Antifish vs Stockfish in cccc

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

Moderators: hgm, Rebel, chrisw

Nay Lin Tun
Posts: 708
Joined: Mon Jan 16, 2012 6:34 am

Antifish vs Stockfish in cccc

Post by Nay Lin Tun »

The next CCC event is 4 days of Stockfish vs Lc0 (Antifish)

Some testing results say Antifish ( Leela derivative training with Stockfish games , rather than self played games like Leela) is as strong as Leela.

Who will win?
Uri Blass
Posts: 10268
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Antifish vs Stockfish in cccc

Post by Uri Blass »

Nay Lin Tun wrote: Mon Jan 21, 2019 8:28 am The next CCC event is 4 days of Stockfish vs Lc0 (Antifish)

Some testing results say Antifish ( Leela derivative training with Stockfish games , rather than self played games like Leela) is as strong as Leela.

Who will win?
I expect antifish to win because it designed to win against stockfish if I understand correctly but it does not mean that anti-fish is stronger than stockfish.

It is possible also to have anti-anti-fish that is designed to win against anti-fish and maybe stockfish is going to win against anti-anti-fish that is designed to win against anti-fish but not against stockfish.
User avatar
AdminX
Posts: 6339
Joined: Mon Mar 13, 2006 2:34 pm
Location: Acworth, GA

Re: Antifish vs Stockfish in cccc

Post by AdminX »

These Neural Networks (DeuX, AntiFish, etc.) reminds me of the Mephisto modules for the dedicated computer chessboards back in the day. I wonder why not just start a test network trained only by user who have Syzygybases installed. I would think it may take longer but you would get better results. IMHO. As for AntiFish, I downloaded it the other day but have not had time to test it out. Seems interesting, but still searching for one engine to rule them all like Stockfish. :D

PS: It's good to be the king, death to the king!
"Good decisions come from experience, and experience comes from bad decisions."
__________________________________________________________________
Ted Summers
Eduard
Posts: 1439
Joined: Sat Oct 27, 2018 12:58 am
Location: Germany
Full name: N.N.

Re: Antifish vs Stockfish in cccc

Post by Eduard »

Can be downloaded this network?
nabildanial
Posts: 126
Joined: Thu Jun 05, 2014 5:29 am
Location: Malaysia

Re: Antifish vs Stockfish in cccc

Post by nabildanial »

Eduard wrote: Mon Jan 21, 2019 6:29 pm Can be downloaded this network?
Taken from Fishcooking. The download link is available in the original post.
https://groups.google.com/forum/#!topic ... sBZnYA6B14
User avatar
AdminX
Posts: 6339
Joined: Mon Mar 13, 2006 2:34 pm
Location: Acworth, GA

Re: Antifish vs Stockfish in cccc

Post by AdminX »

Eduard wrote: Mon Jan 21, 2019 6:29 pm Can be downloaded this network?
AF 2-20000 Mark 154 [Best Mse Policy]
https://drive.google.com/open?id=1Uh31D ... TDHfqrMoDT


AF 3-14500 Mark 183 [Focused Mse Policy Value]
https://drive.google.com/open?id=1brqRj ... z8Vh3rWxCv
"Good decisions come from experience, and experience comes from bad decisions."
__________________________________________________________________
Ted Summers
Eduard
Posts: 1439
Joined: Sat Oct 27, 2018 12:58 am
Location: Germany
Full name: N.N.

Re: Antifish vs Stockfish in cccc

Post by Eduard »

Thank you!

Very interesting match
https://www.chess.com/computer-chess-championship

What do the lc0 experts think about this hardware?

GPU: 4x RTX 2080ti (44 GB GPU memory)
CPU: 2x Intel Xeon Gold 6154
Cores: 36 physical
RAM: 96 GB
Werewolf
Posts: 1795
Joined: Thu Sep 18, 2008 10:24 pm

Re: Antifish vs Stockfish in cccc

Post by Werewolf »

Eduard wrote: Mon Jan 21, 2019 8:33 pm Thank you!

Very interesting match
https://www.chess.com/computer-chess-championship

What do the lc0 experts think about this hardware?

GPU: 4x RTX 2080ti (44 GB GPU memory)
CPU: 2x Intel Xeon Gold 6154
Cores: 36 physical
RAM: 96 GB
Not an expert on Lc0 but: The CPU core count doesn't really help (12 cores would be just as good I think) and I wasn't aware Lc0 could use more than 2 cards, but maybe they've made improvements.
Eduard
Posts: 1439
Joined: Sat Oct 27, 2018 12:58 am
Location: Germany
Full name: N.N.

Re: Antifish vs Stockfish in cccc

Post by Eduard »

Game 18:
[pgn][Event "Stockfish vs Antifish (20|2)"] [Site "?"] [Date "2019.01.21"] [Round "18"] [White "Stockfish"] [Black "Antifish"] [Result "1-0"] [TimeControl "1200+2"] [ECO "C19"] [GameDuration "00:41:00"] [GameEndTime "2019-01-21T23:07:26.834 PST"] [GameStartTime "2019-01-21T22:26:26.748 PST"] [Opening "French"] [PlyCount "127"] [Variation "Winawer, Advance, poisoned pawn Variation"] 1. e4 e6 2. d4 d5 3. Nc3 Bb4 4. e5 c5 5. a3 Bxc3+ 6. bxc3 Ne7 7. Qg4 cxd4 8. Qxg7 Rg8 9. Qxh7 Qc7 10. Ne2 dxc3 11. f4 Bd7 12. Qd3 Na6 13. Nd4 Qa5 14. Nb3 Qa4 15. g3 Rc8 16. Be3 b6 17. Nd4 Nb4 18. Qh7 Rg6 19. Rc1 Nbc6 20. Bb5 Qa5 21. Bxc6 Nxc6 22. Qh8+ Ke7 23. Nxc6+ Rxc6 24. Kf2 Rc4 25. Qh4+ Ke8 26. Qh8+ Ke7 27. Qb8 f5 28. exf6+ Rxf6 29. Rhd1 Re4 30. Qg8 Qa4 31. Rd3 Rf7 32. Qg5+ Ke8 33. Qh6 Ke7 34. Qg5+ Ke8 35. Qh6 Bc6 36. h4 Qxa3 37. Rcd1 Kd7 38. Qh8 Rc4 39. Qe5 Qb2 40. R3d2 Kc8 41. Qxe6+ Rd7 42. Re2 Kb7 43. Bd4 Qb4 44. Qe5 a5 45. g4 a4 46. g5 a3 47. g6 a2 48. g7 Rxd4 49. Rxd4 Qb1 50. g8=Q a1=Q 51. Qg2 Ka6 52. Qe6 Kb7 53. f5 Qa3 54. Kg3 Rc7 55. f6 Qac1 56. f7 Qa3 57. Qf2 Qac1 58. f8=Q Qf4+ 59. Q2xf4 Qg1+ 60. Rg2 Qxg2+ 61. Kxg2 Ka6 62. Qxc7 Kb5 63. Qcxc6+ Ka5 64. Qxb6# 1-0[/pgn]

5 Queens on the board:

[d]5Q2/1kr5/1pb1Q3/3p4/3R3P/2p3K1/2P1RQ2/1qq5 b - - 0 58

funny, funny.

Standing after 18 games 9-9.
peter
Posts: 3185
Joined: Sat Feb 16, 2008 7:38 am
Full name: Peter Martan

Re: Antifish vs Stockfish in cccc

Post by peter »

Hi!
Nay Lin Tun wrote: Mon Jan 21, 2019 8:28 am ...Antifish ( Leela derivative training with Stockfish games , rather than self played games like Leela)...
"Rather" means trained with games against SF only (how many in which hardware- time?) or with SF- games added to an existing LC0- network?
Peter.