[d] N4rk1/1b1p3p/3bpn2/pn3p2/2P1P1p1/P2P2Pq/1B1QBP1P/R4RK1 b - -
Black won this interesting correspondence game despite being down a lot of material at one point.
bm from this position is f4. Rybka 3 gets this fairly rapidly but Arasan takes a long time (over 10 minutes on my Athlon dual core).
--Jon
Muri-Peixoto, corr EU-ch M1231 2001
Moderator: Ras
-
jdart
- Posts: 4406
- Joined: Fri Mar 10, 2006 5:23 am
- Location: http://www.arasanchess.org
-
royb
- Posts: 566
- Joined: Thu Mar 09, 2006 12:53 am
Re: Muri-Peixoto, corr EU-ch M1231 2001
It may be my underpowered hardware, but Rybka 3 on my system only shows f4 with a 0.00 eval. As I played out the suggested moves a few more ply, the 0.00 eval sticks, indicating that Rybka 3 thinks this line is only good for a draw.
What is the winning line?
What is the winning line?
-
jdart
- Posts: 4406
- Joined: Fri Mar 10, 2006 5:23 am
- Location: http://www.arasanchess.org
Re: Muri-Peixoto, corr EU-ch M1231 2001
I didn't say this particular move was winning, only that it was best.
--Jon
--Jon
-
Anil
- Posts: 540
- Joined: Sat Nov 11, 2006 2:56 pm
Re: Muri-Peixoto, corr EU-ch M1231 2001
Can you post the game here? It would be interesting to see the winning moves.
-
jdart
- Posts: 4406
- Joined: Fri Mar 10, 2006 5:23 am
- Location: http://www.arasanchess.org
Re: Muri-Peixoto, corr EU-ch M1231 2001
This is the game:
[Event "EU-ch M1231 corr"]
[Site "ICCF corr"]
[Date "2001.??.??"]
[Round "?"]
[White "Muri, Heinrich"]
[Black "Peixoto, Antonio"]
[Result "0-1"]
[ECO "A00"]
1. b4 b6 2. Bb2 Bb7 3. e3 e6 4. Nf3 f5 5. Be2 Nf6 6. a3 a5 7. b5 Be7 8. O-O O-O
9. d3 Qe8 10. Nd4 c5 11. bxc6 Nxc6 12. Nb5 Qg6 13. Bf3 Qh6 14. c4 g5 15. N1c3
g4 16. Be2 Nd8 17. Na4 Nf7 18. Nxb6 Nd6 19. Nxa8 Nxb5 20. Qd2 Bd6 21. g3 Qh3
22. e4 f4 23. Rfd1 fxg3 24. Qg5+ Kf7 25. hxg3 Bxg3 26. Qxf6+ Ke8 27. Qxf8+ Kxf8
28. fxg3 Qxg3+ 29. Kf1 Nd6 30. Nb6 Nf5 31. Bc1 Ke8 32. Rb1 Bc6 33. Rb2 h5 34.
Nd5 exd5 35. cxd5 Ba4 36. exf5 Bxd1 37. Bxd1 h4 38. Re2+ Kf7 39. Bg5 h3 40. d6
Qxd3 41. Bc2 Qf3+ 42. Ke1 Qg3+ 43. Kf1 h2 44. Re7+ Kf8 45. Be4 Qh3+ 46. Ke1
h1=Q+ 47. Bxh1 Qxh1+ 48. Kd2 Qd5+ 49. Ke3 Qf3+ 50. Kd4 Qxf5 51. Re5 Qf2+ 52.
Be3 Qf6 53. Ke4 Qf3+ 54. Kd3 g3 55. Rxa5 Qd1+ 56. Ke4 Qg4+ 57. Kd3 g2 58. Rd5
0-1
[Event "EU-ch M1231 corr"]
[Site "ICCF corr"]
[Date "2001.??.??"]
[Round "?"]
[White "Muri, Heinrich"]
[Black "Peixoto, Antonio"]
[Result "0-1"]
[ECO "A00"]
1. b4 b6 2. Bb2 Bb7 3. e3 e6 4. Nf3 f5 5. Be2 Nf6 6. a3 a5 7. b5 Be7 8. O-O O-O
9. d3 Qe8 10. Nd4 c5 11. bxc6 Nxc6 12. Nb5 Qg6 13. Bf3 Qh6 14. c4 g5 15. N1c3
g4 16. Be2 Nd8 17. Na4 Nf7 18. Nxb6 Nd6 19. Nxa8 Nxb5 20. Qd2 Bd6 21. g3 Qh3
22. e4 f4 23. Rfd1 fxg3 24. Qg5+ Kf7 25. hxg3 Bxg3 26. Qxf6+ Ke8 27. Qxf8+ Kxf8
28. fxg3 Qxg3+ 29. Kf1 Nd6 30. Nb6 Nf5 31. Bc1 Ke8 32. Rb1 Bc6 33. Rb2 h5 34.
Nd5 exd5 35. cxd5 Ba4 36. exf5 Bxd1 37. Bxd1 h4 38. Re2+ Kf7 39. Bg5 h3 40. d6
Qxd3 41. Bc2 Qf3+ 42. Ke1 Qg3+ 43. Kf1 h2 44. Re7+ Kf8 45. Be4 Qh3+ 46. Ke1
h1=Q+ 47. Bxh1 Qxh1+ 48. Kd2 Qd5+ 49. Ke3 Qf3+ 50. Kd4 Qxf5 51. Re5 Qf2+ 52.
Be3 Qf6 53. Ke4 Qf3+ 54. Kd3 g3 55. Rxa5 Qd1+ 56. Ke4 Qg4+ 57. Kd3 g2 58. Rd5
0-1
-
Eelco de Groot
- Posts: 4673
- Joined: Sun Mar 12, 2006 2:40 am
- Full name: Eelco de Groot
Re: Muri-Peixoto, corr EU-ch M1231 2001
Ancalagon after a very deep search is not so convinced anymore that Black has a draw at most after 22... f4 but maybe it sees mirages flying over the horizon. Also in this version I was experimenting with QSearch beta cut-offs;jdart wrote:I didn't say this particular move was winning, only that it was best.
--Jon
Build 6 has a cutoff value at the beginning of QSearch of 200. I tried 0, 100, - that version finds Rfd1 instead of Rfc1 -, and beta + 50 +depth, beta + 150 + depth in this position, with depth <= 0 because it is QSearch.
I was surprised values > beta had no effect whatsoever even on nodecounts, so I tried other values. Values 0, 100, 200 at least did change the search.
Code: Select all
// Initialize "stand pat score", and return it immediately if it is
// at least beta.
if(pos.is_check())
bestValue = -VALUE_INFINITE;
else {
bestValue = staticValue;
if(bestValue >= 200) // was beta + 250 + depth)
return bestValue;
if(bestValue > alpha)
alpha = bestValue;
[d]N4rk1/1b1p3p/3bpn2/pn3p2/2P1P1p1/P2P2Pq/1B1QBP1P/R4RK1 b - -
Engine: Ancalagon 1.1 Beta 1 (256 MB)
by Tord Romstad, Eelco de Groot
8.00 0:00 -4.05 1...Nxe4 2.dxe4 Bxe4 3.f3 Bxg3 4.Bd1 Qxh2+
5.Qxh2 Bxh2+ 6.Kxh2 g3+ 7.Kg2 Rxa8 (205.909) 239
9.01 0:01 -4.25 1...Nxe4 2.dxe4 Bxe4 3.f3 Bxg3 4.Bd1 gxf3
5.cxb5 Qg2+ 6.Qxg2 fxg2 7.hxg3 gxf1Q+
8.Kxf1 Bd3+ 9.Kf2 Bxb5 (318.444) 261
9.02 0:02 -2.37 1...f4 2.cxb5 Nxe4 3.Bxg4 fxg3
4.hxg3 Qxf1+ 5.Rxf1 Nxd2 6.Nb6 Bxg3
7.Nxd7 (579.831) 279
10.01 0:04 -1.47 1...f4 2.Bxf6 Rxf6 3.Bxg4 Qxg4
4.cxb5 Bxa8 5.Qc3 Rf8 6.f3 Qh3 7.g4 Bxa3 (1.181.835) 292
11.01 0:17 -4.07 1...f4 2.Rfc1 Nxa3 3.Bxa3 Bxa3
4.Rxa3 fxg3 5.fxg3 Rxa8 6.Rxa5 Rxa5
7.Qxa5 d6 8.Rf1 (5.255.243) 306
11.46 0:26 0.00 1...Nxe4 2.dxe4 Bxe4 3.f3 Bxg3
4.hxg3 Qxg3+ 5.Kh1 Qh3+ 6.Kg1 Qg3+
7.Kh1 (8.196.975) 310
12.01 0:28 0.00 1...Nxe4 2.dxe4 Bxe4 3.f3 Bxg3
4.hxg3 Qxg3+ 5.Kh1 Qh3+ 6.Kg1 Qg3+
7.Kh1 (8.976.113) 309
13.01 0:35 -0.29 1...Nxe4 2.dxe4 Bxe4 3.f3 Bxg3 4.Bd3 Bxd3
5.cxb5 Bxb5 6.hxg3 Qxg3+ 7.Qg2 Qxg2+
8.Kxg2 Bxf1+ 9.Rxf1 Rxa8 10.fxg4 Rb8
11.Be5 (10.852.472) 308
14.01 0:57 -3.13 1...Nxe4 2.dxe4 Bxe4 3.f3 Bxg3 4.Bd1 Bd3
5.Ba4 Bxc4 6.Nb6 Qxh2+ 7.Qxh2 Bxh2+
8.Kxh2 Bxf1 9.Rxf1 Nxa3 10.Bxa3 Kf7 (17.477.787) 304
15.01 3:25 -4.45 1...Nxe4 2.dxe4 Bxe4 3.f3 Bxg3 4.Bd1 Bd3
5.Ba4 Qxh2+ 6.Qxh2 Bxh2+ 7.Kxh2 Nxa3
8.Bxa3 Rxa8 9.Rfd1 gxf3 10.Kg1 f2+
11.Kxf2 Bxc4 (64.425.574) 313
15.02 5:21 -1.15 1...f4 2.Rfc1 Nxe4 3.dxe4 Bxe4 4.Bf1 fxg3
5.Qg5+ Kf7 6.fxg3 Qh6 7.h4 Qxg5
8.hxg5 Nxa3 9.Nc7 Bxc7 10.Rxa3 (100.529.040) 312
16.01 6:53 -1.19 1...f4 2.Rfc1 Nxe4 3.dxe4 Bxe4 4.Bf1 fxg3
5.Qg5+ Kf7 6.fxg3 Qh6 7.h4 Qxg5
8.hxg5 Nxa3 9.Nc7 Bxc7 10.Rxa3 h5 (129.247.143) 312
17.01 12:11 -1.21 1...f4 2.Rfc1 Nxe4 3.dxe4 Bxe4 4.Bf1 fxg3
5.Qg5+ Kf7 6.fxg3 Qh6 7.h4 Qxg5
8.hxg5 Bc5+ 9.Kh2 Kg8 10.Bg2 Nxa3
11.Bxa3 Bxg2 12.Bxc5 Bxa8 (228.866.462) 312
With cutoff value 100 Ancalagon was faster but that is probably only in this position. White is still doing better in Ancalagon's eyes, here at depth 17 and other programs also show a plus for White but could only do a shallow search.
After 22... f4 programs apparently have to decide where to move the f1 Rook, other moves are not liked, but exactly where the Rook should go opinion on that seems to vary a lot during the search, also with other programs.
These were all running in parallel with each other and Ancalagon on the Athlon 2009 MHz single CPU:
N4rk1/1b1p3p/3bpn2/pn6/2P1Ppp1/P2P2Pq/1B1QBP1P/R4RK1 w - -
Engine: Toga II Checkov Beta_4/ Mara Beta 6 (64 MB)
by Fabien Letouzey, Thomas Gaksch
10/47 0:33 +0.56 1.Rfb1 Nxe4 2.dxe4 Bxe4 3.Bf1 fxg3
4.Qg5+ Kf7 5.fxg3 Qh6 6.h4 Qxg5
7.hxg5 Bc5+ 8.Kh2 Bxb1 9.cxb5 Be4 (6.735.985) 201
11/47 0:47 +0.28 1.Rfb1 Nxe4 2.dxe4 Bxe4 3.Bf1 fxg3
4.Qg5+ Kf7 5.fxg3 Qh6 6.h4 Qxg5
7.hxg5 Bxb1 8.cxb5 Be4 9.Rd1 Bc5+
10.Kh2 Rxa8 11.Rxd7+ Kg6 (9.599.627) 202
12/49 2:32 +0.28 1.Rfb1 Nxe4 2.dxe4 Bxe4 3.Bf1 fxg3
4.Qg5+ Kf7 5.fxg3 Qh6 6.h4 Qxg5
7.hxg5 Bxb1 8.cxb5 Be4 9.Rd1 Bc5+
10.Kh2 Rxa8 11.Rxd7+ Kg6 (30.347.242) 198
13/49 4:36 0.00 1.Rfb1 Nxe4 2.dxe4 Bxe4 3.Bf1 fxg3
4.Qg5+ Kf7 5.fxg3 Qh6 6.h4 Qxg5
7.hxg5 Bc5+ 8.Kh2 Nd4 9.Rd1 Nf3+
10.Kh1 Ne1+ 11.Kh2 Nf3+ (52.144.642) 188
14/50 15:30 +0.50 1.Rfb1 Nxe4 2.dxe4 Bxe4 3.Bf1 fxg3
4.Qg5+ Kf7 5.fxg3 Qh6 6.h4 Qxg5
7.hxg5 Bc5+ 8.Kh2 Nd4 9.Bxd4 Bxd4
10.Nc7 Bxa1 11.Rxa1 Ke7 (153.730.726) 165
15/52 34:48 +0.51 1.Rfb1 Nxe4 2.dxe4 Bxe4 3.Bf1 fxg3
4.Qg5+ Kf7 5.fxg3 Qh6 6.h4 Qxg5
7.hxg5 Bc5+ 8.Kh2 Nd4 9.Bxd4 Bxd4
10.Nc7 Bxa1 11.Rxa1 Ke7 12.Bg2 (328.985.757) 157
best move: Rf1-b1 time: 42:10.172 min n/s: 155.643 CPU 100.0% n/s(1CPU): 155.643 nodes: 393.780.000
Only choosing 23.Rfb1 at this depth. By the way there seems to be a fix for Multi-PV possible for all the Togas derived from Toga II 1.4beta5c, found by a Chessbase programmer. Maybe if Denis Mendoza has some time free after working twelve hour shifts
in Texas he could compile it for us. I hope they are paying poor Denis all that overtime! The bugfix seems to be provisionally working in Toga Checkov but I can't test it with more than one thread. N4rk1/1b1p3p/3bpn2/pn6/2P1Ppp1/P2P2Pq/1B1QBP1P/R4RK1 w - -
Engine: Glaurung 2.0.1 Mjolnir 001h (64 MB)
by Tord Romstad
17.01 29:08 +0.45 1.Rfc1 Bxe4 2.Bf1 Qh5 3.cxb5 Bxa8
4.Bg2 Bxg2 5.Kxg2 Qxb5 6.Kg1 Nd5
7.Qe2 f3 8.Qd2 Be7 9.Rab1 Rb8 (311.493.375) 178
best move: Rf1-c1 time: 42:10.000 min n/s: 174.690 nodes: 441.960.000
N4rk1/1b1p3p/3bpn2/pn6/2P1Ppp1/P2P2Pq/1B1QBP1P/R4RK1 w - -
Engine: Rybka 2.2n2 mp 32-bit PVtips5menbases (64 MB)
by Vasik Rajlich
14.01 4:41 +0.37 1.Rfb1 Nxe4 2.dxe4 Bxe4 3.Bf1 fxg3
4.Qg5+ Kf7 5.fxg3 Qh6 6.h4 Qxg5
7.hxg5 Bxb1 8.cxb5 Be4 9.Rd1 (7.257.297) 26
15.01 7:34 +0.37 1.Rfb1 Nxe4 2.dxe4 Bxe4 3.Bf1 fxg3
4.Qg5+ Kf7 5.fxg3 Qh6 6.h4 Qxg5
7.hxg5 Bxb1 8.cxb5 Be4 9.Rd1 (11.978.305) 26
16.01 13:56 +0.15 1.Rfb1 Nxe4 2.dxe4 Bxe4 3.Bf1 Qh6
4.Bg2 Bxg2 5.cxb5 Bf3 6.Qc3 fxg3
7.fxg3 Bxg3 8.Qh8+ Kf7 9.Qg7+ Qxg7
10.Bxg7 (22.662.082) 27
16.04 18:01 +0.41 1.Rfc1 Bxe4 2.Bf1 Qh5 3.cxb5 Bxa8
4.Bg2 Bxg2 5.Kxg2 Qxb5 (28.902.645) 27
17.01 25:12 +0.38 1.Rfc1 Bxe4 2.Bf1 Qh5 3.cxb5 Bxa8
4.Bg2 Bxg2 5.Kxg2 Qxb5 (40.053.464) 27
best move: Rf1-c1 time: 42:09.938 min n/s: 63.257.687 nodes: 63.257.687
What square should the Rook go to and does Black have a draw or can he hope for even more? Only after a very deep search Ancalagon 1.1 is slowly having visions of an oasis in the desert appear just over the search-horizont for Black:
after 1... f4
[d]N4rk1/1b1p3p/3bpn2/pn6/2P1Ppp1/P2P2Pq/1B1QBP1P/R4RK1 w - -
Engine: Ancalagon 1.1 Beta 1 (256 MB)
by Tord Romstad, Eelco de Groot
15 1:53 +1.00 2.Rfd1 Bxe4 3.dxe4 Nxe4 4.Bf1 Qh5 5.cxb5 Nxd2 6.Rxd2 fxg3 7.hxg3 Bc5 8.b6 Rxa8 9.Bg2 Bxb6 10.Bxa8 d5 11.Bd4 Bxd4 12.Rxd4 (35.254.503) 310
15 1:53 +0.78 2.Rfc1 Nxe4 3.dxe4 Bxe4 4.Bf1 fxg3 5.Qg5+ Kf7 6.fxg3 Qh6 7.h4 Qxg5 8.hxg5 Bc5+ 9.Kh2 Nd4 10.Bxd4 Bxd4 11.Nc7 Bxa1 12.Rxa1 (35.254.503) 310
15 1:53 +0.72 2.Rfb1 Nxe4 3.dxe4 Bxe4 4.Bf1 fxg3 5.Qg5+ Kf7 6.fxg3 Qh6 7.h4 Qxg5 8.hxg5 Nxa3 9.Bxa3 Bxa3 10.Rxa3 Rxa8 11.Rb5 (35.254.503) 310
___________________________________________________________
16 5:06 +0.98 2.Rfd1 Bxe4 3.dxe4 Nxe4 4.Bf1 Qh5
5.cxb5 Nxd2 6.Rxd2 fxg3 7.hxg3 Bc5
8.Bd4 Rxa8 9.Bxc5 Qxc5 10.Rb1 Qf5
11.Bd3 Rb8 (94.018.331) 306
16 5:06 +0.96 2.Rfb1 Nxe4 3.dxe4 Bxe4 4.Bf1 fxg3
5.Qg5+ Kf7 6.fxg3 Qh6 7.h4 Qxg5
8.hxg5 Bc5+ 9.Kh2 Nxa3 10.Rd1 Bxa8
11.Rxd7+ Ke8 12.Rd2 (94.018.331) 306
16 5:06 +0.78 2.Rfc1 Nxe4 3.dxe4 Bxe4 4.Bf1 fxg3
5.Qg5+ Kf7 6.fxg3 Qh6 7.h4 Qxg5
8.hxg5 Bc5+ 9.Kh2 Nd4 10.Bxd4 Bxd4
11.Nc7 Bxa1 12.Rxa1 (94.018.331) 306
___________________________________________________________
17 17:20 +0.70 2.Rfb1 Nxe4 3.dxe4 Bxe4 4.Bf1 fxg3
5.Qg5+ Kf7 6.fxg3 Qh6 7.h4 Qxg5
8.hxg5 Bc5+ 9.Kh2 Nd4 10.Bxd4 Bxd4
11.Nc7 Bxb1 12.Rxb1 d6 (316.974.468) 304
17 17:20 +0.70 2.Rfc1 Nxe4 3.dxe4 Bxe4 4.Bf1 fxg3
5.Qg5+ Kf7 6.fxg3 Qh6 7.h4 Qxg5
8.hxg5 Bc5+ 9.Kh2 Nd4 10.Bxd4 Bxd4
11.Nc7 Bxa1 12.Rxa1 d6 (316.974.468) 304
17 17:20 +0.45 2.Rfd1 Bxe4 3.dxe4 Nxe4 4.Bf1 Qh5
5.cxb5 Nxd2 6.Rxd2 fxg3 7.hxg3 Bc5
8.Bd4 Rxa8 9.Bxc5 Qxc5 10.Rb1 Rb8
11.Rxd7 Rf8 12.b6 Rxf2 (316.974.468) 304
___________________________________________________________
18 96:11 +1.05 2.Rfd1 Bxe4 3.dxe4 Nxe4 4.Bf1 Qh5
5.cxb5 Bc5 6.gxf4 Nxd2 7.Rxd2 Rxa8
8.Bd4 Rb8 9.a4 d6 10.Rad1 Rb6
11.Bxc5 Qxc5 12.Be2 Qd4 (1.197.409.481) 207
18 96:11 +0.72 2.Rfc1 Nxe4 3.dxe4 Bxe4 4.Bf1 fxg3
5.Qg5+ Kf7 6.fxg3 Qh6 7.h4 Qxg5
8.hxg5 Bc5+ 9.Kh2 Nd4 10.Bxd4 Bxd4
11.Nc7 Bxa1 12.Rxa1 Ke7 13.Rd1 Rf2+
14.Kg1 (1.197.409.481) 207
18 96:11 +0.49 2.Rfb1 Nxe4 3.dxe4 Bxe4 4.Bf1 fxg3
5.Qg5+ Kf7 6.fxg3 Qh6 7.h4 Qxg5
8.hxg5 Bc5+ 9.Kh2 Nd4 10.Bxd4 Bxd4
11.Nb6 Kg6 12.Bg2 Bxb1 13.Rxb1 Bxb6
14.Rxb6 Kxg5 (1.197.409.481) 207
___________________________________________________________
19 146:03 -1.25 2.Rfd1 fxg3 3.Qg5+ Kf7 4.fxg3 Bxg3
5.hxg3 Qxg3+ 6.Kf1 Nh5 7.Bf3 Qxf3+
8.Kg1 Qg3+ 9.Kh1 Qh3+ 10.Kg1 Ke8
11.Rf1 Qg3+ 12.Kh1 Nf4 13.Rxf4 Qh3+
14.Kg1 Qe3+ 15.Kh1 (2.097.693.705) 239
Eelco
Debugging is twice as hard as writing the code in the first
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
-- Brian W. Kernighan
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
-- Brian W. Kernighan
-
Eelco de Groot
- Posts: 4673
- Joined: Sun Mar 12, 2006 2:40 am
- Full name: Eelco de Groot
Re: Muri-Peixoto, corr EU-ch M1231 2001
Three best moves at 19 ply so far, White improved just marginally but Black is still doing better. 22...f4 seems a good testmove, programs should also get bonus points for a positive Black score though!
N4rk1/1b1p3p/3bpn2/pn6/2P1Ppp1/P2P2Pq/1B1QBP1P/R4RK1 w - -
Engine: Ancalagon 1.1 Beta 1 (256 MB)
by Tord Romstad, Eelco de Groot
19 202:14 -1.19 2.Rfb1 Nxe4 3.dxe4 Bxe4 4.Bf1 fxg3
5.Qg5+ Kf7 6.fxg3 Qh6 7.h4 Qxg5
8.hxg5 Bc5+ 9.Kh2 Nd4 10.Bxd4 Bxd4
11.Re1 Bxa1 12.Rxe4 Rxa8 13.Rxg4 Bc3
14.Kg2 (3.092.723.295) 254
19 202:14 -1.25 2.Rfd1 fxg3 3.Qg5+ Kf7 4.fxg3 Bxg3
5.hxg3 Qxg3+ 6.Kf1 Nh5 7.Bf3 Qxf3+
8.Kg1 Qg3+ 9.Kh1 Qh3+ 10.Kg1 Ke8
11.Rf1 Qg3+ 12.Kh1 Nf4 13.Rxf4 Qh3+
14.Kg1 Qe3+ 15.Kh1 (3.092.723.295) 254
19 202:14 -1.78 2.Rfc1 fxg3 3.Qg5+ Kf7 4.hxg3 Bxg3
5.Qxf6+ Ke8 6.Qxf8+ Kxf8 7.fxg3 Qxg3+
8.Kf1 Qe3 9.cxb5 g3 10.Kg2 Qxe2+
11.Kxg3 Qxd3+ 12.Kf2 Qxb5 13.Rab1 Qg5
14.Bg7+ Qxg7 15.Rxb7 (3.092.723.295) 254
Eelco
N4rk1/1b1p3p/3bpn2/pn6/2P1Ppp1/P2P2Pq/1B1QBP1P/R4RK1 w - -
Engine: Ancalagon 1.1 Beta 1 (256 MB)
by Tord Romstad, Eelco de Groot
19 202:14 -1.19 2.Rfb1 Nxe4 3.dxe4 Bxe4 4.Bf1 fxg3
5.Qg5+ Kf7 6.fxg3 Qh6 7.h4 Qxg5
8.hxg5 Bc5+ 9.Kh2 Nd4 10.Bxd4 Bxd4
11.Re1 Bxa1 12.Rxe4 Rxa8 13.Rxg4 Bc3
14.Kg2 (3.092.723.295) 254
19 202:14 -1.25 2.Rfd1 fxg3 3.Qg5+ Kf7 4.fxg3 Bxg3
5.hxg3 Qxg3+ 6.Kf1 Nh5 7.Bf3 Qxf3+
8.Kg1 Qg3+ 9.Kh1 Qh3+ 10.Kg1 Ke8
11.Rf1 Qg3+ 12.Kh1 Nf4 13.Rxf4 Qh3+
14.Kg1 Qe3+ 15.Kh1 (3.092.723.295) 254
19 202:14 -1.78 2.Rfc1 fxg3 3.Qg5+ Kf7 4.hxg3 Bxg3
5.Qxf6+ Ke8 6.Qxf8+ Kxf8 7.fxg3 Qxg3+
8.Kf1 Qe3 9.cxb5 g3 10.Kg2 Qxe2+
11.Kxg3 Qxd3+ 12.Kf2 Qxb5 13.Rab1 Qg5
14.Bg7+ Qxg7 15.Rxb7 (3.092.723.295) 254
Eelco
Debugging is twice as hard as writing the code in the first
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
-- Brian W. Kernighan
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
-- Brian W. Kernighan
-
Eelco de Groot
- Posts: 4673
- Joined: Sun Mar 12, 2006 2:40 am
- Full name: Eelco de Groot
New search changes in Ancalagon 1.1
I used Jon's testposition to develop Ancalagon's search a little, along the lines of MIID, Manual Incremental Iterative Deepening as described in my lay-out for Vas for future Rybka 4 on the Rybka forum 
but then of course in the fully automated version idea that was already developed for Ancalagon. The last search version does also show some signs of instability if the search stays too long with one single line, but maybe that can be further refined. Maybe it's more an effect of (apparent) selectivity that is caused by the latest changes. Also the implementation is not very efficient and may be doing double or triple work here and there. Indeed, if you would look at the search lay-out you'd think it wouldn't work not in a million years! But because of the basic strength of Tord's search-routines, apart from the double work done by my "programming" the efficiency is still there even with all the changes, that really are not so much programming as some rearranging of Tord's basic supersearchroutines. It is not so much better I suppose in Ancalagon but certainly a bit unorthodox. Now if only I had a fast compile and better eval!
The latest version of the search is in Build 19 and finally 22...f4 is found a bit faster. That does not say very much because it could also just point to too much selectivity... I'm curious just how fast Rybka is here?
[d]N4rk1/1b1p3p/3bpn2/pn3p2/2P1P1p1/P2P2Pq/1B1QBP1P/R4RK1 b - -
Engine: Ancalagon 1.1 Beta 1 (Athlon 2009 MHz, 256 MB)
by Tord Romstad, Eelco de Groot
8.03 0:01 -4.05 22...f4 23.Nb6 Nxe4 24.dxe4 Bxe4
25.f3 Bc5+ 26.Kh1 Nd4 (360.817) 278
8.29 0:01 0.00 22...Nxe4 23.dxe4 Bxe4 24.Qg5+ Kf7
25.f3 Bxg3 26.Qg7+ Ke8 27.hxg3 Qxg3+
28.Kh1 Qh3+ 29.Kg1 Qg3+ 30.Kh1 (467.880) 285
9.01 0:02 0.00 22...Nxe4 23.dxe4 Bxe4 24.Qg5+ Kf7
25.f3 Bxg3 26.Qg7+ Ke8 27.hxg3 Qxg3+
28.Kh1 Qh3+ 29.Kg1 Qg3+ 30.Kh1 (595.751) 286
10.01 0:03 0.00 22...Nxe4 23.dxe4 Bxe4 24.Qg5+ Kf7
25.f3 Bxg3 26.Qg7+ Ke8 27.hxg3 Qxg3+
28.Kh1 Qh3+ 29.Kg1 Qg3+ 30.Kh1 (935.263) 296
11.01 0:06 -0.19 22...Nxe4 23.dxe4 Bxe4 24.f3 Bxg3
25.Bd3 Bxd3 26.cxb5 Bxb5 27.hxg3 Qxg3+
28.Qg2 Qxg2+ 29.Kxg2 Bxf1+ 30.Rxf1 Rxa8
31.fxg4 Rb8 (2.028.341) 290
12.01 0:11 -0.29 22...Nxe4 23.dxe4 Bxe4 24.f3 Bxg3
25.Bd3 Bxd3 26.cxb5 Bxb5 27.hxg3 Qxg3+
28.Qg2 Qxg2+ 29.Kxg2 Bxf1+ 30.Rxf1 Rxa8
31.fxg4 Rb8 32.Be5 (3.425.273) 296
13.01 0:21 -0.45 22...Nxe4 23.dxe4 Bxe4 24.f3 Bxg3
25.Bd3 Bxd3 26.cxb5 Bxb5 27.hxg3 Qxg3+
28.Qg2 Qxg2+ 29.Kxg2 Bxf1+ 30.Rxf1 Rxa8
31.fxg4 Rb8 32.Be5 Rb3 (6.306.011) 299
14.01 1:01 -3.80 22...Nxe4 23.dxe4 Bxe4 24.f3 Bxg3
25.Bd1 Bd3 26.Ba4 Bxc4 27.Nb6 Qxh2+
28.Qxh2 Bxh2+ 29.Kxh2 Bxf1 30.Rxf1 Nxa3
31.Bxa3 gxf3 32.Kg1 (18.024.643) 294
14.02 2:56 -0.88 22...f4 23.Rfc1 Nxe4 24.dxe4 Bxe4
25.Bf1 fxg3 26.Qg5+ Kf7 27.fxg3 Qh6
28.h4 Qxg5 29.hxg5 Bc5+ 30.Kh2 Nxa3 (53.580.731) 302
15.01 4:34 -1.68 22...f4 23.Rfc1 Nxe4 24.dxe4 Bxe4
25.Bf1 fxg3 26.Qg5+ Kf7 27.fxg3 Qh6
28.h4 Qxg5 29.hxg5 Bc5+ 30.Kh2 Kg8
31.Bg2 Nxa3 32.Bxa3 Bxa3 33.Bxe4 Bxc1
34.Rxc1 (83.259.226) 302
16.01 8:11 -1.68 22...f4 23.Rfc1 Nxe4 24.dxe4 Bxe4
25.Bf1 fxg3 26.Qg5+ Kf7 27.fxg3 Qh6
28.h4 Qxg5 29.hxg5 Bc5+ 30.Kh2 Kg8
31.Bg2 Nxa3 32.Bxa3 Bxa3 33.Bxe4 Bxc1
34.Rxc1 (151.575.608) 308
17.01 17:56 -1.60 22...f4 23.Rfc1 Nxe4 24.dxe4 Bxe4
25.Bf1 fxg3 26.Qg5+ Kf7 27.fxg3 Qh6
28.h4 Qxg5 29.hxg5 Bc5+ 30.Kh2 Kg8
31.Bg2 Nxa3 32.Bxa3 Bxa3 33.Rxa3 Bxg2
34.Kxg2 Rxa8 (336.238.365) 312
18.01 47:06 -0.72 22...f4 23.Rfc1 Nxe4 24.dxe4 Bxe4
25.Bf1 fxg3 26.Qg5+ Kf7 27.fxg3 Qh6
28.h4 Qxg5 29.hxg5 Bc5+ 30.Kh2 Nd4
31.Bxd4 Bxd4 32.Nc7 Bxa1 33.Rxa1 Ke7
34.Rd1 Rf2+ 35.Kg1 (890.568.044) 315
best move: f5-f4 time: 53:31.125 min n/s: 316.219 nodes: 1.015.410.000
Eelco
but then of course in the fully automated version idea that was already developed for Ancalagon. The last search version does also show some signs of instability if the search stays too long with one single line, but maybe that can be further refined. Maybe it's more an effect of (apparent) selectivity that is caused by the latest changes. Also the implementation is not very efficient and may be doing double or triple work here and there. Indeed, if you would look at the search lay-out you'd think it wouldn't work not in a million years! But because of the basic strength of Tord's search-routines, apart from the double work done by my "programming" the efficiency is still there even with all the changes, that really are not so much programming as some rearranging of Tord's basic supersearchroutines. It is not so much better I suppose in Ancalagon but certainly a bit unorthodox. Now if only I had a fast compile and better eval! The latest version of the search is in Build 19 and finally 22...f4 is found a bit faster. That does not say very much because it could also just point to too much selectivity... I'm curious just how fast Rybka is here?
[d]N4rk1/1b1p3p/3bpn2/pn3p2/2P1P1p1/P2P2Pq/1B1QBP1P/R4RK1 b - -
Engine: Ancalagon 1.1 Beta 1 (Athlon 2009 MHz, 256 MB)
by Tord Romstad, Eelco de Groot
8.03 0:01 -4.05 22...f4 23.Nb6 Nxe4 24.dxe4 Bxe4
25.f3 Bc5+ 26.Kh1 Nd4 (360.817) 278
8.29 0:01 0.00 22...Nxe4 23.dxe4 Bxe4 24.Qg5+ Kf7
25.f3 Bxg3 26.Qg7+ Ke8 27.hxg3 Qxg3+
28.Kh1 Qh3+ 29.Kg1 Qg3+ 30.Kh1 (467.880) 285
9.01 0:02 0.00 22...Nxe4 23.dxe4 Bxe4 24.Qg5+ Kf7
25.f3 Bxg3 26.Qg7+ Ke8 27.hxg3 Qxg3+
28.Kh1 Qh3+ 29.Kg1 Qg3+ 30.Kh1 (595.751) 286
10.01 0:03 0.00 22...Nxe4 23.dxe4 Bxe4 24.Qg5+ Kf7
25.f3 Bxg3 26.Qg7+ Ke8 27.hxg3 Qxg3+
28.Kh1 Qh3+ 29.Kg1 Qg3+ 30.Kh1 (935.263) 296
11.01 0:06 -0.19 22...Nxe4 23.dxe4 Bxe4 24.f3 Bxg3
25.Bd3 Bxd3 26.cxb5 Bxb5 27.hxg3 Qxg3+
28.Qg2 Qxg2+ 29.Kxg2 Bxf1+ 30.Rxf1 Rxa8
31.fxg4 Rb8 (2.028.341) 290
12.01 0:11 -0.29 22...Nxe4 23.dxe4 Bxe4 24.f3 Bxg3
25.Bd3 Bxd3 26.cxb5 Bxb5 27.hxg3 Qxg3+
28.Qg2 Qxg2+ 29.Kxg2 Bxf1+ 30.Rxf1 Rxa8
31.fxg4 Rb8 32.Be5 (3.425.273) 296
13.01 0:21 -0.45 22...Nxe4 23.dxe4 Bxe4 24.f3 Bxg3
25.Bd3 Bxd3 26.cxb5 Bxb5 27.hxg3 Qxg3+
28.Qg2 Qxg2+ 29.Kxg2 Bxf1+ 30.Rxf1 Rxa8
31.fxg4 Rb8 32.Be5 Rb3 (6.306.011) 299
14.01 1:01 -3.80 22...Nxe4 23.dxe4 Bxe4 24.f3 Bxg3
25.Bd1 Bd3 26.Ba4 Bxc4 27.Nb6 Qxh2+
28.Qxh2 Bxh2+ 29.Kxh2 Bxf1 30.Rxf1 Nxa3
31.Bxa3 gxf3 32.Kg1 (18.024.643) 294
14.02 2:56 -0.88 22...f4 23.Rfc1 Nxe4 24.dxe4 Bxe4
25.Bf1 fxg3 26.Qg5+ Kf7 27.fxg3 Qh6
28.h4 Qxg5 29.hxg5 Bc5+ 30.Kh2 Nxa3 (53.580.731) 302
15.01 4:34 -1.68 22...f4 23.Rfc1 Nxe4 24.dxe4 Bxe4
25.Bf1 fxg3 26.Qg5+ Kf7 27.fxg3 Qh6
28.h4 Qxg5 29.hxg5 Bc5+ 30.Kh2 Kg8
31.Bg2 Nxa3 32.Bxa3 Bxa3 33.Bxe4 Bxc1
34.Rxc1 (83.259.226) 302
16.01 8:11 -1.68 22...f4 23.Rfc1 Nxe4 24.dxe4 Bxe4
25.Bf1 fxg3 26.Qg5+ Kf7 27.fxg3 Qh6
28.h4 Qxg5 29.hxg5 Bc5+ 30.Kh2 Kg8
31.Bg2 Nxa3 32.Bxa3 Bxa3 33.Bxe4 Bxc1
34.Rxc1 (151.575.608) 308
17.01 17:56 -1.60 22...f4 23.Rfc1 Nxe4 24.dxe4 Bxe4
25.Bf1 fxg3 26.Qg5+ Kf7 27.fxg3 Qh6
28.h4 Qxg5 29.hxg5 Bc5+ 30.Kh2 Kg8
31.Bg2 Nxa3 32.Bxa3 Bxa3 33.Rxa3 Bxg2
34.Kxg2 Rxa8 (336.238.365) 312
18.01 47:06 -0.72 22...f4 23.Rfc1 Nxe4 24.dxe4 Bxe4
25.Bf1 fxg3 26.Qg5+ Kf7 27.fxg3 Qh6
28.h4 Qxg5 29.hxg5 Bc5+ 30.Kh2 Nd4
31.Bxd4 Bxd4 32.Nc7 Bxa1 33.Rxa1 Ke7
34.Rd1 Rf2+ 35.Kg1 (890.568.044) 315
best move: f5-f4 time: 53:31.125 min n/s: 316.219 nodes: 1.015.410.000
Eelco
Debugging is twice as hard as writing the code in the first
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
-- Brian W. Kernighan
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
-- Brian W. Kernighan
-
Omega-three
Re: New search changes in Ancalagon 1.1
Rybka 3 running in GUI11I'm curious just how fast Rybka is here?
Athlon 4200x2
win XP x64
Analysis by Deep Rybka 3:
1...Nxe4 2.dxe4 Bxe4 3.f3 Bxg3 4.Bd1 Bd3
+- (1.66) Depth: 6 00:00:00 35kN
1...Nxe4 2.dxe4 Bxe4 3.f3 Bxg3 4.Bd1 Bd3 5.Ba4 Bxf1 6.hxg3 Rxa8 7.Rxf1 Qxg3+ 8.Qg2
+- (1.82) Depth: 7 00:00:01 80kN
1...Nxe4 2.dxe4 Bxe4 3.f3 Bxg3 4.Bd1 Bd3 5.Ba4 Bxf1 6.hxg3 Rxa8 7.Rxf1 Qxg3+ 8.Qg2
+- (1.82) Depth: 8 00:00:01 90kN
1...Nxe4
+- (1.62 !) Depth: 9 00:00:02 178kN
1...f4
+- (1.42 !) Depth: 9 00:00:04 310kN
1...f4
+/- (1.02 !) Depth: 9 00:00:05 424kN
1...f4 2.Rfb1 Bxe4[] 3.Bf1 Qh5 4.Bxf6 Bc6 5.Rxb5 Bxb5 6.cxb5 Rxf6 7.Bg2 Rh6 8.gxf4 Qxb5 9.Be4
+/- (1.06) Depth: 9 00:00:09 744kN
1...f4 2.Rfb1 Bxe4[] 3.Bf1 Qh5 4.Bxf6 Bc6 5.Rxb5 Bxb5 6.cxb5 Rxf6 7.Bg2 Rh6 8.gxf4 Qxb5 9.Be4
+/- (1.06) Depth: 10 00:00:11 899kN
1...f4 2.Rfb1 Bxe4 3.Bf1 Qh5 4.Bxf6 Bc6 5.Rxb5 Bxb5 6.cxb5 Rxf6 7.Bg2 Qxb5 8.d4 Qa6 9.Be4
+/- (1.05) Depth: 11 00:00:14 1219kN
1...f4
+/- (0.85 !) Depth: 12 00:00:23 2051kN
1...f4
+/= (0.65 !) Depth: 12 00:00:32 2865kN
1...f4 2.Rfd1 Bxe4[] 3.Bf1
+/= (0.47) Depth: 12 00:00:43 3935kN
1...f4 2.Rfd1 Bxe4[] 3.Bf1 Qh5[] 4.cxb5
+/= (0.55) Depth: 13 00:00:52 4772kN
1...f4 2.Rfc1 Bxe4 3.Bf1 Qh5 4.cxb5
+/= (0.43) Depth: 14 00:01:31 8438kN
1...f4
= (0.23 !) Depth: 15 00:02:32 14488kN
1...f4
= (0.03 !) Depth: 15 00:04:32 25886kN
1...f4 2.Rfc1 Nxe4 3.dxe4 Bxe4[] 4.Bf1 fxg3[] 5.Qg5+ Kf7[] 6.fxg3 Qh6[] 7.h4[] Qxg5[] 8.hxg5 Bc5+[] 9.Kh2[] Kg6 10.Bf6[] Nd4[] 11.Bg2[] Bxa8 12.Bxd4 Bxd4 13.Rf1 Bf3 14.Rad1 Bxd1 15.Rxf8[] Kxg5 16.Rd8 Ba4
= (0.02) Depth: 15 00:05:54 34053kN
1...f4 2.Rfc1 Nxe4 3.dxe4 Bxe4[] 4.Bf1 fxg3[] 5.Qg5+ Kf7[] 6.fxg3 Qh6[] 7.h4[] Qxg5[] 8.hxg5 Bc5+[] 9.Kh2[] Kg6 10.Bf6[] Nd4[] 11.Bg2 Bxa8 12.Bxd4 Bxd4 13.Rf1 Bf3 14.Rad1 Bxd1 15.Rxf8[] Kxg5 16.Rd8 Ba4
= (0.02) Depth: 16 00:07:42 44365kN
1...f4 2.Rfc1 Nxe4 3.dxe4 Bxe4[] 4.Bf1[] fxg3[] 5.Qg5+[] Kf7[] 6.fxg3 Qh6[] 7.h4[] Qxg5[] 8.hxg5[] Bc5+[] 9.Kh2[] Kg6 10.Bf6[] Nd4[] 11.Bg2 Bxa8 12.Bxd4 Bxd4 13.Rf1 Bf3 14.Rad1 Bxd1 15.Rxf8[] Kxg5 16.Rd8 Ba4
= (0.02) Depth: 17 00:11:57 68637kN
1...f4 2.Rfb1 Nxe4 3.dxe4 Bxe4 4.Bf1 Qh5 5.cxb5
= (0.00) Depth: 18 00:26:09 149mN
Regards
Mike