could be doing more here. Just relying on the evaluation at ridiculous depths isn’t very interesting, and I don’t believe those scores necessarily translate to a very good chance to win. 33 Rcc2 would have been very hard to calculate and involves a double pawn sacrifice. 36… Bxb4 probably was the move most humans would have made but I’m still not convinced Nepo always wins from there. I’m sure there were other swings I’m forgetting. But outside of “this line exists at depth 30 if you see it” there’s a lot to learn about the components of that evaluation, like mobility and king safety etc that do mirror some of the human thought process and which you can apply yourself. I’d love to see more detailed metrics like that, especially if they could capture why the humans felt it was a very good Catalan for white with lots of options, and not just zeroes. But as of today, I’d rather hear what Kramnik thinks than Stockfish. And the difficulty of holding a position is far more accurate coming from humans than the machines (game 5 for example). Even when the computer thinks there’s only one move in a position, that’s often based on the opponent playing perfectly to a great depth afterwards.
I don’t want to sound too much of a curmudgeon here, I like engines, I write my own and of course I have Sesse and my own copy of Leela open most games. But I think too often they short circuit one’s brain, and I’ve found my enjoyment increases when I follow the human thought process and not the machine one. I’d love to see a richer synthesis of the two though!
My jaw dropped when Magnus ignored his rook being attacked at the end in favor of the potential fork with his knight. It is ultimately what led to the sure win.
It is likely that this is the most evenly matched world championship of the 21st century, as both players are extremely well-prepared to face each other at the highest level.
The rapid tie-breaker was a different story however.
Where/which move did Ian make the mistake?
Anyways, hats off to everyone involved. Magnus, Ian, commenters, and the chess twitterati. What a game! ~8hrs!!!
Q. What was the decisive moment?
“I don’t know,” he [Carlsen] says. “But it felt like at the end when I got [133. e6] and maybe there was still some miracle defense there, but it didn’t really feel that way. At that point I felt very, very good about my chances.”'
https://syzygy-tables.info/?fen=4k3/8/8/4PR2/5P2/6NK/q7/8%20...
>"From May to August 2018 Bojun Guo generated 7-piece tables. The 7-piece tablebase contains 423,836,835,667,331 unique legal positions in about 18 Terabytes."
I wouldn't say that Ian made a mistake. That position was winning for white after pawn on h4 was traded. It's not Leela vs Stockfish, it's two humans playing. Defending with solo queen against RNPP w/ connected pawns is extremely hard unless perpetual check is unstoppable.
Maybe that would be a fun concept. A game where the players would start with 1 year on the clock?
If not, maybe there is some insight to be gained to raise high level play even more.
It's unknown what would happen if a top human with serious preparation played against a modern engine. They'd have basically no chance of winning and a fairly low chance of a draw, with the unknown being "how low is fairly low". My wild-ass guess is maybe 1 game in 10 when the human has the white pieces and much lower with the black pieces. There is not much chess theory about what happens when a human plays a game from the beginning aiming for a draw, even though they do that all the time in real chess. It is something of a gap in the "literature".
This is one of the intuitive reasons for hypothesizing that chess is drawn with perfect play, even though we can’t prove it.
(I gather that the number of draws may have been lower at the time, but they showed few or none.)
But titling "unlike the last time (and only the last time) there has been a win and it's also the longest game ever" would have been less clickbaity.
I love both playing and watching rapid chess, but I think it's better for the world championship to be decided using classical time limits. That way, we get more interesting games, rather than the indignity of top players blundering while under time pressure.
I think I agree with you. Chess is only a thing because people have played chess for a long time. Different time limits are a different kind of game -- neither better nor worse, but different from the things that made tournament-style chess popular.
Perhaps with computers it's time to retire the existing notion of chess tournaments. But people still seem interested, much as we still watch Usain Bolt despite him being outclassed by cars and even bicycles.
A machine can throw a fastball at speeds well above what humans can do. That doesn't mean we retire pitchers completely and end the game of baseball. A car can cover 100m in a fraction of the time humans can, doesn't mean we end track and field.
Chess is a game between humans and that's what makes it exciting.
And with computers we have a whole additional layer added where we can see what an essentially perfect game would look like and compare to that (although computers can still occasionally be surprised by human moves).
You clearly don't follow chess, given your question. So I have to ask: why are you so quick to take a stance on something you don't follow or understand?
[0]: https://www.chess.com/article/view/the-draw-rule-is-classica...
Classical games: Ian Nepomniachtchi beat Magnus Carlsen 4 to 2, with 14 draws.
[0]: https://www.chessgames.com/perl/chess.pl?page=1&player=carls....
Look instead at the next highest level tournament, the WCC candidates. There's money on the line, but really, the chance for the title is so much more important, winning the tournament is all that matters, and you aren't getting there with just draws, or even just a couple of wins. The very top players in the world, other than the current WCC are the only ones that participate, and yet the draw rate is lower than that average 75% too: In the last tournament, Nepo lost 1 and won 5 out of 14 games. Same with Caruana in the 2018 candidates.
That said, we'd see fewer draws too with shorter time controls, but then the top players also change. In 2018, if the championship was an infinite classic series until someone has 3 wins, Caruana was probably the favorite, given how he had Carlsen against the ropes over and over again, just a bit shy of victory. But in Rapid or Blitz, He's nowhere near as good.
> Known for always being quite fit, the Norwegian GM admitted this was a tiring affair: "Sure, but I think that's the way it is. As I said, it shouldn't be easy in a world championship match. You have to try for every chance no matter how small it is. Part of it was by design at some point that I thought I should make the game as long as possible so that we would both be as tired as possible when the critical moment came. That turned out to be a good strategy."
Turned the game into a sport. Which ended up paying off better than the counter-strategy of playing quick moves to apply extra time pressure. From the description it seemed like it could easily have gone either way, so clearly a win and yet by very narrow margins where replaying, if possible to go back in time, would likely yield different results.
CORRECTION (see below): this was game 6, not game 5 (I had lost count), so current score is 3.5 to 2.5, and there are potentially 8 classical games left.
Now he will likely have to change tactics and enter into more aggressive and risky ideas. It should increase the odds of a decisive result one way or the other going forward, regardless of whether Magnus is "playing for a draw" or not.
Tough luck for Ian though. It’ll be hard not to tilt after this but he’s a champion.
In fast games (bullet, blitz), anything goes and it's much more common to see games end in a checkmate - or very close to one - as players will often play to the bitter end to maximize the chance of their opponent making a mistake under time pressure, or running out of time completely.
In reply black can only try to give endless checks with his queen, but the white king can hide behind the Knight and they will end.
Can we edit the title to say "Magnus Carlsen and Ian Nepomnitchi play the longest world chess championship game"?
This is the game mode that has been decided and agreed upon by all parties.