chess 960 showdown in St. Louis

[lkaufman]

The Rapid chess 960 RR in St. Louis, featuring most of the world's top players other than MC and challenger Nepo, is quite remarkable. On the first day, only 3 of the 15 games ended in draws, and engine analysis shows that only one of those three was really a likely draw all the way; in one of the draws Kasparov could have won immediately with a rather obvious pawn push, and in another White was probably winning early on though hard to spot. In several of the other games the advantage turned around, wins and draws becoming losses. It seems even the best players in the world make countless huge errors in Rapid when they have to think for themselves from move 1, with no playing by memory or pattern. It is obvious to me that if engines played against GMs in chess 960, the engines would get much higher ratings than they would in normal chess; I'm speaking of engines that are rated in the human range, not the top five or so. While we are still quite far from the day when any engine can win a Rapid match giving knight odds to a top ten human player in standard chess, seeing these games makes me think that perhaps Komodo Dragon could already do so in chess 960! Whether chessplayers see this as a plus or minus point for 960 is debatable; for me it's a huge plus to see humans playing without the aid of memorization, but others may feel that the poor play shows that 960 is just too difficult for humans. It is unfortunate that 960 is not a solution to the draw problem in engine chess; recall the 2 to 1 victory by Komodo over Stockfish with 47 draws early this year in TCEC 960 championship. We need a modification for it to solve that problem.

[Modern Times]

It seems even the best players in the world make countless huge errors in Rapid when they have to think for themselves from move 1, with no playing by memory or pattern.

Yes exactly, thinking from move 1 is exactly why I prefer chess960. And in computer chess, there are no arguments about choice of book - balanced, unbalanced, range of openings, and all of that. As you say though amongst the very top engines it doesn't solve the draw "problem" which isn't in fact a problem at all in my opinion, it is just reality of very high level chess and doesn't require an artificial solution like horribly unbalanced openings that some rave about. Maybe the draw rates in chess960 matches is lower amongst the mid range engines, I haven't looked at that.

[CMCanavessi]

FRC is the future, whether you like it or not.

[lkaufman]

It seems even the best players in the world make countless huge errors in Rapid when they have to think for themselves from move 1, with no playing by memory or pattern.

Yes exactly, thinking from move 1 is exactly why I prefer chess960. And in computer chess, there are no arguments about choice of book - balanced, unbalanced, range of openings, and all of that. As you say though amongst the very top engines it doesn't solve the draw "problem" which isn't in fact a problem at all in my opinion, it is just reality of very high level chess and doesn't require an artificial solution like horribly unbalanced openings that some rave about. Maybe the draw rates in chess960 matches is lower amongst the mid range engines, I haven't looked at that.

Chess960 does produce lower draw rates in engine play than normal chess with normal opening books, but higher than normal chess with highly unbalanced opening books. For me the ideal would be to simply forbid repetition of position in chess960, which would substantially lower the draw rate even between top engines. As long as we're changing the start position and castling rules, why not make this natural improvement as is used in GO and with some other conditions as to which side needs to avoid the rep in Chinese chess?

[lkaufman]

FRC is the future, whether you like it or not.

On the second day five of fifteen games were drawn, but in three out of these five games one side was pretty clearly winning at some point, so only two out of fifteen were likely drawn from start to finish. Pretty favorable sign for FRC, only ten percent of thirty games looked drawn all the way.

[MikeB]

It seems even the best players in the world make countless huge errors in Rapid when they have to think for themselves from move 1, with no playing by memory or pattern.

Indeed, that is the beauty of Chess960. RJF was on to something way before most others, if not, all others.

[kranium]

[pgn][Event "Champions Showdown 9LX"]
[Site "Saint Louis USA"]
[Date "2021.09.09"]
[Round "1.2"]
[White "So, Wesley"]
[Black "Vachier-Lagrave, Maxime"]
[Result "1-0"]
[EventDate "2021.09.08"]
[Variant "Chess960"]
[SetUp "1"]
[FEN "qnrbbnkr/pppppppp/8/8/8/8/PPPPPPPP/QNRBBNKR w HChc - 0 1"]

1. b4 d6 2. c4 e5 3. d4 exd4 4. Qxd4 Nc6 5. Qb2 Ne6 6. Ne3 Bg5 7. Nd5 Bxc1
8. Qxc1 O-O 9. b5 Ncd8 10. Nf6+ gxf6 11. Qh6 f5 12. Qf6 Ng7 13. Bc3 Nde6
14. h4 Bxb5 15. h5 Rce8 16. h6 Qd8 17. hxg7 Qxf6 18. gxf8=Q+ Nxf8 19. Bxf6
Bxc4 20. Nc3 b5 21. Rh4 Re6 22. Bd8 c5 23. Bb3 Bxb3 24. axb3 b4 25. Nd5 Ng6
26. Rh3 Rxe2 27. Re3 Rd2 28. Re8+ Kg7 29. Bf6+ Kh6 30. Ne3 f4 31. Nc4 Rd3
32. Ra8 f3 33. g4 Nf4 34. Rg8 Ne6 35. Ne3 1-0[/pgn]

10. Nf6+ ! wins quickly, and is being called a brilliancy
white's bishops will likely come to c3 and c2, exerting tremendous pressure on the black king

seems logical though considering blacks cramped position, backrank pieces, and completely out-of-play queen
9. ...Ncd8 seem to be a big mistake

[kranium]

BTW - Youtube has complete live coverage

2021 Champions Showdown | Chess 9LX
https://www.youtube.com/c/STLChessClub

[lkaufman]

Today, the final day, six of the fifteen games were drawn, but only one of the six looked like it should have been drawn thruout according to engines; in all the other five one side had a substantial, probably winning, advantage, at some stage, sometimes even in the final position. So only 4 out of 45 games were never looking like probable wins at some point. I noticed that White's opening advantage varies quite a bit from position to position; this means that for engine vs. engine play, if we have too many draws in chess960, all we need do is prune out the more balanced positions, and the draw percentage will go down while it will still be fair if each engine plays White once from each selected position. I doubt that any of the 960 positions is actually won for White, but I'll bet that the 100 most unbalanced ones would produce a pretty decent share of wins even with the two best engines playing at long time controls on big hardware. So maybe for engine play (or correspondence), we'll have "Chess100" (or some other number in that ballpark). Is there a list somewhere of all 960 positions evaluated by a strong engine (I seem to recall seeing one once)?

[CMCanavessi]

Today, the final day, six of the fifteen games were drawn, but only one of the six looked like it should have been drawn thruout according to engines; in all the other five one side had a substantial, probably winning, advantage, at some stage, sometimes even in the final position. So only 4 out of 45 games were never looking like probable wins at some point. I noticed that White's opening advantage varies quite a bit from position to position; this means that for engine vs. engine play, if we have too many draws in chess960, all we need do is prune out the more balanced positions, and the draw percentage will go down while it will still be fair if each engine plays White once from each selected position. I doubt that any of the 960 positions is actually won for White, but I'll bet that the 100 most unbalanced ones would produce a pretty decent share of wins even with the two best engines playing at long time controls on big hardware. So maybe for engine play (or correspondence), we'll have "Chess100" (or some other number in that ballpark). Is there a list somewhere of all 960 positions evaluated by a strong engine (I seem to recall seeing one once)?

There's something here: https://www.chess.com/article/view/what ... 0-position