GM Awonder Liang vs Leela Knight Odds Rapid match

[lkaufman]

A Rapid knight odds match between GM Awonder Liang and LeelaKnightOdds has been scheduled on LiChess for Dec.12. starting at 3 pm Eastern Standard Time (EST), or 9 pm CET. Initial time limit 10'5" (bullet chess for Leela!). Liang may request a change in the time limit after every game pair, based on results. Alternating b1/g1 odds. First player to win six games wins the match (same rule as Kasparov/Karpov 1984-1985, let's hope the FIDE president doesn't have to halt the match after 48 games!). If the match is not finished with the game in progress after 3 hours, we plan to resume on the next suitable afternoon.
GM Liang is rated 2687 FIDE, recently tied for second in the U.S. Championship, recently won the chess.com chess 960 championship, and has had several major successes this year, most notably winning the World Open.
He will be the highest rated player ever to take knight odds from an engine on record. So far no one has a plus or even score against LeelaKnightOdds at any time control, although one anonymous GM from Israel managed to score 4.5 to 5.5 at 20' + 5" recently, so a close match is likely.

[Father]

A Rapid knight odds match between GM Awonder Liang and LeelaKnightOdds has been scheduled on LiChess for Dec.12. starting at 3 pm Eastern Standard Time (EST), or 9 pm CET. Initial time limit 10'5" (bullet chess for Leela!). Liang may request a change in the time limit after every game pair, based on results. Alternating b1/g1 odds. First player to win six games wins the match (same rule as Kasparov/Karpov 1984-1985, let's hope the FIDE president doesn't have to halt the match after 48 games!). If the match is not finished with the game in progress after 3 hours, we plan to resume on the next suitable afternoon.
GM Liang is rated 2687 FIDE, recently tied for second in the U.S. Championship, recently won the chess.com chess 960 championship, and has had several major successes this year, most notably winning the World Open.
He will be the highest rated player ever to take knight odds from an engine on record. So far no one has a plus or even score against LeelaKnightOdds at any time control, although one anonymous GM from Israel managed to score 4.5 to 5.5 at 20' + 5" recently, so a close match is likely.

Thank you Mr. Larry Kaufman. I look forward to such a wonderful battle... To GM Awonder Liang, to the LeelaOdds team, to Mr. Larry Kaufman: "Good wind, fair sea."

[Father]

A Rapid knight odds match between GM Awonder Liang and LeelaKnightOdds has been scheduled on LiChess for Dec.12. starting at 3 pm Eastern Standard Time (EST), or 9 pm CET. Initial time limit 10'5" (bullet chess for Leela!). Liang may request a change in the time limit after every game pair, based on results. Alternating b1/g1 odds. First player to win six games wins the match (same rule as Kasparov/Karpov 1984-1985, let's hope the FIDE president doesn't have to halt the match after 48 games!). If the match is not finished with the game in progress after 3 hours, we plan to resume on the next suitable afternoon.
GM Liang is rated 2687 FIDE, recently tied for second in the U.S. Championship, recently won the chess.com chess 960 championship, and has had several major successes this year, most notably winning the World Open.
He will be the highest rated player ever to take knight odds from an engine on record. So far no one has a plus or even score against LeelaKnightOdds at any time control, although one anonymous GM from Israel managed to score 4.5 to 5.5 at 20' + 5" recently, so a close match is likely.

Thank you Mr. Larry Kaufman. I look forward to such a wonderful battle... To GM Awonder Liang, to the LeelaOdds team, to Mr. Larry Kaufman: "Good wind, fair sea."

[pgn] [Event "Casual bullet game"]
[Site "https://lichess.org/t5ErcDK7"]
[Date "2024.12.07"]
[White "Catecan"]
[Black "LeelaQueenForKnight"]
[Result "1/2-1/2"]
[UTCDate "2024.12.07"]
[UTCTime "14:52:23"]
[WhiteElo "2030"]
[BlackElo "2000"]
[BlackTitle "BOT"]
[Variant "From Position"]
[TimeControl "60+0"]
[ECO "?"]
[Opening "?"]
[Termination "Normal"]
[FEN "rnb1kbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1"]
[SetUp "1"]
[Annotator "lichess.org"]

1. d4 Nf6 2. e3 d6 3. c3 e5 4. a3 Be6 5. f4 e4 6. g3 Be7 7. Bh3 Bc4 8. Bf1 Be6 9. Bh3 Bd5 10. Bg2 c5 11. Bd2 Nbd7 12. Qe2 c4 13. Bc1 a5 14. Qc2 Bc6 15. Bd2 d5 16. Ne2 h5 17. h4 a4 18. Bh3 Nb6 19. Rg1 Ng4 20. Bxg4 hxg4 21. Rg2 f5 22. Ng1 Nd7 23. Ke2 Nf6 24. Be1 Kf7 25. Bf2 Ne8 26. Ra2 Rh6 27. Ra1 Nf6 28. Rc1 Be8 29. Ke1 Ng8 30. Ke2 Rf6 31. Ra1 Rb6 32. Ra2 Rb3 33. Kd1 Bd6 34. Be1 Nf6 35. Re2 Nd7 36. Bd2 Nf6 37. Rg2 Ra6 38. Ke1 Nd7 39. Kf1 Rab6 40. Re2 Nf6 41. Kg2 Bd7 42. Be1 Ke8 43. Bd2 Kd8 44. Be1 Kc8 45. Bd2 Be8 46. Be1 Nh5 47. Bd2 Be7 48. Re1 Rh6 49. Re2 Rbb6 50. Be1 Rh8 51. Qc1 Rbh6 52. Qd2 Bd7 53. Qc1 Kc7 54. Qc2 b5 55. Ra1 Nf6 56. Ra2 Ne8 57. Ra1 Nd6 58. Ra2 Nb7 59. Ra1 Na5 60. Rd1 Nb3 61. Rb1 Na5 62. Rd1 Nb3 63. Rb1 Be8 64. Rd1 Bh5 65. Rb1 Be8 66. Rd1 Bf7 67. Rb1 Bh5 68. Rd1 Be8 69. Rb1 Bd7 70. Rd1 Be8 71. Rb1 Bh5 72. Rd1 Kd7 73. Rc1 Ke6 74. Rd1 Kf7 75. Rc1 Kg8 76. Rd1 Kf7 77. Rdd2 Ke6 78. Rd1 Kd7 79. Rdd2 Kc6 80. Rd1 Kb6 81. Rdd2 Be8 82. Rd1 Bd7 83. Rdd2 Nxd2 84. Qxd2 Be8 85. Qc2 Bh5 86. Bf2 Be8 87. Be1 Bd7 88. Bf2 Be8 89. Be1 { The game is a draw. } 1/2-1/2[/pgn]

[Father]

A Rapid knight odds match between GM Awonder Liang and LeelaKnightOdds has been scheduled on LiChess for Dec.12. starting at 3 pm Eastern Standard Time (EST), or 9 pm CET. Initial time limit 10'5" (bullet chess for Leela!). Liang may request a change in the time limit after every game pair, based on results. Alternating b1/g1 odds. First player to win six games wins the match (same rule as Kasparov/Karpov 1984-1985, let's hope the FIDE president doesn't have to halt the match after 48 games!). If the match is not finished with the game in progress after 3 hours, we plan to resume on the next suitable afternoon.
GM Liang is rated 2687 FIDE, recently tied for second in the U.S. Championship, recently won the chess.com chess 960 championship, and has had several major successes this year, most notably winning the World Open.
He will be the highest rated player ever to take knight odds from an engine on record. So far no one has a plus or even score against LeelaKnightOdds at any time control, although one anonymous GM from Israel managed to score 4.5 to 5.5 at 20' + 5" recently, so a close match is likely.

Before falling asleep I have been meditating about LeelaQueenOdds. I have a question for chess Grandmasters. Is this/ LeelaQueenOdds is giving some gambling lessons playing without a queen. In my opinion. Truly incredible and wonderful. Potentializes the dynamics of the game at a very high level. Has this changed the theory of openings and defenses? That is to say, if we add the queen to this way of playing, doesn't it generate super chess?

[Uri Blass]

A Rapid knight odds match between GM Awonder Liang and LeelaKnightOdds has been scheduled on LiChess for Dec.12. starting at 3 pm Eastern Standard Time (EST), or 9 pm CET. Initial time limit 10'5" (bullet chess for Leela!). Liang may request a change in the time limit after every game pair, based on results. Alternating b1/g1 odds. First player to win six games wins the match (same rule as Kasparov/Karpov 1984-1985, let's hope the FIDE president doesn't have to halt the match after 48 games!). If the match is not finished with the game in progress after 3 hours, we plan to resume on the next suitable afternoon.
GM Liang is rated 2687 FIDE, recently tied for second in the U.S. Championship, recently won the chess.com chess 960 championship, and has had several major successes this year, most notably winning the World Open.
He will be the highest rated player ever to take knight odds from an engine on record. So far no one has a plus or even score against LeelaKnightOdds at any time control, although one anonymous GM from Israel managed to score 4.5 to 5.5 at 20' + 5" recently, so a close match is likely.

I think that if you talk about the result against killer840 then it is slightly different.

It seems that one of the losses of him was at time control 3+2 and result at 15+5(not 20+5) is 4.5-4.5

[lkaufman]

A Rapid knight odds match between GM Awonder Liang and LeelaKnightOdds has been scheduled on LiChess for Dec.12. starting at 3 pm Eastern Standard Time (EST), or 9 pm CET. Initial time limit 10'5" (bullet chess for Leela!). Liang may request a change in the time limit after every game pair, based on results. Alternating b1/g1 odds. First player to win six games wins the match (same rule as Kasparov/Karpov 1984-1985, let's hope the FIDE president doesn't have to halt the match after 48 games!). If the match is not finished with the game in progress after 3 hours, we plan to resume on the next suitable afternoon.
GM Liang is rated 2687 FIDE, recently tied for second in the U.S. Championship, recently won the chess.com chess 960 championship, and has had several major successes this year, most notably winning the World Open.
He will be the highest rated player ever to take knight odds from an engine on record. So far no one has a plus or even score against LeelaKnightOdds at any time control, although one anonymous GM from Israel managed to score 4.5 to 5.5 at 20' + 5" recently, so a close match is likely.

I think that if you talk about the result against killer840 then it is slightly different.

It seems that one of the losses of him was at time control 3+2 and result at 15+5(not 20+5) is 4.5-4.5

Sorry, I remembered 20 as the time but that was the total of base plus increment, not the base, which has a 15 max for the bot. I hadn't noticed that one of the games was blitz. At least the book will be updated for the match, not yet sure about the net itself. Looks like we will have coverage of the match by GM Matthew Sadler "Silicon Road", perhaps more. Any guess as to who "killer840" might be?

[Jouni]

Why are odds matches played with Leela? Are Komodo or Stockfish not suitable?

[Uri Blass]

Why are odds matches played with Leela? Are Komodo or Stockfish not suitable?

because Leela is the best in these conditions.

Other engines will prefer to lose a rook by a simple capture and not to allow mate that the opponent may miss.
see 34...Re8 of Kaissa

https://en.wikipedia.org/wiki/Kaissa

[Uri Blass]

Here is the test position that I believe LeelaKnightOdds can solve and other top engines cannot solve
[d]Q5k1/4rp1p/1p1q1bp1/1B1n4/3Pp1P1/4B2P/PP3P2/2R3K1 b - - 0 1

find Kg7.

I tried KomodoDragon also with MCTS and it does not help and Dragon blunders like all engines with Re8?? that does not give a practical chance(at least against strong GM's at 10+5 when I believe Kg7 gives a small practical chance to save the game).

[cbash]

Why are odds matches played with Leela? Are Komodo or Stockfish not suitable?

Expanding a little bit on Uri's answer Leelaknightodds is currently using a specialized network tuned on knight odds games which has shown improvements at odds over the previous use of wdl contempt (itself significantly better than stockfish or komodo at knight odds). This is better at odds because it leads to Leela playing moves which maximize expected score against a human strength opponent rather than moves best against engine level opponents. Uri has provided an example position showing this, you can also look at some of its games on lichess to get a feel for its playing style and see how well it is doing in practice. One final note Leela currently has to limit search in order to maintain this style of play.