test position

[kgburcham]

25.g4


[D] 3r2k1/p1r1npp1/qp2p2p/4P2P/3PN3/P2R1Q2/5PP1/3R2K1 w - - 0 25

kgburcham

[yanquis1972]

is this too difficult for today's engines to find at reasonable time? even after playing g4 the best i have to offer practically go haywire for a bit.

[Werewolf]

Is there any proof g4 is best?
even after 2 hours on an overclocked quad Rybka 3 doesn't get it

[kgburcham]

the g4 move was part of a Kingside attack. The move was made in a white winning game, so I wanted to see if black blundered for the white win. I also wanted to see what Rybka though about the move--which it will play, and also what rybka thinks about the line played after g4. In this game the g4 move seems convincing after playing through the game moves.

?
kgburcham

[D] 3r2k1/p1r1npp1/qp2p2p/4P2P/3PN3/P2R1Q2/5PP1/3R2K1 w - - 0 25

[Event "XIII CC Olympiad ICCF, 29.04.2"]
[White "Baumbach,Dr. F"]
[Black "Valent,P"]
[Result "1-0"]

1. d4 Nf6 2. c4 e6 3. Nf3 b6 4. Nc3 Bb7 5. a3 d5
6. cxd5 Nxd5 7. Qc2 Nxc3 8. bxc3 Be7 9. e4 O-O 10. Bd3 c5 11. O-O Qc8 12. Qe2
Ba6 13. Rd1 Bxd3 14. Rxd3 Nd7 15. e5 cxd4 16. cxd4 Re8 17. Bg5 Qa6 18. Bxe7
Rxe7 19. Rad1 Nf8 20. Ng5 Rc7 21. Qf3 Rd8 22. h4 h6 23. Ne4 Ng6 24. h5 Ne7 25.
g4 Qb7 26. g5 hxg5 27. Qg4 Rc4 28. Nd6 Rxd6 29. exd6 Nf5 30. Qxg5 Qd7 31. h6 f6
32. Qg6 Nxh6 33. Rh3 Nf7 34. Qh7+ Kf8 35. Rg3 Ng5 36. Qd3 b5 37. f4 Nf7 38. Qh7
g5 39. fxg5 Qxd6 40. Rg4 Qxa3 41. g6 Qe3+ 42. Kh1

Rybka 3:
Hash 2048

25.Nf6+ Kf8 26.d5 Nxd5 27.Nxd5 exd5 28.Rxd5[] Rxd5 29.Qxd5 Qc8 30.Kh2 Re7 31.Qe4 Re8 32.a4 Qe6 33.Rd5
= (0.13) Depth: 12 00:00:01 572kN
25.Nf6+ Kf8 26.d5 Nxd5 27.Nxd5 exd5 28.Rxd5[] Rxd5 29.Qxd5 Qc8 30.Kh2 Re7 31.Qe4 Re8 32.a4 Qe6 33.Rd5
= (0.13) Depth: 13 00:00:03 1017kN
25.Nf6+ Kf8 26.d5 Nxd5 27.Nxd5 Rxd5 28.Rxd5 exd5 29.Qxd5 Qc8 30.Kh2 Re7 31.Qe4 Re8 32.a4 Qe6 33.Rd5 Kg8 34.Qd4
= (0.11) Depth: 14 00:00:08 2315kN
25.Nf6+ Kf8 26.d5 Nxd5 27.Nxd5 Rxd5 28.Rxd5[] exd5 29.Qxd5 Qc8 30.Kh2 Re7 31.Qe4 Re8 32.a4 Qe6 33.Rd5 Kg8 34.Qd4
= (0.11) Depth: 15 00:00:11 3328kN
25.Nf6+ Kf8 26.d5 Nxd5 27.Nxd5 Rxd5 28.Rxd5[] exd5 29.Qxd5 Qc8 30.Kh2 Re7 31.Qe4 Re8 32.a4 Kg8 33.Rd5 Qc1[] 34.f4 Qc8[]
= (0.09) Depth: 16 00:00:17 5411kN
25.d5 Nxd5 26.Qg4 Kh8 27.Rg3 Rg8 28.Nf6 Rc4 29.Qf3 Qa4 30.Nxg8 Kxg8 31.Qd3 Rc3 32.Qd4 Rxa3 33.Rxa3 Qxa3 34.Ra1 Qe7 35.Rc1 Qf8 36.Qa4 a5 +/= (0.28) Depth: 16 00:00:23 7483kN
25.Nf6+ Kf8 26.d5 Nxd5 27.Nxd5 Rxd5 28.Rxd5[] exd5 29.Qxd5 Qc8 30.Qd8+ Qxd8[] 31.Rxd8+ Ke7[] 32.Rg8 Rc5 33.f4 Rc4 34.g3 Rc3 35.Kf2 Rxa3 36.Rxg7 b5 37.Rh7 b4 38.Rxh6 b3 39.Rc6 a5 40.Rc7+
= (0.17) Depth: 17 00:00:50 16250kN
25.Nf6+ Kf8 26.d5 Nxd5 27.Nxd5 Rxd5 28.Rxd5 exd5 29.Qxd5 Qc8 30.Kh2 Re7 31.Qe4 Re8 32.a4 Kg8 33.Rd5 Qe6[] 34.Kg3 Qc6[] 35.f4 Qc3+[] 36.Rd3 Qc2[]
= (0.10) Depth: 18 00:02:07 41015kN
25.g4 Qa4 26.g5 hxg5 27.h6 Nd5
+/= (0.28) Depth: 18 00:04:13 73362kN
25.Kh2
+/= (0.37 !) Depth: 18 00:05:08 87037kN
25.Kh2 Qa4 26.Qg4 Kh8 27.Nd6 Nd5 28.Rc1 f5 29.Qg6 Nf4[] 30.Rxc7[] Nxg6[] 31.Nf7+[] Kg8[] 32.hxg6 Rf8 33.Rh3
+/= (0.36) Depth: 18 00:05:21 90872kN
25.Kh2 Qa4 26.Qg4 Kh8 27.Nd6 Nc8 28.Rg3 f5
+/= (0.28) Depth: 19 00:05:49 98610kN
25.g4 Qa4 26.g5 hxg5 27.h6 Nd5 28.Nxg5 g6[] 29.h7+ Kg7[] 30.Kg2 Qc6 31.Kh2 Rdc8 32.Rh1 a6 33.Qg4 Kh8 34.Rf3 Qe8 35.Kg2 b5
+/= (0.28) Depth: 20 00:13:49 242mN
25.g4 Qa4 26.g5 hxg5 27.h6 Nd5 28.Nxg5 g6[] 29.Rb1 Qd7 30.h7+ Kh8 31.Qg4 Rf8 32.Rf3 Qe7
+/= (0.31) Depth: 21 00:26:20 464mN
25.g4 Qa4 26.g5 hxg5 27.h6 Nd5 28.Nxg5 g6[] 29.Rb1 Qd7 30.h7+ Kh8 31.Qg4 Rf8 32.Rf3 Qe7 33.Rbb3 Rcc8 34.Qh4 Kg7 35.a4 Kh8 36.Kg2 Qd7
+/= (0.31) Depth: 22 00:37:54 675mN
25.g4 Qa4 26.g5 hxg5 27.h6 Nd5 28.Nxg5 g6[] 29.Rb1 Qd7 30.h7+ Kh8 31.Qg4 Rf8 32.Rf3 Qe7 33.Rbb3 Rcc8 34.Qh4 Kg7 35.a4 Kh8 36.Kg2 Qd7
+/= (0.33) Depth: 23 01:07:25 1221mN

[beachknight]

1561: 3r2k1/p1r1npp1/qp2p2p/4P2P/3PN
3r2k1/p1r1npp1/qp2p2p/4P2P/3PN3/P2R1Q2/5PP1/3R2K1 w - - 0 1

Analysis by Deep Fritz 11:

+/= (0.43) Depth: 19/34 00:00:26 194mN
25.Qf4 Qc8 26.Qg4 Rc1 27.Nd6 Rxd1+ 28.Qxd1 Qd7 29.Qd2 Nd5 30.Rg3 Kh8 31.Qc2 Rf8 32.Rg4
+/= (0.46) Depth: 20/37 00:00:40 306mN
25.Qf4 Qc8 26.Qg4 Rc6 27.Nd6 Rcxd6 28.exd6 Rxd6 29.Qf4 Qd7 30.Rc1 Nc6 31.Qe4 Rxd4 32.Rcc3
+/= (0.41) Depth: 21/38 00:01:01 467mN
25.Qf4 Qc8 26.Qg4 Rc6 27.Nd6 Rcxd6 28.exd6 Rxd6 29.Re1 Nf5 30.Qf4 Qd7 31.Re4 Rc6
+/= (0.29) Depth: 22/45 00:01:40 781mN
25.Kh2 Qa4 26.Qg4 b5 27.Re1 Qc4 28.Rc3 Qxc3 29.Nxc3 Rxc3 30.Rb1 a6 31.a4 bxa4 32.d5
+/= (0.49) Depth: 22/45 00:02:41 1257mN
25.Kh2 Qa4 26.Qg4 b5 27.Re1 Qc4 28.Rc3 Qxc3 29.Nxc3 Rxc3 30.Qe2 Rc4 31.a4 Rxa4 32.Qxb5 Raxd4
+/= (0.52) Depth: 23/40 00:03:56 1858mN
25.Kh2 Qa4 26.Qg4 b5 27.Re1 Qc4 28.Rc3 Qxc3 29.Nxc3 Rxc3 30.Qe2 Rc4 31.a4 a6 32.axb5 axb5
+/= (0.50) Depth: 24/41 00:06:18 3011mN
25.Kh2 Qa4 26.Qg4 b5 27.d5 exd5 28.Rg3 Nf5 29.Rxd5 Rxd5 30.Qxf5 Qd4 31.Nf6+ Kf8 32.Nxd5 Qxd5 33.Rd3
+/= (0.39) Depth: 25/45 00:11:32 5550mN
25.g4 Rf8 26.g5 hxg5 27.Nxg5 Qb5 28.Qf4 Nf5 29.Ne4 Kh8 30.Nd6 Nxd6 31.exd6 Rd7 32.h6 Rg8 33.Rh3
+/- (0.72) Depth: 25/49 00:31:53 15509mN

(, chessengines20 13.07.2009)

[lmader]

Analysis by Deep Fritz 11:

25.Qf4 Qc8 26.Qg4 Rc6 27.Nd6 Rcxd6 28.exd6 Rxd6 29.Re1 Nf5 30.Qf4 Qd7 31.Re4 Rc6
+/= (0.29) Depth: 22/45 00:01:40 781mN
25.Kh2 Qa4 26.Qg4 b5 27.d5 exd5 28.Rg3 Nf5 29.Rxd5 Rxd5 30.Qxf5 Qd4 31.Nf6+ Kf8 32.Nxd5 Qxd5 33.Rd3
+/= (0.39) Depth: 25/45 00:11:32 5550mN
25.g4 Rf8 26.g5 hxg5 27.Nxg5 Qb5 28.Qf4 Nf5 29.Ne4 Kh8 30.Nd6 Nxd6 31.exd6 Rd7 32.h6 Rg8 33.Rh3
+/- (0.72) Depth: 25/49 00:31:53 15509mN

That's an impressive performance by Deep Fritz 11.
What are the specs of the hardware on which you are running it?

[lmader]

Analysis by Deep Fritz 11:

25.Qf4 Qc8 26.Qg4 Rc6 27.Nd6 Rcxd6 28.exd6 Rxd6 29.Re1 Nf5 30.Qf4 Qd7 31.Re4 Rc6
+/= (0.29) Depth: 22/45 00:01:40 781mN
25.Kh2 Qa4 26.Qg4 b5 27.d5 exd5 28.Rg3 Nf5 29.Rxd5 Rxd5 30.Qxf5 Qd4 31.Nf6+ Kf8 32.Nxd5 Qxd5 33.Rd3
+/= (0.39) Depth: 25/45 00:11:32 5550mN
25.g4 Rf8 26.g5 hxg5 27.Nxg5 Qb5 28.Qf4 Nf5 29.Ne4 Kh8 30.Nd6 Nxd6 31.exd6 Rd7 32.h6 Rg8 33.Rh3
+/- (0.72) Depth: 25/49 00:31:53 15509mN

That's an impressive performance by Deep Fritz 11.
What are the specs of the hardware on which you are running it?

On my Intel dual core laptop, I noticed that bumping the hashtable size of DF11 from 2048mb to the practical limit on my machine of 2335mb made a HUGE difference in the search time. It found Qf4, Kh2, and then G4 in a fraction of the time needed at 2048mb of hashtable. Anyone else notice this sort of thing with a seemingly modest change in hashtable size?

[ernest]

It found Qf4, Kh2, and then G4 in a fraction of the time needed at 2048mb of hashtable. Anyone else notice this sort of thing with a seemingly modest change in hashtable size?

Don't forget this is multiprocessor, so there is no reproducibility (you can get anything...).

But even if this happens on single core, nobody has been able to give a theory about the relation between hash size and solution time (or time to get the PV at depth n). For the observer, it just looks unpredictable.

[lmader]

Ahhh, thanks for that info. I am aware of the non-deterministic effect that multiprocessor chess algorithms exhibit, but I didn't know that hashtable size was also so difficult to quantify with regard to the search.

I repeated the test many times with both hashtable sizes, being careful to both clear the hash and exit the program after each run. I did notice that the search looked consistently different between the two hash sizes: The ply depth at which the moves were found was always fewer with the larger hashtable, and it got there much faster, as I mentioned.

Anyway :p