Details about Leela Chalange in Jerusalem

Discussion of anything and everything relating to chess playing software and machines.

Moderator: Ras

Uri Blass
Posts: 11086
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Details about Leela Chalange in Jerusalem

Post by Uri Blass »

some analysis of game 14 last game of the pieceodds bot(other rook odds gasmes are played with the rook odds bot)

[pgn][Event "casual classical game"]
[Site "https://lichess.org/BXBrdCuh"]
[Date "2025.11.29"]
[Round "?"]
[White "LeelaPieceOdds"]
[Black "jeruchess"]
[Result "1-0"]
[BlackElo "2400"]
[WhiteElo "2200"]
[TimeControl "5400+30"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/1NBQKBNR w Kkq - 0 1"]
[Termination "normal"]
[PlyCount "99"]
[WhiteType "program"]
[BlackType "human"]

1. e4 d5 2. exd5 Nf6 3. d4 c6 4. c3 Bf5 5. Nd2 Qxd5 6. Ngf3 e6 7. Nh4 Bg4
8. Qc2 Be7 (8. .. c5) (8. .. Qxa2) 9. Bc4 Qd8 10. h3 Bh5 11. g4 Bg6 12.
Nxg6 hxg6 13. Nf3 Qd6 14. g5 Nd5 15. O-O Nd7 (15. .. b5 16. Be2 Rxh3) 16.
h4 O-O-O (16. .. e5) 17. Re1 Rh5? (17. .. e5 18. dxe5 Qe6 19. Re4 Qh3 20.
Bf1 Qxf3 21. Be2 Qh3 22. Bg4 Qxh4 23. Bxd7+ Rxd7 24. Rxh4 Rxh4) (17. .. Qc7
{with the idea Bd6=f4 is better}) 18. Qe4 Rdh8 19. Bb3 N7f6 (19. .. e5) 20.
gxf6 gxf6 21. c4 Nc7? (21. .. f5 22. Qe5 (22. Qd3 Nf4) 22. .. Qxe5) 22. Bf4
Qd7 23. d5 e5 (23. .. cxd5) 24. Bg3 Qf5 (24. .. cxd5) 25. Qe2 Bc5 (25. ..
g5) 26. Kg2 Qd7 (26. .. g5) 27. Ba4 g5 (27. .. b5) 28. dxc6 bxc6 29. Qe4
gxh4 30. Nxh4 Qd4 (30. .. Qe6) 31. Qxd4 Bxd4 32. Nf3 Bxb2 (32. .. c5) 33.
Rb1 Ba3 34. Bxc6 Bc5 (34. .. Ne6) 35. Nd2 Ne6 (35. .. Rg8) 36. Ne4 Kc7?
{(-2.08 ? -0.78) Mistake. R5h6 was best.} (36. .. R5h6 37. Bd5 Be7 38. Rb7
Nc7 39. Rxa7 Rd8 40. Bxf7 Rd4 41. f3 Rh7 42. Bg6) (36. .. Nf4+ 37. Kf3 Kc7)
37. Bd5 Bd4 (37. .. Be7 38. Rb7+ Kd8 39. Rxa7 Nc7 40. Nc3 Bc5 41. Rb7 Bd4
42. Ne4 f5 43. Nd6 {is another way for equality}) 38. Rb7+ Kd8 39. Nd6 Nc7
40. Bf3 R5h7 41. Bg4 Ne6?! {Inaccuracy. Bc5 was best.} (41. .. Bc5 42. Ne4
Bd4 43. Nd6 Bc5) 42. Nxf7+ Rxf7 43. Rxf7 Re8? {a losing move . Nc5 was
best.} (43. .. Nc5 44. Rxa7) 44. Rxf6 Kd7 (44. .. Nc5 45. Rc6) 45. Rh6 Kd6
46. Kf3 Bc3 47. Ke4 a5 48. c5+ Kxc5 49. Bxe6 Rd8 50. Bxe5 {. Black resigns}
1-0
[/pgn]
Uri Blass
Posts: 11086
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Details about Leela Chalange in Jerusalem

Post by Uri Blass »

analysis of game 15
Hunan decided to repeat moves so a fast draw.

[pgn][Event "casual classical game"]
[Site "https://lichess.org/abstxHfM"]
[Date "2025.11.29"]
[Round "?"]
[White "LeelaRookOdds"]
[Black "Messithegoat2024"]
[Result "1/2-1/2"]
[BlackElo "2400"]
[WhiteElo "2000"]
[TimeControl "5400+30"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/1NBQKBNR w Kkq - 0 1"]
[Termination "normal"]
[PlyCount "22"]
[WhiteType "program"]
[BlackType "human"]

1. d4 Nf6 2. e3 d5 3. Bd3 e6 4. f4 c5 5. c3 Bd6 6. Nh3 b6 7. Nd2 a5 8. Bb5+
Bd7 9. Bd3 Bc8 10. Qa4+ Bd7 11. Qd1 Bc8 {The game is a draw.} (11. .. Bc6
{was a better move}) 1/2-1/2
[/pgn]
User avatar
towforce
Posts: 12671
Joined: Thu Mar 09, 2006 12:57 am
Location: Birmingham UK
Full name: Graham Laight

Re: Details about Leela Chalange in Jerusalem

Post by towforce »

Uri Blass wrote: Thu Dec 11, 2025 9:03 am analysis of game 15
Hunan decided to repeat moves so a fast draw.

[pgn][Event "casual classical game"]
[Site "https://lichess.org/abstxHfM"]
[Date "2025.11.29"]
[Round "?"]
[White "LeelaRookOdds"]
[Black "Messithegoat2024"]
[Result "1/2-1/2"]
[BlackElo "2400"]
[WhiteElo "2000"]
[TimeControl "5400+30"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/1NBQKBNR w Kkq - 0 1"]
[Termination "normal"]
[PlyCount "22"]
[WhiteType "program"]
[BlackType "human"]

1. d4 Nf6 2. e3 d5 3. Bd3 e6 4. f4 c5 5. c3 Bd6 6. Nh3 b6 7. Nd2 a5 8. Bb5+
Bd7 9. Bd3 Bc8 10. Qa4+ Bd7 11. Qd1 Bc8 {The game is a draw.} (11. .. Bc6
{was a better move}) 1/2-1/2
[/pgn]

Leela needs a contempt factor: she should not have allowed this draw!
Human chess is partly about tactics and strategy, but mostly about memory
Uri Blass
Posts: 11086
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Details about Leela Chalange in Jerusalem

Post by Uri Blass »

towforce wrote: Thu Dec 11, 2025 12:36 pm
Uri Blass wrote: Thu Dec 11, 2025 9:03 am analysis of game 15
Hunan decided to repeat moves so a fast draw.

[pgn][Event "casual classical game"]
[Site "https://lichess.org/abstxHfM"]
[Date "2025.11.29"]
[Round "?"]
[White "LeelaRookOdds"]
[Black "Messithegoat2024"]
[Result "1/2-1/2"]
[BlackElo "2400"]
[WhiteElo "2000"]
[TimeControl "5400+30"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/1NBQKBNR w Kkq - 0 1"]
[Termination "normal"]
[PlyCount "22"]
[WhiteType "program"]
[BlackType "human"]

1. d4 Nf6 2. e3 d5 3. Bd3 e6 4. f4 c5 5. c3 Bd6 6. Nh3 b6 7. Nd2 a5 8. Bb5+
Bd7 9. Bd3 Bc8 10. Qa4+ Bd7 11. Qd1 Bc8 {The game is a draw.} (11. .. Bc6
{was a better move}) 1/2-1/2
[/pgn]

Leela needs a contempt factor: she should not have allowed this draw!
I suspect that there is some bug because in the next game even part of the old engines do not allow repetition and leela also allowed a fast repetition. I will post that game in the next post
Uri Blass
Posts: 11086
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Details about Leela Chalange in Jerusalem

Post by Uri Blass »

game 16 again Leela allowed a fast repetition.

[pgn][Event "casual classical game"]
[Site "https://lichess.org/ed2Cc3DE"]
[Date "2025.11.29"]
[Round "?"]
[White "LeelaRookOdds"]
[Black "Dana_kochavi"]
[Result "1/2-1/2"]
[BlackElo "2400"]
[WhiteElo "2200"]
[TimeControl "5400+30"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/1NBQKBNR w Kkq - 0 1"]
[Termination "normal"]
[PlyCount "48"]
[WhiteType "program"]
[BlackType "human"]

1. c4 Nf6 2. Nc3 g6 3. d4 d6 4. e4 Bg7 5. Be2 O-O 6. g4 c5 7. d5 b5 8. g5
Nfd7 (8. .. Nh5) 9. Nxb5 Ba6 (9. .. a6) 10. Na3 Qb6 11. b3 e6 (11. .. Bc8)
12. h4 exd5 13. exd5 Re8 (13. .. Bc8) 14. Kf1 Nf8 15. Bd2 Bc8 16. h5 Bf5
17. Nf3 Bg4 18. Ng1 Bf5 19. Nf3 Bg4 20. h6 Bh8 21. Rh4 Bf5 22. Rh1 Bg4 23.
Rh4 Bf5 (23. .. Bxf3) 24. Rh1 (24. Nc2) 24. .. Bg4?? {The game is a draw.}
(24. .. a6 25. Nh4 Bc8 26. f4 Ra7 27. Bd3 Rae7 28. f5 Nbd7 29. Nc2 Ne5 30.
Bc3) (24. .. Na6) 1/2-1/2
[/pgn]


last iteration by some engine with no NN suggest different moves than 24.Rh1 that is a repetition but not third repetition because the opponent does not have to repeat and can play 24..Na6

[d]rn2rnkb/p4p1p/1q1p2pP/2pP1bP1/2P4R/NP3N2/P2BBP2/3Q1K2 w - - 7 24

Wasp450-x64-modern:

33/47 27:59 14,255,474k 8,486k -4.36 Nf3-h2 Nb8-d7 Be2-g4 Bf5xg4 Rh4xg4 f7-f6 Nh2-f3 f6-f5 Rg4-g3 Nd7-e5 Nf3-h2 Ne5-f7 f2-f4 Nf8-d7 Nh2-f3 Bh8-d4 Na3-b5 Re8-e4 Qd1-c2 a7-a6 Nb5-c3 Re4-e7 Kf1-g2 Ra8-e8 Kg2-h1 Qb6-c7 Rg3-g2 Bd4-e3 Rg2-e2 Be3xd2 Re2xe7 Re8xe7 Qc2xd2
33/47 23:59 12,176,927k 8,461k -4.30 Bd2-f4 Nb8-a6 Na3-b5 Ra8-d8 Bf4-d2 Na6-c7 Nb5-c3 Qb6-a6 Qd1-c1 Nf8-d7 Bd2-e3 Re8-e7 a2-a3 Rd8-b8 Be2-d1 Nd7-e5 Nf3xe5 Bh8xe5 Kf1-g1 Be5-d4 Be3xd4 c5xd4 Rh4xd4 Re7-e1+ Kg1-g2 Rb8-e8 Rd4-h4 Re8-e5 Qc1-d2 Qa6-a5 Rh4-h1 Re1xh1 Kg2xh1 Qa5xa3
33/47 22:42 11,509,206k 8,449k -4.26 Na3-c2 Nb8-d7 Nc2-e3 Bf5-e4 Nf3-h2 Qb6-d8 Rh4-g4 Bh8-d4 Ne3-c2 Be4xc2 Qd1xc2 Qd8-e7 Be2-d1 f7-f5 Rg4-h4 Nd7-e5 Kf1-g1 Nf8-d7 Kg1-g2 a7-a5 a2-a4 Ra8-a7 Bd2-f4 Ra7-b7 Qc2-d2 Re8-b8 Nh2-f1 Ne5-f7 Nf1-g3 Nd7-e5 Ng3-e2 Rb7-b4 Ne2xd4 c5xd4
---------------------------------------------------------------------------
34/48 38:54 19,960,993k 8,552k -4.39 Na3-b5 a7-a6 Nb5-a3 Nb8-d7 Rh4-h2 Nd7-e5 Nf3-h4 Bf5-d3 Be2xd3 Ne5xd3 Na3-b1 Bh8-e5 Rh2-g2 Nd3-f4 Rg2-g4 Nf4-h5 Nb1-c3 Nf8-d7 Nh4-f3 f7-f5 Rg4-h4 Qb6-d8 Nc3-e2 Qd8-e7 Rh4-h1 Be5-d4 Rh1-h4 Bd4-b2 Ne2-f4 Nh5xf4 Bd2xf4 Bb2-d4 b3-b4 Ra8-c8 Nf3xd4 c5xd4
34/47 32:43 16,715,090k 8,513k -4.44 Na3-c2 Nb8-d7 Nc2-e3 Bf5-e4 Rh4-h3 Qb6-d8 Ne3-g4 Be4-f5 Rh3-g3 Bf5xg4 Rg3xg4 Nd7-e5 Nf3xe5 Bh8xe5 f2-f4 Be5-d4 Kf1-g2 Nf8-d7 Rg4-g3 Ra8-b8 Be2-d3 a7-a5 a2-a4 f7-f5 Bd3-c2 Re8-e7 Rg3-h3 Qd8-b6 Rh3-h4 Rb8-e8 Bc2-d3 Nd7-f8 Rh4-h3 Nf8-d7
34/47 32:01 16,353,153k 8,510k -4.29 Bd2-f4 Nb8-a6 Na3-b5 Ra8-d8 Bf4-d2 Na6-c7 Nb5-c3 Qb6-a6 Qd1-c1 Nf8-d7 Bd2-e3 Re8-e7 a2-a3 Rd8-b8 Be2-d1 Nd7-e5 Nf3xe5 Bh8xe5 Kf1-g2 Be5-h8 a3-a4 Bf5-d3 Kg2-g1 Qa6-b6 Kg1-g2 Re7-e8 Qc1-d2 Bd3-f5 Nc3-e2 Re8-e4 Ne2-f4 Qb6-b4 Qd2xb4 Rb8xb4
---------------------------------------------------------------------------
35/49 50:25 25,949,929k 8,578k -4.44 Na3-c2 Nb8-d7 Nc2-e3 Bf5-e4 Rh4-h3 Qb6-d8 Ne3-g4 Be4-f5 Rh3-g3 Bf5xg4 Rg3xg4 Nd7-e5 Nf3xe5 Bh8xe5 Be2-d3 Be5-d4 Rg4-g3 Nf8-d7 f2-f4 f7-f5 Rg3-h3 Re8-e7 Kf1-g2 Qd8-b6 Rh3-h2 Ra8-e8 Rh2-h1 Qb6-a6 a2-a4 Qa6-b6 a4-a5 Qb6-b7 Kg2-g3 Kg8-f7 Qd1-c2
35/49 47:13 24,287,235k 8,572k -4.32 Na3-b5 a7-a6 Nb5-a3 Nb8-d7 Rh4-h2 Nd7-e5 Nf3-h4 Bf5-d3 Be2xd3 Ne5xd3 Na3-b1 Bh8-e5 Rh2-h3 Nd3-f4 Bd2xf4 Be5xf4 Qd1-g4 Bf4-e5 Nh4-f3 Qb6-a5 a2-a4 Qa5-d8 Nb1-d2 Nf8-d7 Kf1-g2 f7-f5 g5xf6/ep Nd7xf6 Qg4-g5 Kg8-f8 Rh3-h4 Ra8-b8 Nf3xe5 Re8xe5 Qg5-f4
35/49 44:41 22,982,243k 8,570k -4.41 Bd2-f4 Nb8-a6 Na3-b5 Ra8-d8 Bf4-d2 Na6-c7 Nb5-c3 Qb6-a6 Qd1-c1 Nf8-d7 Bd2-e3 Re8-e7 Be2-d1 Nd7-e5 Nf3xe5 Re7xe5 Kf1-g2 Re5-e7 Bd1-f3 Rd8-e8 Rh4-h1 Bh8-e5 Qc1-d2 Qa6-a5 Nc3-e2 Qa5xd2 Be3xd2 Be5-b2 Ne2-g3 Bf5-d3 Bd2-e3 f7-f5 g5xf6/ep Bb2xf6 Rh1-d1
---------------------------------------------------------------------------
36/53 1:32:45 47,860,993k 8,599k -4.60 Na3-b5 a7-a6 Nb5-a3 Nb8-d7 Rh4-h2 Nd7-e5 Nf3-h4 Bf5-d3 Be2xd3 Ne5xd3 Na3-b1 Bh8-e5 Rh2-h3 Nd3-f4 Bd2xf4 Be5xf4 Qd1-g4 Bf4-e5 Nh4-f3 Qb6-a5 a2-a4 Qa5-d8 Nb1-d2 Nf8-d7 Kf1-g2 f7-f5 Qg4-h4 Be5-c3 Rh3-h1 Bc3xd2 Nf3xd2 Re8-e2 Nd2-f1 Kg8-f7 Nf1-g3 Re2-d2
36/50 1:05:51 33,849,588k 8,566k -4.48 Na3-c2 Nb8-d7 Nc2-e3 Bf5-e4 Rh4-h3 Qb6-d8 Ne3-g4 Be4-f5 Nf3-h4 Bf5xg4 Be2xg4 Nd7-e5 Bg4-e2 f7-f6 g5xf6 Bh8xf6 Nh4-f3 Ne5xf3 Be2xf3 Bf6-d4 Bf3-g4 Qd8-f6 Qd1-f3 Qf6xf3 Rh3xf3 Ra8-b8 Bd2-g5 Rb8-b6 Bg5-d2 Re8-e4 Rf3-f4 Re4xf4 Bd2xf4 Kg8-f7 Kf1-g2 Rb6-a6
36/50 57:18 29,468,303k 8,571k -4.37 Bd2-f4 Nb8-a6 Na3-b5 Ra8-d8 Bf4-d2 Na6-c7 Nb5-c3 Qb6-a6 Qd1-c1 Nf8-d7 Nf3-e1 Rd8-b8 Ne1-f3 Re8-e7 Be2-d1 Nc7-e8 a2-a4 Qa6-b7 Nf3-e1 Nd7-e5 Kf1-g2 Ne5-d3 Ne1xd3 Bf5xd3 Bd2-e3 Bd3-f5 Kg2-g1 Bh8-e5 Nc3-e2 a7-a5 Ne2-g3 Be5xg3 f2xg3 Re7-e5 Be3-d2 Qb7-b6
---------------------------------------------------------------------------
37/55 3:05:21 95,794,691k 8,614k -4.67 Rh4-h1 Nb8-a6 Nf3-h4 Bf5-d7 Na3-b1 Re8-e7 Nb1-c3 Na6-b4 Kf1-g1 Ra8-e8 Be2-f1 Qb6-a6 a2-a4 Qa6-b6 Qd1-c1 a7-a6 Rh1-h2 Bh8-d4 Nh4-f3 Bd7-f5 a4-a5 Qb6-b8 Nf3xd4 c5xd4 Nc3-a4 Nb4-d3 Bf1xd3 Bf5xd3 Na4-b6 Re7-e4 f2-f3 Re4-e2 Rh2xe2 Re8xe2 b3-b4 Bd3-f5 Qc1-d1
37/55 2:36:37 80,887,768k 8,607k -4.65 Na3-c2 Nb8-d7 Nc2-e3 Bf5-e4 Rh4-h3 Qb6-d8 Ne3-g4 Be4-f5 Rh3-g3 Bf5xg4 Rg3xg4 Nd7-e5 Nf3xe5 Bh8xe5 Be2-d3 Be5-d4 Rg4-g3 Nf8-d7 f2-f4 f7-f6 f4-f5 Nd7-e5 Bd3-c2 f6xg5 Bd2xg5 Qd8-d7 Kf1-g2 Re8-f8 Qd1-f1 Kg8-h8 f5-f6 Ne5-g4 Rg3-f3 Ng4-f2 Bg5-d2 Nf2-e4 Bc2xe4 Qd7-g4+ Kg2-h1 Qg4xe4
37/53 1:42:34 52,955,390k 8,604k -4.38 Bd2-f4 Nb8-a6 Na3-b5 Ra8-d8 Bf4-d2 Na6-c7 Nb5-c3 Qb6-a6 Qd1-c1 Nf8-d7 Be2-d1 Nd7-e5 Nf3-e1 Qa6-c8 Bd1-c2 Ne5-g4 Qc1-b1 Bf5xc2 Qb1xc2 Bh8-d4 Ne1-f3 f7-f5 Qc2-b2 Bd4-h8 Qb2-c2 Re8-e7 Nc3-e2 Rd8-e8 Ne2-g3 Nc7-a6 Qc2-b1 Bh8-d4 Nf3xd4 c5xd4 Qb1-d3 Qc8-c5 Kf1-g1
---------------------------------------------------------------------------
.
.
38/55 3:24:47 105,982,106k 8,625k -4.56 Na3-c2 Nb8-d7 Nc2-e3 Bf5-e4 Rh4-h3 Qb6-d8 Ne3-g4 Be4-f5 Rh3-g3 Bf5xg4 Rg3xg4 Nd7-e5 Rg4-g3 Ne5xf3 Be2xf3 Bh8-e5 Rg3-h3 Be5-d4 Bf3-g4 Qd8-e7 Qd1-f3 Nf8-d7 Qf3-f4 Qe7-e5 Qf4-f3 Ra8-d8 Kf1-g2 Qe5-e7 Qf3-f4 Qe7-e4+ Qf4xe4 Re8xe4 Bg4-f3 Re4-e7 Bd2-a5 Rd8-c8 Ba5-d2 Nd7-e5 Bf3-e2
Uri Blass
Posts: 11086
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Details about Leela Chalange in Jerusalem

Post by Uri Blass »

Larry Kaufman explained about the last draw:

"remember it is using a mixed knight odds and rook odds net, so it may still take a draw if down between knight and rook"

Note that it is not iclear if it is llogical decision to allow repetition when the advantage is between a knight and a rook because otherwise it is possible that Leela is going to play some worse move to avoid repetition against a strong player when the strong player even did not mean to make a repetition and the weak move may help the strong player to win.
Uri Blass
Posts: 11086
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Details about Leela Chalange in Jerusalem

Post by Uri Blass »

Tיhe explanation of larry is simply that the net fror the rookOdds bot and the knight odd bots is the same in the last year and improvement in the LeelaRookOdds and LeelaKnightOdds were only in the search