leelapieceoddsfrc

[Father]

Question please: ... Is it a new engine ?

[lkaufman]

It is a new bot, using whatever net is best at the same odds in standard chess. The knight odds net has special training for FRC, which also helps at rook odds since they use the same net; at the higher handicaps it has no training for FRC, but still seems stronger against humans than at the same odds in standard chess, since players can't rely on playing their favorite openings, they have to think for themselves. Aleady in one bullet game a player rated over 2300 got checkmated in four moves despite taking the maximum handicap of Queen and Two Rooks!! Sounds impossible, but it happened. We encourage you and others reading this to give it a try, you will probably need double or triple whatever time you need in standard chess at the same odds.

[lucario6607]

Due to hardware problems it isn't online all the time, and probably has a very low concurrency.

[Ajedrecista]

Hello Larry:

[...] Aleady in one bullet game a player rated over 2300 got checkmated in four moves despite taking the maximum handicap of Queen and Two Rooks!! Sounds impossible, but it happened. [...]

FWIW, here is the game:

[pgn][Event "casual bullet game"]
[Site "https://lichess.org/"]
[Date "2025.11.07"]
[White "LeelaPieceOddsFRC"]
[Black ""]
[Result "1-0"]
[UTCDate "2025.11.07"]
[UTCTime "18:05:02"]
[WhiteElo "2000"]
[BlackElo "2302"]
[WhiteTitle "BOT"]
[Variant "From Position"]
[TimeControl "60+1"]
[ECO "?"]
[Opening "?"]
[Termination "Normal"]
[FEN "brknqbnr/pppppppp/8/8/8/8/PPPPPPPP/B1KN1BN1 w kq - 0 1"]
[SetUp "1"]
[Annotator "lichess.org"]

1. b4?? { (0.15 → -9.70) Blunder. e4 was best. } (1. e4 e6 2. Nf3 Nf6 3. Nc3 Nc6 4. d4 Kd8 5. g4 a6 6. g5 Ng4) 1... f6 2. Nc3 e5 3. Nb5 a6?? { (-9.97 → Mate in 1) Checkmate is now unavoidable. Nc6 was best. } (3... Nc6 4. d4 a6 5. e4 axb5 6. dxe5 Kd8 7. a3 fxe5 8. Bxb5 Qe7 9. Bb2) 4. Na7# { White wins by checkmate. } 1-0[/pgn]

Thank goodness Bobby Fischer is not aware of this game, otherwise he would call the human player 'a disgrace to the human race', as he already did in August, 1968, when a human player lost against MacHack VI computer:


Source: Fischer's chess column at the Boys' Life issue of August, 1968 (screenshot courtesy of Chess Notes site).

[pgn][Event ""]
[Site ""]
[Round ""]
[Date "1967.??.??"]
[White "MacHack VI"]
[Black "Landey, Ben"]
[PlyCount "41"]
[Result "1-0"]

{Notes by Fischer} 1.e4 c5 2.d4 cxd4 3.Qxd4 {Leads to loss of time. Try to avoid early queen moves. 3.Nf3 was better.} 3...Nc6 4.Qd3 Nf6 5.Nc3 g6
6.Nf3 d6 7.Bf4 e5? {Anti-positional, weakens square on d5.} 8.Bg3? {8.Bg5 pinning knight and indirectly gaining control on d5 was logical.} 8...a6 9.O-O-O b5 10.a4 Bh6+? {Bad because it removes the d pawn's protection – the machine takes advantage of this. Simply 10...bxa4! and Black's still OK.}
11.Kb1 b4 12.Qxd6! Bd7 13.Bh4! Bg7 14.Nd5 Nxe4 15.Nc7+ Qxc7 16.Qxc7 Nc5 17.Qd6 Bf8 18.Qd5 Rc8 19.Nxe5 Be6 20.Qxc6!! Rxc6
21.Rd8# {A very powerful finish by the machine.} 1-0[/pgn]

I also found the game at page 276 of the September, 1967 issue of Chess Life magazine. I corrected the version of MackHack in the PGN as well as added the first name of the human player and partially guessed the date. Ben Landey could be Benjamin M. Landey, which appears in the issue of August, 1967 of the same magazine.

Fun fact 1: you are cited at the end of the Chess Life article! 'Larry Kaufman, one of MacHack's best friends, [...]'.

Fun fact 2: the 1967 game has some coincidences with the oldest game that we know, from circa 1475: both games end with with a white win by checkmate in 21 moves, both giving checkmate with a piece on d8 square.

Fun fact 3: the checkmate pattern of the 1967 game is Réti's, as one of the two possible checkmates in the very famous game against Tartakower in 1910 (10.- ..., Ke8; 11.- Rd8# or 10.- ..., Kc7; 11.- Bd8#).

Regards from Spain.

Ajedrecista.

[lkaufman]

Hello Larry:

[...] Aleady in one bullet game a player rated over 2300 got checkmated in four moves despite taking the maximum handicap of Queen and Two Rooks!! Sounds impossible, but it happened. [...]

FWIW, here is the game:

[pgn][Event "casual bullet game"]
[Site "https://lichess.org/"]
[Date "2025.11.07"]
[White "LeelaPieceOddsFRC"]
[Black ""]
[Result "1-0"]
[UTCDate "2025.11.07"]
[UTCTime "18:05:02"]
[WhiteElo "2000"]
[BlackElo "2302"]
[WhiteTitle "BOT"]
[Variant "From Position"]
[TimeControl "60+1"]
[ECO "?"]
[Opening "?"]
[Termination "Normal"]
[FEN "brknqbnr/pppppppp/8/8/8/8/PPPPPPPP/B1KN1BN1 w kq - 0 1"]
[SetUp "1"]
[Annotator "lichess.org"]

1. b4?? { (0.15 → -9.70) Blunder. e4 was best. } (1. e4 e6 2. Nf3 Nf6 3. Nc3 Nc6 4. d4 Kd8 5. g4 a6 6. g5 Ng4) 1... f6 2. Nc3 e5 3. Nb5 a6?? { (-9.97 → Mate in 1) Checkmate is now unavoidable. Nc6 was best. } (3... Nc6 4. d4 a6 5. e4 axb5 6. dxe5 Kd8 7. a3 fxe5 8. Bxb5 Qe7 9. Bb2) 4. Na7# { White wins by checkmate. } 1-0[/pgn]

Thank goodness Bobby Fischer is not aware of this game, otherwise he would call the human player 'a disgrace to the human race', as he already did in August, 1968, when a human player lost against MacHack VI computer:


Source: Fischer's chess column at the Boys' Life issue of August, 1968 (screenshot courtesy of Chess Notes site).

[pgn][Event ""]
[Site ""]
[Round ""]
[Date "1967.??.??"]
[White "MacHack VI"]
[Black "Landey, Ben"]
[PlyCount "41"]
[Result "1-0"]

{Notes by Fischer} 1.e4 c5 2.d4 cxd4 3.Qxd4 {Leads to loss of time. Try to avoid early queen moves. 3.Nf3 was better.} 3...Nc6 4.Qd3 Nf6 5.Nc3 g6
6.Nf3 d6 7.Bf4 e5? {Anti-positional, weakens square on d5.} 8.Bg3? {8.Bg5 pinning knight and indirectly gaining control on d5 was logical.} 8...a6 9.O-O-O b5 10.a4 Bh6+? {Bad because it removes the d pawn's protection – the machine takes advantage of this. Simply 10...bxa4! and Black's still OK.}
11.Kb1 b4 12.Qxd6! Bd7 13.Bh4! Bg7 14.Nd5 Nxe4 15.Nc7+ Qxc7 16.Qxc7 Nc5 17.Qd6 Bf8 18.Qd5 Rc8 19.Nxe5 Be6 20.Qxc6!! Rxc6
21.Rd8# {A very powerful finish by the machine.} 1-0[/pgn]

I also found the game at page 276 of the September, 1967 issue of Chess Life magazine. I corrected the version of MackHack in the PGN as well as added the first name of the human player and partially guessed the date. Ben Landey could be Benjamin M. Landey, which appears in the issue of August, 1967 of the same magazine.

Fun fact 1: you are cited at the end of the Chess Life article! 'Larry Kaufman, one of MacHack's best friends, [...]'.

Fun fact 2: the 1967 game has some coincidences with the oldest game that we know, from circa 1475: both games end with with a white win by checkmate in 21 moves, both giving checkmate with a piece on d8 square.

Fun fact 3: the checkmate pattern of the 1967 game is Réti's, as one of the two possible checkmates in the very famous game against Tartakower in 1910 (10.- ..., Ke8; 11.- Rd8# or 10.- ..., Kc7; 11.- Bd8#).

Regards from Spain.

Ajedrecista.

I remember this game, I was on the MacHack team; I wrote the opening book (apparently it was not ready when this game was played), and regularly played against MachHack myself giving it queen odds, with competitive results, suggesting ways to improve it. I think Fischer was too harsh in his comment, Landey was only rated in the 1500s or so if I remember correctly, and didn't seem to play below that level here. Now I'm again playing queen odds games as part of the team making LeelaXOdds, only this time I'm the receiver! How times have changed.

[Father]

It is a new bot, using whatever net is best at the same odds in standard chess. The knight odds net has special training for FRC, which also helps at rook odds since they use the same net; at the higher handicaps it has no training for FRC, but still seems stronger against humans than at the same odds in standard chess, since players can't rely on playing their favorite openings, they have to think for themselves. Aleady in one bullet game a player rated over 2300 got checkmated in four moves despite taking the maximum handicap of Queen and Two Rooks!! Sounds impossible, but it happened. We encourage you and others reading this to give it a try, you will probably need double or triple whatever time you need in standard chess at the same odds.

Thursday, November 27, 2025.
Good morning, Mr. Larry Kaufman and chess community. I hope you are all well and continuing to do well. Thank you, Mr. Larry Kaufman, for your message. I want to thank you, as well as the developers, creators, and collaborators, for the LeelaOdds and Odds robots in general; without a doubt, this is a "great step in the search for understanding and access to the source of wisdom."

Today is a great day, Thanksgiving Day. "I share my gratitude and best wishes for the great nation that teaches us about the solemnity of Thanksgiving."

You, Mr. Larey Kaufman, and the entire Leela team have already given me the best gift in my chess dreams: "Playing among many greats and competing against robots and managing to achieve two first places, a sixth place, and excellent positions in 4 robots. I have already reached the crown I dreamed of. Thanks without hesitation to all. Since last October 7th, I have abandoned chess due to the implementation of the new rules through which my positions in the draws were reduced. I would only be interested in returning when a draw has the value of 0.5 and a victory the value of 1. That way I would compete again. Catecan and his rider have been brought down and abandoned at the edge of the road. But we have not stopped, we continue very united in prayer, traveling through valleys and mountains after a star that It illuminates with special beauty and leads us to Bethlehem. Thank you very much, Mr. Larry Kaufman.

[lkaufman]

It is a new bot, using whatever net is best at the same odds in standard chess. The knight odds net has special training for FRC, which also helps at rook odds since they use the same net; at the higher handicaps it has no training for FRC, but still seems stronger against humans than at the same odds in standard chess, since players can't rely on playing their favorite openings, they have to think for themselves. Aleady in one bullet game a player rated over 2300 got checkmated in four moves despite taking the maximum handicap of Queen and Two Rooks!! Sounds impossible, but it happened. We encourage you and others reading this to give it a try, you will probably need double or triple whatever time you need in standard chess at the same odds.

Thursday, November 27, 2025.
Good morning, Mr. Larry Kaufman and chess community. I hope you are all well and continuing to do well. Thank you, Mr. Larry Kaufman, for your message. I want to thank you, as well as the developers, creators, and collaborators, for the LeelaOdds and Odds robots in general; without a doubt, this is a "great step in the search for understanding and access to the source of wisdom."

Today is a great day, Thanksgiving Day. "I share my gratitude and best wishes for the great nation that teaches us about the solemnity of Thanksgiving."

You, Mr. Larey Kaufman, and the entire Leela team have already given me the best gift in my chess dreams: "Playing among many greats and competing against robots and managing to achieve two first places, a sixth place, and excellent positions in 4 robots. I have already reached the crown I dreamed of. Thanks without hesitation to all. Since last October 7th, I have abandoned chess due to the implementation of the new rules through which my positions in the draws were reduced. I would only be interested in returning when a draw has the value of 0.5 and a victory the value of 1. That way I would compete again. Catecan and his rider have been brought down and abandoned at the edge of the road. But we have not stopped, we continue very united in prayer, traveling through valleys and mountains after a star that It illuminates with special beauty and leads us to Bethlehem. Thank you very much, Mr. Larry Kaufman.

Nice to hear from you. Yes, doing well, looking forward to Saturday's event organized by Uri. Rules regarding draws are up to Noah, I haven't heard of any planned changes. Currently I believe they are rated except after more than six games without a win. As long as you play at a time control and handicap where you win at least 20% or so of the games your draws will be rated. Currently I have the top spot on most of the two-piece handicaps, while Uri is on top at the very large handicaps involving three pieces, but I'm sure that any young IM or even strong FM could achieve the top place on most of these lists with enough games. But so far strong titled players have mostly stuck to playing the single-piece odds bots. Further improvements in many or even all of the bots are likely in the near future due to some very recent developments.