Stockfish 10 beta

[Eduard]

Yes!

The new Stockfish seems to be real very good. I played on playchess.com (account "SolistaChess") only with a Laptop Dual core 2,4 GHz against much stronger Computers! Have not lost last 13 games. Very funny!

Played first with Version 121118, then Beta 10, and last 2 games with 201118. Here as example the last game, against on of the highest rated computers on playchess:

[Event "Wertungspartie, 5 min"]
[Site "Maschinenraum"]
[Date "2018.11.22"]
[Round "?"]
[White "IBM, Toltec Omega"]
[Black "SolistaChess, Stockfish 201118 64"]
[Result "1/2-1/2"]
[ECO "A28"]
[WhiteElo "2585"]
[BlackElo "2419"]
[Annotator "?ev?;0.10"]
[PlyCount "119"]
[EventDate "2018.11.22"]
[SourceTitle "playchess.com"]
[TimeControl "300"]

{Toltec Omega (24 threads): 34.9 plies; 28.684kN/s Intel(R) Xeon(R) CPU
X5650 @ 2.67GHz 2660MHz, (12 cores, 24 threads), enMatch.ctg, 128 MB} 1. c4 {0}
e5 {B 0} 2. Nc3 {0} Nf6 {B 0} 3. Nf3 {0} Nc6 {B 0} 4. e4 {0} Bb4 {B 0} 5. d3 {0
} d6 {B 0} 6. a3 {0} Bxc3+ {B 0} 7. bxc3 {0} Ne7 {B 0} 8. g3 {0} h6 {0.10/23 24
} 9. Nh4 {0 (Lg2)} a6 {0.04/20 5} 10. Rb1 {0} Rb8 {0.18/22 9} 11. Bg2 {8 (a4)}
b5 {0.09/23 8} 12. cxb5 {13} Rxb5 {0.31/23 0} 13. Rxb5 {10} axb5 {0.27/22 0}
14. O-O {3} Qd7 {0.08/20 2} 15. f4 {10} O-O {0.00/23 0} 16. fxe5 {14 (h3)} dxe5
{-0.09/22 2} 17. Bxh6 {2} Rd8 {0.00/26 2} 18. Bg5 {0} Qxd3 {0.00/30 12} 19.
Qxd3 {0} Rxd3 {0.08/31 6} 20. Rc1 {0} Rd7 {0.00/30 5} 21. Nf5 {6} Nxf5 {
0.00/33 0} 22. exf5 {3} Rd6 {0.00/34 3} 23. Bxf6 {3} gxf6 {0.00/34 2} 24. g4 {2
} Bd7 {0.00/33 1} 25. Bf1 {7 (c4) SolistaChess,Stockfish 201118 64 bietet Remis
} Rd2 {0.00/34 5} 26. c4 {3 (Lg2)} bxc4 {0.00/31 3} 27. Bxc4 {1} Ba4 {0.00/33 4
} 28. h3 {0 (Lf1)} Kg7 {0.00/33 16} 29. Bf1 {6 (Tc3)} c6 {0.00/33 8} 30. Rc3 {0
} Kh6 {0.00/33 2} 31. h4 {0} Rb2 {0.00/34 8} 32. Bd3 {4} Bd1 {0.00/35 0} 33.
Rxc6 {4} Bxg4 {0.00/37 0} 34. Rxf6+ {4} Kg7 {0.00/37 0} 35. Rc6 {4} Bf3 {
0.00/37 0} 36. f6+ {3} Kh6 {0.00/38 0} 37. Rc4 {3 (Tc2)} Rb3 {0.00/33 5} 38.
Bf5 {0} Rxa3 {0.00/36 4} 39. Rc7 {4} Bd5 {0.00/37 0} 40. Rc8 {3 (Te7)} Ra8 {
0.00/36 4} 41. Rxa8 {37 (Tc5)} Bxa8 {0.00/51 4} 42. Kf2 {25} Bd5 {0.00/64 0}
43. Bc2 {5 (Ke3)} Kh5 {0.00/46 4} 44. Kg3 {5} e4 {0.00/46 0} 45. Kf4 {2} Kxh4 {
0.00/66 1} 46. Bxe4 {22} Bxe4 {0.00/97 0} 47. Kxe4 {0} Kg4 {0.00/0 0} 48. Ke5 {
1} Kg5 {0.00/0 0} 49. Ke4 {0} Kxf6 {0.00/0 0} 50. Kf4 {0} Ke6 {0.00/0 0} 51.
Ke4 {0} f5+ {0.00/0 0} 52. Kd3 {1} Ke5 {0.00/0 0} 53. Ke3 {0} f4+ {0.00/0 0}
54. Kd2 {0} Ke4 {0.00/0 0} 55. Ke2 {0} f3+ {0.00/0 0} 56. Kf1 {0} Ke3 {0.00/0 0
} 57. Ke1 {0} Ke4 {0.00/0 0} 58. Kf1 {1} Ke3 {0.00/0 0} 59. Ke1 {0} Ke4 {
0.00/0 0} 60. Kf1 {1} 1/2-1/2

[corres]

Contempt has value when you have an inferior opponent. Suppose, for instance, that you come out of an opening book and your opponent is a little bit ahead of you (perhaps a quarter of a pawn). If you have an equal opponent, and you can find nothing better, it is a good idea to strive for a draw.

But suppose that your opponent is much weaker than you are. Now it is probably not a good idea to strive for a draw because you will probably win the game if you play on and avoid a draw.

So if (for instance) you have contempt on during self-play, you will see absurd things like both sides being ahead at the same time, according to the evaluation.

Unfortunately, with the standard Stockfish, you cannot turn off all of the contempt, but only part of it.

So I modify Stockfish so that I can use it for analysis and get the right answer.

Now, there is clearly a real affect of having contempt when it comes to game play. So it is a good idea for a chess engine. But when analyzing, it is telling you a small lie, which is bad.

IMO-YMMV

You are right and I did it also for better analysis.

[Hurnavich]

Played 3 matches with the latest development version against SF9

Code: Select all

  Match      games    Time   Perc  ELO
sf10 - sf9  10,000  15+0.1  60.2%  +72
sf10 - sf9  10,000  30+0.2  59.9%  +69
sf10 - sf9  10,000  60+0.4  59.5%  +67

Impressive!

Any chance for a link to the games? thanks

[Jouni]

What was this SF 10 beta thing? No release so far. Nothing is said in fishcooking forum .

[Graham Banks]

What was this SF 10 beta thing? No release so far. Nothing is said in fishcooking forum .

Author: Stéphane Nicolet
Date: Mon Nov 19 11:18:21 2018 +0100
Timestamp: 1542622701

Stockfish 10-beta

Preparation commit for the upcoming Stockfish 10 version, giving a chance to catch last minute feature bugs and evaluation regression during the one-week code freeze period. Also changing the copyright dates to include 2019.

No functional change

[Werner]

From our CEGT 40/4 list:
1 Stockfish 20181025 x64 1CPU 3449 +15 -15 2000 games
2 Stockfish 9.0 x64 1CPU 3389

[schack]

Any idea on when SF10 is going to be officially released?

[stavros]

Any idea on when SF10 is going to be officially released?

"...Also changing the copyright dates to include 2019"

expect it the next year.they used to be lazy.

[Eelco de Groot]

Any idea on when SF10 is going to be officially released?

At some point, the codebase will be frozen so that there can be some final testing. But right now some commits are still added, others are postponed for after the release. Some final checking will have to be done. For instance, I read somewhere that the makefile seems to not support 32 bit makes anymore. It has to be decided if that is still supportable. I think it was somewhere in Uri's thread about his project in the programming subforum.

[Scally]

For info, it still compiles on the 32bit Raspberry Pi O.K.

Al.