Amazing results of LeelaKnightOdds-Dev

[Marcus91]

It seems that fixing the bug change the table and now this is the table(I do not complain about reduction in the place of me that is logical but how it is possible that Catecan went down to 2291.9?)

I thought that even with dividing K by 2 considering the fact that he played so many games his rating should not go down by a lot because it already converged to his performance against the bot.
The only case when his rating should go down by a lot is the case that he improved significantly in the last games so the rating still did not converge to be close to the right rating.

Even I went down only by 30 elo or something like that and I could see 2246 or something close earlier after part of the update so I guess it was my rating before fixing the bug.

LeelaQueenOdds as white 1+0 (Reference) 2895.2
LeelaQueenOdds as black 1+0 (Reference) 2795.2
LeelaQueenOdds as white 1+1 (Reference) 2634.5
LeelaQueenOdds as white 3+0 (Reference) 2619.5
LeelaQueenOdds as black 1+1 (Reference) 2534.5
1 Orichess88 2524.5 73
LeelaQueenOdds as black 3+0 (Reference) 2519.5
2 Master_Unknown 2514.6 92
LeelaQueenOdds as white 3+2 (Reference) 2450.0
3 Ushkothebear 2443.7 184
4 Hissha 2398.4 94
5 Omnivor 2396.8 30
LeelaQueenOdds as white 5+3 (Reference) 2353.7
LeelaQueenOdds as black 3+2 (Reference) 2350.0
6 Respectful_Dave 2299.6 74
7 itore 2297.1 43
8 Catecan 2291.9 1288
LeelaQueenOdds as black 5+3 (Reference) 2253.7
9 Sakali 2251.1 146
10 Spreek 2229.8 45
11 sfy2011 2223.6 25
12 SilverMitt 2223.1 26
LeelaQueenOdds as white 10+5 (Reference) 2220.3
13 uriblass 2216.4 182
14 nn1133 2205.6 6346

Catecan went down because its recent performance is better than its older performance, all it has to do is maintain this performance and it will continue to rise.

I see the leader has 73 games in the table but search found only 65 games

https://lichess.org/@/Orichess88/search ... order=desc

What is the reason for the difference?

The bug likely occurs during the download of new games to update the leaderboard. Despite a check in the code to ensure only new games are added, it seems some are being added regardless. I will investigate how to fix it

[lkaufman]

Today was perhaps the most amazing result yet for Leela. It played 35 games at 1' plus 1" increment against Polish GM Marcel Kanarak, FIDE 2478, Lichess bullet rating 2716, giving Queen odds, and won by 33.5 to 1.5 (he played until he finally won the 35th game!). Scores like this have been seen at pure 1' 0" bullet, but to make such a score at queen odds with a minute plus increment against a young (age 31) GM is just incredible. This sure makes "Catecan" look good by comparison, although it's not a totally fair comparison since he usually plays White (queen and move odds) whereas these games were with Leela playing White as is traditional.

[Uri Blass]

Today was perhaps the most amazing result yet for Leela. It played 35 games at 1' plus 1" increment against Polish GM Marcel Kanarak, FIDE 2478, Lichess bullet rating 2716, giving Queen odds, and won by 33.5 to 1.5 (he played until he finally won the 35th game!). Scores like this have been seen at pure 1' 0" bullet, but to make such a score at queen odds with a minute plus increment against a young (age 31) GM is just incredible. This sure makes "Catecan" look good by comparison, although it's not a totally fair comparison since he usually plays White (queen and move odds) whereas these games were with Leela playing White as is traditional.

I think a lot of weaker players can do clearly better result at 1+1
Maybe the GM gave a weaker player to play in his usename.

[lkaufman]

Today was perhaps the most amazing result yet for Leela. It played 35 games at 1' plus 1" increment against Polish GM Marcel Kanarak, FIDE 2478, Lichess bullet rating 2716, giving Queen odds, and won by 33.5 to 1.5 (he played until he finally won the 35th game!). Scores like this have been seen at pure 1' 0" bullet, but to make such a score at queen odds with a minute plus increment against a young (age 31) GM is just incredible. This sure makes "Catecan" look good by comparison, although it's not a totally fair comparison since he usually plays White (queen and move odds) whereas these games were with Leela playing White as is traditional.

I think a lot of weaker players can do clearly better result at 1+1
Maybe the GM gave a weaker player to play in his usename.

Well, of course such things are possible, but it's hard to see what motivation he would have to do so, it only makes him look bad. I think that is very unlikely. Anyone can make a free account to play, a student wouldn't need to embarrass his teacher like this just to play the bot. My guess is that he just had a bad day for unknown reasons.

[Uri Blass]

I suspect that even I have chances to do better than losing 33.5-1.5 against LeelaQueenOdds with black at 1+1

I lost against it 8.5-0.5 at 3+2 after the upgrade and later lost 9.5-2.5 at 5+3 but I believe I can do clearly better than it and my problem was lack of experience.

I do not want to try 1+1 against Leela Queen odds because it means that even if I get 5% against it I lose rating points and I wonder if playing against the piece odds is basically the same except the fact that games are unrated or maybe the piece odds does not play or play in a different way than the queen odds bot.

Edit:I see the piece oddalready won a queen odds game with white

https://lichess.org/3lVxjv75

[lkaufman]

I suspect that even I have chances to do better than losing 33.5-1.5 against LeelaQueenOdds with black at 1+1

I lost against it 8.5-0.5 at 3+2 after the upgrade and later lost 9.5-2.5 at 5+3 but I believe I can do clearly better than it and my problem was lack of experience.

I do not want to try 1+1 against Leela Queen odds because it means that even if I get 5% against it I lose rating points and I wonder if playing against the piece odds is basically the same except the fact that games are unrated or maybe the piece odds does not play or play in a different way than the queen odds bot.

Edit:I see the piece oddalready won a queen odds game with white

https://lichess.org/3lVxjv75

I find that I need a few games to get used to playing against Leela at queen odds, but then I can win nearly every game (11.5 out of the last 12, the draw relating to stalemate tricks) with a ten second increment (and the required one minute basetime). I don't repeat openings or win by memory, but I learn how to avoid tricks, how to simplify, how to stay out of danger, and even when things go badly I can usually find a way to return material and reach a won endgame. This might even be enough for top place on the list, we'll see. I think this is actually useful experience for regular chess (at least quick chess), it's better than playing Carlsen in blitz (with or without odds). Now that I've shown the way, I expect others to pass me, I'm sure that many GMs and some IMs and FMs can run up the score with 5 seconds increment (as Orichess88 has done at 2' plus 3") and pass me that way. Maybe with enough practice I could start to score heavily with just 5 seconds increment, but I rather doubt it. As for the 1'1" series, I'm sure the GM could make many more draws if that was his goal, but most GMs will have the attitude that they "should" win at queen odds, even in bullet, and so will go all out to win every game, with the result we observed. As for the piece odds bot, currently it uses the queen odds net and book, although that could change if a net is trained for other large odds. My experience so far with it is that at 5'3", I am a big favorite at rook plus knight, but a big underdog at two minor pieces. Assuming that I am roughly at FM level now at age 77, this suggests that only players of IM level (regardless of actual title) or better can score well at 5'3" with odds of two minor pieces. That sounds about right to me.

[Uri Blass]

I suspect that even I have chances to do better than losing 33.5-1.5 against LeelaQueenOdds with black at 1+1

I lost against it 8.5-0.5 at 3+2 after the upgrade and later lost 9.5-2.5 at 5+3 but I believe I can do clearly better than it and my problem was lack of experience.

I do not want to try 1+1 against Leela Queen odds because it means that even if I get 5% against it I lose rating points and I wonder if playing against the piece odds is basically the same except the fact that games are unrated or maybe the piece odds does not play or play in a different way than the queen odds bot.

Edit:I see the piece oddalready won a queen odds game with white

https://lichess.org/3lVxjv75

I find that I need a few games to get used to playing against Leela at queen odds, but then I can win nearly every game (11.5 out of the last 12, the draw relating to stalemate tricks) with a ten second increment (and the required one minute basetime). I don't repeat openings or win by memory, but I learn how to avoid tricks, how to simplify, how to stay out of danger, and even when things go badly I can usually find a way to return material and reach a won endgame. This might even be enough for top place on the list, we'll see. I think this is actually useful experience for regular chess (at least quick chess), it's better than playing Carlsen in blitz (with or without odds). Now that I've shown the way, I expect others to pass me, I'm sure that many GMs and some IMs and FMs can run up the score with 5 seconds increment (as Orichess88 has done at 2' plus 3") and pass me that way. Maybe with enough practice I could start to score heavily with just 5 seconds increment, but I rather doubt it. As for the 1'1" series, I'm sure the GM could make many more draws if that was his goal, but most GMs will have the attitude that they "should" win at queen odds, even in bullet, and so will go all out to win every game, with the result we observed. As for the piece odds bot, currently it uses the queen odds net and book, although that could change if a net is trained for other large odds. My experience so far with it is that at 5'3", I am a big favorite at rook plus knight, but a big underdog at two minor pieces. Assuming that I am roughly at FM level now at age 77, this suggests that only players of IM level (regardless of actual title) or better can score well at 5'3" with odds of two minor pieces. That sounds about right to me.

I believe players can improve significantly and also that leela can improve.

I analyzed your draw and it seems that leela made a mistake.

35.Rxb6 was better than 35.Bf4 that allowed you to simplify
black is still objectively winning after 35.Rxb6 but I believe practically leela could win that game with this move.

Edit:Not sure if Leela could win but Leela could have better chances(a possible draw line was 35.Rxb6 Bc8 36.Rb2 Qxg3 37.Na6+ with a perpetual check).
https://lichess.org/yAJqs2OW/black

[Father]

I suspect that even I have chances to do better than losing 33.5-1.5 against LeelaQueenOdds with black at 1+1

I lost against it 8.5-0.5 at 3+2 after the upgrade and later lost 9.5-2.5 at 5+3 but I believe I can do clearly better than it and my problem was lack of experience.

I do not want to try 1+1 against Leela Queen odds because it means that even if I get 5% against it I lose rating points and I wonder if playing against the piece odds is basically the same except the fact that games are unrated or maybe the piece odds does not play or play in a different way than the queen odds bot.

Edit:I see the piece oddalready won a queen odds game with white

https://lichess.org/3lVxjv75

I find that I need a few games to get used to playing against Leela at queen odds, but then I can win nearly every game (11.5 out of the last 12, the draw relating to stalemate tricks) with a ten second increment (and the required one minute basetime). I don't repeat openings or win by memory, but I learn how to avoid tricks, how to simplify, how to stay out of danger, and even when things go badly I can usually find a way to return material and reach a won endgame. This might even be enough for top place on the list, we'll see. I think this is actually useful experience for regular chess (at least quick chess), it's better than playing Carlsen in blitz (with or without odds). Now that I've shown the way, I expect others to pass me, I'm sure that many GMs and some IMs and FMs can run up the score with 5 seconds increment (as Orichess88 has done at 2' plus 3") and pass me that way. Maybe with enough practice I could start to score heavily with just 5 seconds increment, but I rather doubt it. As for the 1'1" series, I'm sure the GM could make many more draws if that was his goal, but most GMs will have the attitude that they "should" win at queen odds, even in bullet, and so will go all out to win every game, with the result we observed. As for the piece odds bot, currently it uses the queen odds net and book, although that could change if a net is trained for other large odds. My experience so far with it is that at 5'3", I am a big favorite at rook plus knight, but a big underdog at two minor pieces. Assuming that I am roughly at FM level now at age 77, this suggests that only players of IM level (regardless of actual title) or better can score well at 5'3" with odds of two minor pieces. That sounds about right to me.

No place for
Without a doubt, LeelaQueenOdds is an extraordinary machine that gives us a lot of joy and fun playing chess. The achievements of LeelaQueenOdds are incredible to me, they are a fantasy, I love watching LeelaQueenOdds' games, it features high-level players, I marvel at watching the bullet games of 1 minute without increment and 1+1, they are true works of art, LeelaQueenOdds It enhances the dynamics and power of the stones exponentially and geometrically to levels never seen before.

[Father]

I suspect that even I have chances to do better than losing 33.5-1.5 against LeelaQueenOdds with black at 1+1

I lost against it 8.5-0.5 at 3+2 after the upgrade and later lost 9.5-2.5 at 5+3 but I believe I can do clearly better than it and my problem was lack of experience.

I do not want to try 1+1 against Leela Queen odds because it means that even if I get 5% against it I lose rating points and I wonder if playing against the piece odds is basically the same except the fact that games are unrated or maybe the piece odds does not play or play in a different way than the queen odds bot.

Edit:I see the piece oddalready won a queen odds game with white

https://lichess.org/3lVxjv75

I find that I need a few games to get used to playing against Leela at queen odds, but then I can win nearly every game (11.5 out of the last 12, the draw relating to stalemate tricks) with a ten second increment (and the required one minute basetime). I don't repeat openings or win by memory, but I learn how to avoid tricks, how to simplify, how to stay out of danger, and even when things go badly I can usually find a way to return material and reach a won endgame. This might even be enough for top place on the list, we'll see. I think this is actually useful experience for regular chess (at least quick chess), it's better than playing Carlsen in blitz (with or without odds). Now that I've shown the way, I expect others to pass me, I'm sure that many GMs and some IMs and FMs can run up the score with 5 seconds increment (as Orichess88 has done at 2' plus 3") and pass me that way. Maybe with enough practice I could start to score heavily with just 5 seconds increment, but I rather doubt it. As for the 1'1" series, I'm sure the GM could make many more draws if that was his goal, but most GMs will have the attitude that they "should" win at queen odds, even in bullet, and so will go all out to win every game, with the result we observed. As for the piece odds bot, currently it uses the queen odds net and book, although that could change if a net is trained for other large odds. My experience so far with it is that at 5'3", I am a big favorite at rook plus knight, but a big underdog at two minor pieces. Assuming that I am roughly at FM level now at age 77, this suggests that only players of IM level (regardless of actual title) or better can score well at 5'3" with odds of two minor pieces. That sounds about right to me.

No place for
Without a doubt, LeelaQueenOdds is an extraordinary machine that gives us a lot of joy and fun playing chess. The achievements of LeelaQueenOdds are incredible to me, they are a fantasy, I love watching LeelaQueenOdds' games, it features high-level players, I marvel at watching the bullet games of 1 minute without increment and 1+1, they are true works of art, LeelaQueenOdds It enhances the dynamics and power of the stones exponentially and geometrically to levels never seen before.

...more and more powerful players are seen against LeelaQueenOdds... on this occasion it is a super player with 3168 bulet elo in Lichess who falls at the feet of the Queen.

[pgn][Event "Casual bullet game"]
[Site "https://lichess.org/pkXvKRYN"]
[Date "2024.12.03"]
[White "LeelaQueenOdds"]
[Black "catask"]
[Result "1-0"]
[UTCDate "2024.12.03"]
[UTCTime "19:36:07"]
[WhiteElo "2000"]
[BlackElo "3168"]
[WhiteTitle "BOT"]
[Variant "From Position"]
[TimeControl "60+0"]
[ECO "?"]
[Opening "?"]
[Termination "Normal"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNB1KBNR w KQkq - 0 1"]
[SetUp "1"]
[Annotator "lichess.org"]

1. d4 Nf6 2. g4 g6 3. h4 Nxg4 4. h5 Nf6 5. hxg6 fxg6 6. Nc3 Bg7 7. e4 d6 8. Bg5 h6 9. Bd2 c6 10. O-O-O b5 11. Bd3 b4 12. Nce2 Ba6 13. e5 dxe5 14. Bxg6+ Kf8 15. dxe5 Nd5 16. Nd4 Qc8 17. Bf5 Qe8 18. Ne6+ Kg8 19. Nf3 { Black resigns. } 1-0[/pgn]

[lkaufman]

I suspect that even I have chances to do better than losing 33.5-1.5 against LeelaQueenOdds with black at 1+1

I lost against it 8.5-0.5 at 3+2 after the upgrade and later lost 9.5-2.5 at 5+3 but I believe I can do clearly better than it and my problem was lack of experience.

I do not want to try 1+1 against Leela Queen odds because it means that even if I get 5% against it I lose rating points and I wonder if playing against the piece odds is basically the same except the fact that games are unrated or maybe the piece odds does not play or play in a different way than the queen odds bot.

Edit:I see the piece oddalready won a queen odds game with white

https://lichess.org/3lVxjv75

I find that I need a few games to get used to playing against Leela at queen odds, but then I can win nearly every game (11.5 out of the last 12, the draw relating to stalemate tricks) with a ten second increment (and the required one minute basetime). I don't repeat openings or win by memory, but I learn how to avoid tricks, how to simplify, how to stay out of danger, and even when things go badly I can usually find a way to return material and reach a won endgame. This might even be enough for top place on the list, we'll see. I think this is actually useful experience for regular chess (at least quick chess), it's better than playing Carlsen in blitz (with or without odds). Now that I've shown the way, I expect others to pass me, I'm sure that many GMs and some IMs and FMs can run up the score with 5 seconds increment (as Orichess88 has done at 2' plus 3") and pass me that way. Maybe with enough practice I could start to score heavily with just 5 seconds increment, but I rather doubt it. As for the 1'1" series, I'm sure the GM could make many more draws if that was his goal, but most GMs will have the attitude that they "should" win at queen odds, even in bullet, and so will go all out to win every game, with the result we observed. As for the piece odds bot, currently it uses the queen odds net and book, although that could change if a net is trained for other large odds. My experience so far with it is that at 5'3", I am a big favorite at rook plus knight, but a big underdog at two minor pieces. Assuming that I am roughly at FM level now at age 77, this suggests that only players of IM level (regardless of actual title) or better can score well at 5'3" with odds of two minor pieces. That sounds about right to me.

No place for
Without a doubt, LeelaQueenOdds is an extraordinary machine that gives us a lot of joy and fun playing chess. The achievements of LeelaQueenOdds are incredible to me, they are a fantasy, I love watching LeelaQueenOdds' games, it features high-level players, I marvel at watching the bullet games of 1 minute without increment and 1+1, they are true works of art, LeelaQueenOdds It enhances the dynamics and power of the stones exponentially and geometrically to levels never seen before.

...more and more powerful players are seen against LeelaQueenOdds... on this occasion it is a super player with 3168 bulet elo in Lichess who falls at the feet of the Queen.

[pgn][Event "Casual bullet game"]
[Site "https://lichess.org/pkXvKRYN"]
[Date "2024.12.03"]
[White "LeelaQueenOdds"]
[Black "catask"]
[Result "1-0"]
[UTCDate "2024.12.03"]
[UTCTime "19:36:07"]
[WhiteElo "2000"]
[BlackElo "3168"]
[WhiteTitle "BOT"]
[Variant "From Position"]
[TimeControl "60+0"]
[ECO "?"]
[Opening "?"]
[Termination "Normal"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNB1KBNR w KQkq - 0 1"]
[SetUp "1"]
[Annotator "lichess.org"]

1. d4 Nf6 2. g4 g6 3. h4 Nxg4 4. h5 Nf6 5. hxg6 fxg6 6. Nc3 Bg7 7. e4 d6 8. Bg5 h6 9. Bd2 c6 10. O-O-O b5 11. Bd3 b4 12. Nce2 Ba6 13. e5 dxe5 14. Bxg6+ Kf8 15. dxe5 Nd5 16. Nd4 Qc8 17. Bf5 Qe8 18. Ne6+ Kg8 19. Nf3 { Black resigns. } 1-0[/pgn]

Yes, he scored just one draw in six 1'0" bullet games with LeelaQueenOdds. I happen to know who he is, but I can't reveal his identity, other than to say that he is indeed a very strong GM on the top fifty FIDE classical rating list. It is becoming pretty obvious that no human, not even Magnus or Hikaru, would win a 1'0" bullet match from LeelaQueenOdds (Leela playing White of course). You might actually have a better percent score than anyone in the world so far at 1'0" games, though only your games where you played Black can count for such a comparison as that is how almost all of these games have been played. Anyone trying to check this out should search for games from Nov 2 on, since that is the day the Leela queen odds net went live.