CCC Bullet Torch vs lc0

[Eelco de Groot]

Purely from an evaluation viewpoint, the game, or opening seems over after 22.f5. I have not checked if that was a difficult to find move from Torch (It would look good as a human GM brilliancy though ) but the reply by Lc0 seems below par. 22...b5? But is there anything that still draws. Not sure, but Crystal 7 thinks it's not very likely. These are the 3 best moves after a deep think:

[d]1r3r2/1p2ppbk/p1np2pp/5P1q/4P2P/1PP1B2N/1P1Q2P1/1K1R3R b - -
47 475:46 +1.82 22...gxf5 23.Pg5+ Kh8 24.exf5 e6 25.f6 Dg6+ 26.Ka2 Dxf6 27.Tdf1 De7 28.Ph3 Pe5 29.Lxh6 a5 30.Lxg7+ Kxg7 31.Pf4 Th8 32.Dd4 a4 33.bxa4 b5 34.Pd3 bxa4 35.Pxe5 dxe5 (13.650.221.525) 478
47 475:46 +1.84 22...e6 23.fxg6+ Dxg6 24.Pf2 f5 25.h5 Df7 26.Dxd6 Tbd8 27.Dh2 Txd1+ 28.Txd1 f4 29.Lb6 Tg8 30.Pd3 Le5 31.Dh1 Tg5 32.Pc5 Kh8 33.Td7 Dxh5 34.Dxh5 Txh5 35.Pxe6 Th1+ (13.650.221.525) 478
47 475:46 +2.38 22...Tbc8 23.Tdf1 Pe5 24.Tfg1 gxf5 25.g4 Pf3 26.gxh5 Pxd2+ 27.Lxd2 fxe4 28.Pf4 Le5 29.Tf1 Lxf4 30.Txf4 f5 31.Thf1 e6 32.c4 b5 33.Lb4 bxc4 34.bxc4 Tc6 35.La3 Tfc8 (13.650.221.525) 478

48 501:07 +1.91 22...e6 23.fxg6+ Dxg6 24.Pf2 f5 25.h5 Df7 26.Dxd6 Tbd8 27.Dh2 Le5 28.Dh3 Tg8 29.g4 f4 30.Df3 De7 31.Lc1 Tdf8 32.Pd3 Kh8 33.Tde1 Dg5 34.Thg1 Tf6 35.De2 Tgf8 (14.345.054.594) 477

Eval for e6 and gxf5 is sort of stable but if you just take the line or PV deeper, it might fall I think. I have not tried it. Only using 512 MB hash.

[marsell]

why only prefer white and not alternate with black? That would reduce the draw rate even more.

[Graham Banks]

In football, there are too many draws. Should we let the away team start with a 1-0 advantage?

[CornfedForever]

This match cannot be taken seriously. Because of the wrong book.

What exactly does that mean?

This guy thinks, using UHO Openings is something really bad and we all should enjoy tournaments with 90%+ draws instead...
Ridiculous. Dont waste your time.

90% drawn does not seem a problem. For human tournaments, sure. But these are tireless little pieces of software you can put thru their paces whiles you sleep. No one (with a life) sits and watches them as they are being played.

I occasionally watch engine v engine games.

TCEC and CCC have hundreds of spectators too.

Oh, I have peaked myself while they are in progress.

However for myself and about everyone else, I would think the number of human games (usually commentated on on by humans) we have spent watching (not all are uneventful of course) might be a problem for spectators...but 90%+ draws in engine vs engine tournaments....really isn't much of an issue (from a 'spectator' viewpoint) because it really is relatively few people who would have their eyed glued to such a thing.

[Krzysztof Grzelak]

The opening book must be fair to all game participants. So that it is not that one side has an advantage because it has nothing to do with an honest treatment of engine testing.

[dkappe]

The opening book must be fair to all game participants. So that it is not that one side has an advantage because it has nothing to do with an honest treatment of engine testing.

The idea with these UHO openings is that the exit position is on the edge of a winning advantage. You then have to give each engine the chance to play as white. Here is the same opening where lc0 played as white.

[pgn][Event "CCC21 Bullet: Would-be Challenger Match (1|1)"] [Site "Chess.com"] [Date "2023.12.16"] [Round "496"] [White "Lc0"] [Black "Torch"] [Result "1/2-1/2"] [ECO "B53"] [GameDuration "00:05:44"] [GameEndTime "2023-12-16T18:23:11.952 PST"] [GameStartTime "2023-12-16T18:17:27.683 PST"] [Opening "Sicilian, Chekhover Variation"] [PlyCount "176"] [Termination "adjudication"] [TimeControl "60+1"] 1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Qxd4 a6 5. Be3 Nc6 6. Qd2 g6 7. Nc3 Bg7 8. O-O-O Bg4 9. Nd5 Rc8 10. Nb6 Nf6 11. Nxc8 Qxc8 12. Bc4 O-O 13. Bd5 Nxd5 14. exd5 Ne5 15. Nxe5 Bxd1 16. Nxg6 hxg6 17. Rxd1 Qc4 18. Kb1 Qe4 19. g3 b5 20. Qd3 Qg4 21. h4 Rc8 22. a3 Rc4 23. Bb6 Kf8 24. c3 Qh3 25. Kc2 Qg2 26. Re1 Be5 27. Kb3 Bf6 28. Re2 Qh1 29. Rd2 Re4 30. Rd1 Qg2 31. Qf1 Qf3 32. Rd3 Qg4 33. Rd2 Qf3 34. Ka2 Kg7 35. Qd3 Qh1 36. Rd1 Qg2 37. Qf1 Qf3 38. Rd2 Kf8 39. Qd1 Qf5 40. Qb1 Qf3 41. Qd3 Qh1 42. Rd1 Qg2 43. Qf1 Qf3 44. Rd2 Kg8 45. Qd3 Qh1 46. Rd1 Qg2 47. Qf1 Qf3 48. Rd2 Kf8 49. Qd1 Qf5 50. Qb1 Qg4 51. Be3 Rc4 52. Qd3 a5 53. Qe2 Qf5 54. Qd3 Qh3 55. Rd1 a4 56. Rg1 Qg4 57. Rh1 Kg7 58. Kb1 Kg8 59. Bf4 Be5 60. Bxe5 dxe5 61. Re1 e4 62. Qe3 Qh5 63. Qd2 Kg7 64. Kc2 Qf5 65. Kc1 Qf3 66. Rf1 Rc8 67. Rg1 Kh7 68. Re1 Rc7 69. Qc2 Rd7 70. Qxe4 Qxf2 71. g4 Rxd5 72. h5 Qd2+ 73. Kb1 Qd3+ 74. Qxd3 Rxd3 75. Rxe7 Rf3 76. hxg6+ Kxg6 77. Re4 Rf1+ 78. Ka2 f5 79. gxf5+ Rxf5 80. Rb4 Rc5 81. c4 bxc4 82. Rxa4 Kf5 83. Kb1 Ke5 84. Kc2 Kd4 85. b3 Rc7 86. Kd2 Rc5 87. Rb4 Kd5 88. Ke3 cxb3 1/2-1/2[/pgn]

As you can see, lc0 was unable to convert the advantage. So in this “game pair,” the score was 1.5-0.5 in favor of Torch.