Armageddon Openings released

[lkaufman]

It's quite normal to sac a pawn to prevent castling, and many respected gambits seem to offer less compensation than NBC. But I'm rather expecting a close result this time. Maybe engines are better at defense (relative to other engines) than humans are (relative to other humans). What's fair between engines may not be fair between humans.

Looking forward to the results...
Of course, Armageddon openings were designed for cumpterchess = engines. For humans it seems much better to me, to use different thinking times for Armageddon, like it is done already today. Human playing strength is more affected by a reduced thinking-time, than engine playing strenth is.

After 80 games NBC has 49 White wins, 31 draws (counting as Black wins). This seems too one-sided to me, almost surely winning for White with correct play, in line with my subjective opinion. I'll add more games to the NSBC version. If Black stays far ahead with that one, I don't know what would be in between and sensible.

[lkaufman]

It's quite normal to sac a pawn to prevent castling, and many respected gambits seem to offer less compensation than NBC. But I'm rather expecting a close result this time. Maybe engines are better at defense (relative to other engines) than humans are (relative to other humans). What's fair between engines may not be fair between humans.

Looking forward to the results...
Of course, Armageddon openings were designed for cumpterchess = engines. For humans it seems much better to me, to use different thinking times for Armageddon, like it is done already today. Human playing strength is more affected by a reduced thinking-time, than engine playing strenth is.

After 80 games NBC has 49 White wins, 31 draws (counting as Black wins). This seems too one-sided to me, almost surely winning for White with correct play, in line with my subjective opinion. I'll add more games to the NSBC version. If Black stays far ahead with that one, I don't know what would be in between and sensible.

Well, after 80 games with the NSBC version White won 34, and Black drew (counting as wins) 46. So most likely NSBC isn't quite winning for White while NBC is, but this is far from certain. Since NSBC was closer to even than NBC, and since it is closer to normal chess (Black can still castle queenside, and it is closer to normal in terms of White's advantage) I think that NSBC is the best Armageddon option based on the evidence so far. Also I could imagine GM opinion being divided as to which side has the better chances in human top level chess with NSBC, whereas I'm pretty sure that almost all GMs would pick White with NBC. Being limited to castling long is a severe restriction, but not nearly as severe as no castling rights at all.

[pohl4711]

From what I have seen, I believe, that the difference of results NBC and NSBC should be really small. Otherwise, not enough games were played (80 games mean nothing):
If I counted right, for example, in my 500 SALC_Armageddon_IM_4moves testgames with long thinking time, black castled only 116 times long and in the SALC_Armageddon_IM_4moves testgames with short thinking time, black castled only 110 times long. So only less than 25% of all games black castled and in more than 75% black did not castle. So in that SALC-games less than 25% were NSBC and more than 75% were NBC.
So, if you play NSBC-positions in testgames, most of the time the games will played as NBC, because the engine dont castle long with black.
So, the results should be very similar to NBC, otherwise the testing conditions (number of games) are not good.

[lkaufman]

From what I have seen, I believe, that the difference of results NBC and NSBC should be really small. Otherwise, not enough games were played (80 games mean nothing):
If I counted right, for example, in my 500 SALC_Armageddon_IM_4moves testgames with long thinking time, black castled only 116 times long and in the SALC_Armageddon_IM_4moves testgames with short thinking time, black castled only 110 times long. So only less than 25% of all games black castled and in more than 75% black did not castle. So in that SALC-games less than 25% were NSBC and more than 75% were NBC.
So, if you play NSBC-positions in testgames, most of the time the games will played as NBC, because the engine dont castle long with black.
So, the results should be very similar to NBC, otherwise the testing conditions (number of games) are not good.

If I remember my statistics correctly, standard deviation is square root of npq which for 80 games is square root of 20 or about 4.5. So a 49 to 31 result is right at two standard deviations plus for White, and a 34 to 46 score is 1 1/3 sigma for Black. So NBC is "proven" to favor White by standard statistical measure, while NSBC probably favors Black but more games are needed to be confident of this, it may well be about even. You say Black castles long in less than 25% of the games, but nearly a quarter of the games is a very significant portion, surely enough to swing the win percentage by several percentage points. Probably the difference in the win percentage from 61.25% to 42.5% is too large to be due solely to the difference between NBC and NSBC, but perhaps a third or a half of the difference is real, and the rest due to sample error. That seems most likely to me. I agree that the difference is not huge, but it is far from trivial.

[lkaufman]

From what I have seen, I believe, that the difference of results NBC and NSBC should be really small. Otherwise, not enough games were played (80 games mean nothing):
If I counted right, for example, in my 500 SALC_Armageddon_IM_4moves testgames with long thinking time, black castled only 116 times long and in the SALC_Armageddon_IM_4moves testgames with short thinking time, black castled only 110 times long. So only less than 25% of all games black castled and in more than 75% black did not castle. So in that SALC-games less than 25% were NSBC and more than 75% were NBC.
So, if you play NSBC-positions in testgames, most of the time the games will played as NBC, because the engine dont castle long with black.
So, the results should be very similar to NBC, otherwise the testing conditions (number of games) are not good.

I added another 20 games to the NSBC test to make the planned 100 games, and the result of the last 20 was 7 wins for White, 12 draws (counting as Black wins), and one actual Black win (the first in 180 games of these NBC/NSBC tests). So the NSBC total is 41 to 59, which is 1.8 sigma in Black's favor. So it's almost proven that NSBC favors Black and NBC favors White. Which is more fair is pretty much a tossup at this point.

[leavenfish]

I had a good laugh today when it occurred to me that it is Larry Kaufman who is responsible for the (mythical) disease he seeks to cure - draws in chess!

Let me explain.

I was watching today's Sinquefield Cup coverage and at a moment when Anand could apparently have unleashed 19 Rd1 Qb6, 20 Rd7! (about 18 min in to the coverage found on the Chessbase website) with some tactical fireworks in the air. After Maruice and his magical engine tooled around a bit, the opinion was that Anand would probably not go in that direction....and such was the case.

The reason why was tied into a controversial comment of Nakamura's from a while back that players of the 21st century are not as tactically well versed as those of the 20th century because engines have 'taught' the players that there area almost always ''defensive resources"...so often when they get an edge they just play the most positionally harmonious moves instead of the most incisive. (I quibble with this...but get their and Naka's point)

So, there you have it: Larry, his Komdo and the rest...are the true culprits!

[lkaufman]

I had a good laugh today when it occurred to me that it is Larry Kaufman who is responsible for the (mythical) disease he seeks to cure - draws in chess!

Let me explain.

I was watching today's Sinquefield Cup coverage and at a moment when Anand could apparently have unleashed 19 Rd1 Qb6, 20 Rd7! (about 18 min in to the coverage found on the Chessbase website) with some tactical fireworks in the air. After Maruice and his magical engine tooled around a bit, the opinion was that Anand would probably not go in that direction....and such was the case.

The reason why was tied into a controversial comment of Nakamura's from a while back that players of the 21st century are not as tactically well versed as those of the 20th century because engines have 'taught' the players that there area almost always ''defensive resources"...so often when they get an edge they just play the most positionally harmonious moves instead of the most incisive. (I quibble with this...but get their and Naka's point)

So, there you have it: Larry, his Komdo and the rest...are the true culprits!

Engines certainly have a lot to do with the draws, for example they have pretty much shown how Black can draw with the traditional lines of the Berlin or the Marshall or various defenses to the Queen's Gambit. Of course this would be true even if I never learned chess, there would still be Stockfish etc. The fact that so much of the games between top GMs is computer preparation is even more of a problem than the draw percentage. Chess 960 does go a long way to solving both problems, but whether it will win broad acceptance remains to be seen. If not, other ideas are needed. Let's see what the reaction is to the Caruana-Kasparov 960 match (and the others), followed shortly afterward by the first FIDE World 960 championship.

[leavenfish]

I had a good laugh today when it occurred to me that it is Larry Kaufman who is responsible for the (mythical) disease he seeks to cure - draws in chess!

Let me explain.

I was watching today's Sinquefield Cup coverage and at a moment when Anand could apparently have unleashed 19 Rd1 Qb6, 20 Rd7! (about 18 min in to the coverage found on the Chessbase website) with some tactical fireworks in the air. After Maruice and his magical engine tooled around a bit, the opinion was that Anand would probably not go in that direction....and such was the case.

The reason why was tied into a controversial comment of Nakamura's from a while back that players of the 21st century are not as tactically well versed as those of the 20th century because engines have 'taught' the players that there area almost always ''defensive resources"...so often when they get an edge they just play the most positionally harmonious moves instead of the most incisive. (I quibble with this...but get their and Naka's point)

So, there you have it: Larry, his Komdo and the rest...are the true culprits!

Engines certainly have a lot to do with the draws, for example they have pretty much shown how Black can draw with the traditional lines of the Berlin or the Marshall or various defenses to the Queen's Gambit. Of course this would be true even if I never learned chess, there would still be Stockfish etc. The fact that so much of the games between top GMs is computer preparation is even more of a problem than the draw percentage. Chess 960 does go a long way to solving both problems, but whether it will win broad acceptance remains to be seen. If not, other ideas are needed. Let's see what the reaction is to the Caruana-Kasparov 960 match (and the others), followed shortly afterward by the first FIDE World 960 championship.

So few people really care about 960...it's not 'chess' and chess itself is hard enough to master. It really seems that the more patterns that one knows, the more they appreciate a game of chess. 960...takes even that away from them. Certainly the two should never mingle.

I was just thinking about all of Reshevsky's games that degenerated into time pressure - and all sorts of other players as well. Then I was thinking about the G/130 with a 30 second delay we see at the Sinquefield Cup. That is really pretty luxurious given the depth of theory which is so often played. Classical chess is great...but really it is too darn slow for the top level where they prep (with the help of engines) their games against largely the same old faces every tournament. G/120 (G/90 would be good) with a 10 or 15 second delay is more than enough time for players of this caliber.

[lkaufman]

I had a good laugh today when it occurred to me that it is Larry Kaufman who is responsible for the (mythical) disease he seeks to cure - draws in chess!

Let me explain.

I was watching today's Sinquefield Cup coverage and at a moment when Anand could apparently have unleashed 19 Rd1 Qb6, 20 Rd7! (about 18 min in to the coverage found on the Chessbase website) with some tactical fireworks in the air. After Maruice and his magical engine tooled around a bit, the opinion was that Anand would probably not go in that direction....and such was the case.

The reason why was tied into a controversial comment of Nakamura's from a while back that players of the 21st century are not as tactically well versed as those of the 20th century because engines have 'taught' the players that there area almost always ''defensive resources"...so often when they get an edge they just play the most positionally harmonious moves instead of the most incisive. (I quibble with this...but get their and Naka's point)

So, there you have it: Larry, his Komdo and the rest...are the true culprits!

Engines certainly have a lot to do with the draws, for example they have pretty much shown how Black can draw with the traditional lines of the Berlin or the Marshall or various defenses to the Queen's Gambit. Of course this would be true even if I never learned chess, there would still be Stockfish etc. The fact that so much of the games between top GMs is computer preparation is even more of a problem than the draw percentage. Chess 960 does go a long way to solving both problems, but whether it will win broad acceptance remains to be seen. If not, other ideas are needed. Let's see what the reaction is to the Caruana-Kasparov 960 match (and the others), followed shortly afterward by the first FIDE World 960 championship.

So few people really care about 960...it's not 'chess' and chess itself is hard enough to master. It really seems that the more patterns that one knows, the more they appreciate a game of chess. 960...takes even that away from them. Certainly the two should never mingle.

I was just thinking about all of Reshevsky's games that degenerated into time pressure - and all sorts of other players as well. Then I was thinking about the G/130 with a 30 second delay we see at the Sinquefield Cup. That is really pretty luxurious given the depth of theory which is so often played. Classical chess is great...but really it is too darn slow for the top level where they prep (with the help of engines) their games against largely the same old faces every tournament. G/120 (G/90 would be good) with a 10 or 15 second delay is more than enough time for players of this caliber.

I agree with you that the time limit can be a bit faster now due to all the prep. I also agree that chess 960 is not an ideal solution to the prep problem, but I think that it is better than ignoring the problem. Basically we are just watching computers play each other for the first 15 to 20 moves or more, and then watching the humans take over. So why not let the computers pick randomly from a thousand or so reasonable, short opening sequences (ones with low drawing tendencies) and let the players start thinking from there (like we do with engine vs engine testing)? The positions can all be normal ones from GM play, but the players won't know which ones to prepare for so only general preparation will matter, not specific prep for an opponent. This is Kramnik's solution, and my first choice as well.

[leavenfish]

I agree with you that the time limit can be a bit faster now due to all the prep. I also agree that chess 960 is not an ideal solution to the prep problem, but I think that it is better than ignoring the problem. Basically we are just watching computers play each other for the first 15 to 20 moves or more, and then watching the humans take over. So why not let the computers pick randomly from a thousand or so reasonable, short opening sequences (ones with low drawing tendencies) and let the players start thinking from there (like we do with engine vs engine testing)? The positions can all be normal ones from GM play, but the players won't know which ones to prepare for so only general preparation will matter, not specific prep for an opponent. This is Kramnik's solution, and my first choice as well.

Not a terribly bad idea really. It would give an edge to the true renaissance men/women of chess. The Kramnik's of the world would arguably play professionally longer. Kramnik famously long cut down what he need to prepare with his 1.Nf3 approach.

The 'low drawing tendencies' thing though worries me. Should Ding Liren (playing for 1st place tomorrow) thru (bad) luck of the draw be forced to play a line that may not suit his style while Giri might gift Nepo a better chance of a win and catching or surpassing Liren because Giri is less good at playing a particular line he gets stuck with? If anything the simple randomizing of lines forced upon a player (without regard to 'drawing tendencies)should be enough.

I'll pick faster time controls and players playing for their pay ever game (wins only counting towards prize distribution) and let nature take its course...but there is going to be good and bad to any approach.