Potpourri: Carlsen Wins Again...
Plus: Following advice, the end of correspondence chess, coming events, and some games.
Perhaps today’s post has too few topics to really be a “potpourri”. Oh well. I’m tempted to quote Humpty Dumpty in Alice in Wonderland and say that “[w]hen I use a word, it means just what I choose it to mean — neither more nor less.” Tempted, but since I don’t hold to Humpty Dumptyism as a theory of meaning, I wouldn’t do so, except for humorous purposes. [An aside: There’s far too much Humpty Dumptyism in what passes for contemporary political discourse, and it’s a source of endless mischief. The classic source on this is George Orwell’s brilliant essay “Politics and the English Language”, which you should read immediately if not sooner if you’ve never read it before.] So let’s proceed - and I think that, with this preamble, we have enough variety to label the post a potpourri after all. [Really silly aside: perhaps a future post might describe what, if anything, Pope Francis, Pope Benedict XVI, Pope John Paul II and their predecessors had to do with our beloved game. In that case we’d have a popery potpourri. I know, I know, “don’t quit your day job.”]
Moving on from that rather self-indulgent paragraph to the world of chess, here’s some exciting news: Magnus Carlsen won a chess tournament! While that’s a dog-bites-man story (or maybe a man-bites-hamburger story, given its frequency), there was some drama in the “Freestyle GOAT Challenge” or whatever it was called. Even though Carlsen, who chose the other seven participants, took a page out of the pre-WWII World Champions’ book and did his best to avoid his toughest competition (the last two Chess960 World Champions and the most recent runner-up weren’t invited), it was still a strong field. A rapid round-robin was played to determine the seeds for the knockout stage, and Carlsen scored an unimpressive 3.5/7, losing to Nodirbek Abdusattorov (who came in first with an undefeated 5.5 points) and Dommaraju Gukesh. (Of course, 3.5/7 is a fantastic score compared to Ding Liren’s half a point out of seven, with that half-point earned in the last game after starting with six straight losses.)
When it mattered, though, Carlsen rose to the occasion. The knockout stage consisted of two-game mini-matches played with a classical time control, moving on to rapid (and faster) mini-matches if necessary to break a tie. In that first match, Carlsen needed the rapid games, as he lost the first game to Alireza Firouzja before striking back in the second game. From then on, it was all Carlsen: he went 2-0 in the tiebreaker, and then defeated Abdusattorov 1.5-0.5 in their semi-final match (winning the first game with White and drawing game 2) and defeating Fabiano Caruana 1.5-0.5 to win the final and the tournament. (The first game was drawn, and then Carlsen won with the white pieces to win the game, the match, and the tournament.)
Everyone played all three rounds of matches, which is an interesting system that can be good for players and spectator alike. A player (like Firouzja) might have the bad luck of facing off against Carlsen in the first round, but might be able to defeat everyone else. In fact, Firouzja didn’t get to play “everyone” else, but won his second and third matches to finish fifth. (He won what you might call the quarterfinals losers’ bracket.) Levon Aronian barely lost to Caruana, 5-4, with their semifinal match reaching an Armageddon game. Aronian defeated Abdusattorov to win the “semifinals losers’ bracket”, taking third. And the last-place finisher? You probably guessed it, even if you didn’t already know from following the event when it happened: Ding Liren. After going .5/7 in the rapid prelims, he lost 1.5-0.5 to Abdusattorov, 1.5-0.5 to Firouzja, and 2-0 to Vincent Keymer (the seventh-place winner, despite coming in clear second in the prelims). There’s our world champion, ladies and gentlemen, scoring three draws, ten losses, and zero wins. Oof. (For more on the event, go here and find item #2.)
I’m not on Twitter/X, so all I’m able to see is the first tweet in the series, so maybe someone would be kind enough to add further details in the comments. The paper that’s cited looks interesting, suggesting that “low ability” players (what the rating cutoff between “low” and “high” ability is isn’t given in the portion accessible from the first tweet) tend (to what degree?) to disregard “high quality” advice from stronger players. It’s well-known that humans often fail to follow good advice, especially when it’s costly in terms of time, energy, or money. But it’s a bit surprising that they would fail to do so in situations where it not only costs them nothing (and could easily profit them) but the very slight dent in their ego resulting from the acknowledgement that the other person (or engine/AI?) might know something they don’t. This is even weirder when you consider that there is something objective that we can point to: ratings.
This reminds me of a story I think I told on the Perpetual Chess podcast. When I was a (reasonably rapidly) improving youngster, some of the adults in my circle would occasionally offer a little teasing disparagement of me as a “book” player. This, my friends, is one of the weirdest put-downs ever. Chess books (and their 21st century rivals and counterparts) are generally compilations of the best games and commentaries by professional players all the way to the world champions. What kind of fool would not want to learn from them? Life is too short to only learn from one’s own mistakes. (And if you only learn from your own mistakes, your life will be very short indeed.)
[Note: This is not a rejection of trying to learn things for oneself. That has value too - and I would add that I did that as well. But if you’re interested in getting as strong as you can, and in competitive success, there’s no time to learn everything on your own. No offense, but you’re not intelligent enough, either. How do I know that? Because no one is that intelligent. Not even Bobby Fischer, Garry Kasparov, and Magnus Carlsen weren’t intelligent enough to learn everything. All had teachers and assistants, all worked very hard at the game, and all were (and in the latter two cases, are) brilliant. And yet none of them exhausted what there is to know about chess. Fischer’s chess looks simplistic at times compared to contemporary chess, and that’s not an insult. The explosion of opening theory and, more broadly, our understanding of the game is such that even as incandescent a mind as Fischer’s cannot match the combined work of thousands of chess players and, later, engines over the course of 50 years. Kasparov is still thought by many to be the greatest player of all time, and he was certainly the most rigorous analyst of all time. And yet the engines found holes all through his analysis, and for that matter later engines found plenty of holes in his earlier engine-driven analysis. As for his positional understanding of the game, it has clearly been exceeded by Carlsen. But even Carlsen makes errors and loses games from time to time, and it’s very likely that 50 years after his career ends almost no one outside of Norway will claim that he is the greatest player of all time. That’s not an insult; it’s a recognition that the game will continue to evolve.]
On to the next story. Chess engines are pretty amazing, but there are still a very few things where we’ve got them beat. I just set up a position in Stockfish where both sides have eight pawns that are blocking each other. White has only a king, while Black has a bishop, knight, and two rooks, but none of these pieces can sac itself for a white pawn. The bishop and knight can’t even offer themselves up for capture, while the rooks can. All White has to do is not capture one of the rooks if and when it offers itself, and it’s 100% drawn. Stockfish isn’t so sure, however, initially giving Black a +4 score which has now dropped to +1.33 at depth 63. If it can make it to depth 100, then presumably it will finally say 0.00 due to the 50-move rule.
Of course, we don’t need to think 100 half-moves ahead to know this: we know it instantly, and understand that an evaluation of +4 or +1.33 is nonsense here. Either Black has some way in, in which case it’s mate-in-whatever, or - as is the case - it’s 0.00. There are a number of problems like this, which are called Penrose problems, and engines don’t do well on them. Programmers have tried for decades to deal with this, without much luck, but over in AlphaZero land they’re making serious progress by letting the neural net solve it for itself. Interestingly, it’s not a unified AI, but a sort of quasi-collaborative AI model that’s doing the trick. If AI system 1 isn’t making progress, then it switches to AI system 2, etc., and “they” figure out which system works best for problem 1, which works best for problem 2, and as a team they get the job done.
When it comes to correspondence chess, the ICCF and just about everyone else except for USCF (unless they’ve finally decided to stop penalizing honest players) allows engine use. It’s tempting to think that all correspondence chess players do is leave their computers running on their positions 24 hours a day, but that’s not really fair. They have other important functions, like paying the power bill, making sure too must dust doesn’t accumulate around the back of the machine, and physically entering their opponents’ moves.
Ha ha, I kid! (A little.) While no one could compete in high level correspondence play without an engine, and it’s not exactly a coincidence that 90% (or more) of top correspondence games end in a draw, the human contribution is genuine. A strong correspondence player using an engine will dominate the caricature given above. Humans have a role to play: deciding which engines are better in which positions, and deciding where to analyze and for how long.
The latter point needs a little unpacking. Let’s say the game is on move 28. You run the engine for a while and see its candidates. You don’t just let it keep running and do nothing but eventually pick the one with the best score. You go into the variations and start exploring from there. Let’s say we’re exploring the main line, and maybe there’s another critical juncture seven moves in. We let the engine explore from that position - and likewise with other critical points deeper in this variation and in other lines.
The problem, then, is that these two functions - knowing which engine to use in what circumstance, and knowing how to fruitfully steer the engines - may be taken over by the new approach described in the article. The meta-AI takes over the human job, reducing the human to the guy or gal paying the power bill, doing the dusting, and entering the moves from the other bill-paying computer custodian. Yippee. For now, that meta-AI is only available to the AlphaZero team, but it’s just a matter of time before the rest of us have it. It will be very useful for us in our opening preparation, but it may be the death knell for correspondence chess as we know it.
[HT: Marginal Revolution for the second and third items, here and here.]
The Chessable Masters that finished a couple of weeks ago was the first leg of this year’s Champions Chess Tour. There’s still the complementary tour featuring in-person chess, the Grand Chess Tour, and the Tour has announced its full tour participants for 2024: Fabiano Caruana, Maxime Vachier-Lagrave, Wesley So, Dommaraju Gukesh, Ian Nepomniachtchi, Alireza Firouzja, Anish Giri, Rameshbabu Praggnanandhaa, and Nodirbek Abdusattorov. (“Full tour participants” as opposed to wildcards, who will also participate and vary from event to event.)
The Tour comprises five events running from May 6 (the start of the Superbet Rapid & Blitz) through August 31 (the end of the Sinquefield Cup). Three of the five tournaments are rapid & blitz events, the other two feature classical chess.
You might notice several prominent names that are missing, most obviously the guy who wins all the tournaments he plays in. It really looks as if he is committed to doing what he can to avoid classical chess (other than the possibly one-off Chess960 event mentioned above). Too bad. Classical chess will survive his absence and Ding Liren’s horrible form, but it’s still damaging to the cachet of the World Championship title.
You might also wonder why the tour isn’t getting underway until May, but it shouldn’t take too long to solve that puzzle when you reflect on the list of participants. Five of the players have something important in common, the significance of which will pass on or around April 24. (If you’re not sure what the answer is, I might offer you a slate of possibilities, and you can pick the best option from the various candidates.)
More here.
Let’s end with some games. A couple of weeks ago I presented the games from the Carlsen-Firouzja match at the end of the Chessable Masters while expressing my intention to look at more games from that rapid event. Between the play-in and all the divisions there a lot of games - 1326, to be exact. If all of you become paying subscribers I’ll annotate the whole collection. None of us wants to look at all of those games, so let’s look at a dozen entertaining games instead. (You’re still welcome to become a paying subscriber, though.) Have a look, here.
Back in 2013, Carlsen complained, "the difference [between me and Anand] is that I've been winning tournaments and he's been holding on to his title." Now, in this chess 960 tournament, it clearly may be argued that Carlsen won, at least in part, by excluding his most serious competition. Under these circumstances, Carlsen's victory is not terribly convincing.
I loved your self indulgent paragraph 👍