Endgame position: progress in eval?

[ernest]

Found this in the German CSS-Forum
http://forum.computerschach.de/cgi-bin/ ... l?tid=4956

How long does it take for your program (and machine) to see a progress in evaluation - if any - in this endgame position?

[d]7n/6k1/5pp1/2Q2n1p/7P/5PP1/6K1/8 b - - 0 0

[Eelco de Groot]

I have three different versions that I let loose on the position. They all seem to stabilize on a value but the value is pretty low so I would not say it is conclusive, from this evidence, it is a draw. In general in Rainbow serpent the material value of the knight is still pretty low, compared especially to some pawns if they have big bonuses and to material eval of the bishop pair, especially without a bad bishop. For three or more knights the material imbalance table does give an extra bonus, see also the experiments from HGM about that, but for two knights there is a bonus only in the presence of the bishop pair -a bonus for still having all the minors, I put that in mainly to improve setting up kingattacks, at least I hope it will help do that in the long run- but which in this position does not really help of course. So, the material imbalance table that is surprisingly pessimistic about the value of the knights in the Stockfish material imbalance code, maybe in Rainbow Serpent's case, at least here, is still undervalueing the blacks partly for the same reason and this shows in a high score for white, compared to Junior for instance that I believe was used in the CSS thread.

Output from the three versions of Rainbow Serpent

The first version, build 167 which I have been using for some time now, goes to depth 35 okay, and reasonably fast but then I think the search explodes somewhere and no new output is received for twelve hours or more There is no real crash, or threads at low percentages of CPU use, but for the enduser still a bit frustrating...

7n/6k1/5pp1/2Q2n1p/7P/5PP1/6K1/8 b - -

Engine: Rainbow Serpent 2.2.4 Build 167 (512 MB)
by Tord Romstad, Marco Costalba and Joona Kiiski
.
.
.
24/41 0:31 -5.63 1...Nf7 2.Kh3 N5h6 3.Qa7 Nf5 4.g4 N5d6
5.Qc5 Nb7 6.Qd4 Nbd6 7.Qd3 Kh6 8.Qd5 hxg4+
9.fxg4 Kg7 10.Qd4 Nb5 11.Qd2 Nbd6
12.Qb4 Kf8 13.Qc3 Ne5 14.Qa3 Nef7 (121.918.110) 3898

25/43 1:19 -5.79 1...Nf7 2.Kh3 N5h6 3.Qa7 Nf5 4.g4 N5d6
5.Qd4 hxg4+ 6.fxg4 Nb5 7.Qd5 Nbd6
8.Qd4 (322.778.788) 4069

26/47 2:44 -5.80 1...Nf7 2.Kh3 N5h6 3.Qa7 Kf8 4.Qa3+ Kg7
5.g4 Ne5 6.Qe7+ Nhf7 7.gxh5 gxh5
8.Kg3 Kg6 9.Qe6 Nh6 10.Kf4 Nd3+
11.Ke3 Ne5 12.Ke4 Kg7 13.Qe7+ Kg6
14.Qe8+ Kg7 (696.808.352) 4234

27/50 3:12 -5.93 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qa7 Kf8 8.Qb6 Kg7
9.Qb7 Kf8 10.Qc8+ Ke7 11.Qc5+ Ke8
12.Qc6+ Ke7 13.Qe4+ Kf8 14.Qe6 Kg7 (804.370.984) 4186

28/50 3:14 -5.93 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qa7 Kf8 8.Qb6 Kg7
9.Qb7 Kf8 10.Qc8+ Ke7 11.Qc5+ Ke8
12.Qc6+ Ke7 13.Qc5+ (811.979.440) 4183

29/59 6:17 -5.93 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qa7 Kf8 8.Qb6 Kg7
9.Qb7 (1.638.893.001) 4341

30/59 6:30 -5.91 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qa7 Kf8 8.Qb6 Kg7
9.Qb7 Kf8 10.Qb6 (1.694.313.992) 4339

31/59 7:09 -5.91 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qa7 Kf8
8.Qc5+ Kg8 9.Qd5 Kg7 10.Qc5 (1.863.949.794) 4344

32/74 7:53 -5.91 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qa7 Kf8
8.Qc5+ Kg8 9.Qd5 Kg7 10.Qd7 Kg8
11.Kg3 Kh8 12.Qe7 Kg7 13.Qe6 Ne5
14.Qe7+ Nef7 (2.076.063.826) 4382

33/74 8:06 -5.91 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qa7 Kf8
8.Qc5+ Kg8 9.Qd5 Kg7 10.Qd7 Kf8
11.Qe6 Kg7 12.Qe7 f5 13.Qe6 fxg4+
14.Kg3 Kg8 (2.135.629.615) 4390

34/74 8:20 -5.91 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qa7 Kf8
8.Qc5+ Kg8 9.Qd5 Kg7 10.Qc5 (2.209.352.395) 4417

35/74 8:36 -5.91 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qa7 Kf8
8.Qc5+ Kg8 9.Qd5 Kg7 10.Qe6 Ne5
11.Qe7+ (2.281.712.504) 4420

best move: Nh8-f7 time: 802:16.822 min n/s: 4.420.621 nodes: 2.281.712.504


The next version, I'm skipping build 168 so this is 169, does cross the psychological barrier of -6 but also has trouble going deep, more false fail highs at the start of every iteration which are still an unsolved problem also in the case of Stockfish I dare guess, but it is still unclear to me at least where they are coming from. The 0.00 value can be explained because Rainbow Serpent will adapt the search window in case the search goes below VALUE_DRAW. (I am not being very precise here because I have no access to the Rainbow Swerpent code on this computer. Anyway, the sourcecode of Rainbow Serpent has to be posted one of these days, maybe for the next build I made just now, build 170.

7n/6k1/5pp1/2Q2n1p/7P/5PP1/6K1/8 b - -

Engine: Rainbow Serpent 2.2.4 Build 169 (512 MB)
by Tord Romstad, Marco Costalba and Joona Kiiski
.
.
.
24/36 0:30 -5.75 1...Nf7 2.Kh3 N5h6 3.Qc7 Nf5 4.Qd7 Ne3
5.Qd4 Nf5 6.Qa7 g5 7.Qc5 N5h6 8.Qd5 gxh4
9.Kxh4 Kh7 10.Qxh5 Nd6 11.g4 Kg7
12.Qc5 Nb7 (118.704.453) 3840

25/38 0:32 0.00++ 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 Ne5
5.Qa7+ Nhf7 6.Qg1 Nf3 7.Qb6 N7e5
8.Qb7+ Kf8 9.Qa8+ Kg7 10.Qe8 g5
11.Qe7+ Kg6 12.h5+ Kh6 13.Qxf6+ Ng6
14.hxg6 (125.882.055) 3836

25/45 0:40 -5.73 1...Nf7 2.Kh3 N5h6 3.Qc7 Nf5 4.Qd7 N5d6
5.Qc6 Nf5 6.g4 hxg4+ 7.fxg4 N5d6
8.Qd5 f5 9.Qd4+ Kh7 10.Qf6 fxg4+
11.Kxg4 Nh6+ 12.Kf4 Nhf7 13.Kg4 (157.556.305) 3938

26/45 0:44 +0.02++ 1...Nf7 2.Kh3 Ne5 3.Qc7+ Nf7 4.g4 hxg4+
5.fxg4 N5h6 6.Kg3 Nxg4 7.Kxg4 f5+
8.Kf4 Kf6 9.Qb6+ Kg7 10.Qa7 Kf6
11.Qc7 g5+ 12.hxg5+ Nxg5 13.Qb6+ Ne6+
14.Kg3 f4+ (174.507.133) 3964

26/45 0:54 -5.89 1...Nf7 2.Kh3 Ne5 3.Qc7+ Nf7 4.g4 hxg4+
5.fxg4 N5h6 6.Qa7 g5 7.h5 Kf8 8.Qa3+ Ke8
9.Qb3 Ke7 10.Qb7+ Kf8 11.Qb6 Kg7
12.Qb4 Ne5 13.Qe7+ Nef7 14.Qd7 Kf8 (228.603.746) 4185

27/45 0:57 0.00++ 1...Nf7 2.Kh3 Ng5+ 3.hxg5 Nxg3
4.gxf6+ Kh6 5.Kxg3 Kh7 6.Kf4 Kg8 (242.575.782) 4182

27/46 1:27 -5.88 1...Nf7 2.Kh3 N5h6 3.Qc7 Nf5 4.g4 N5h6
5.Qd7 hxg4+ 6.fxg4 g5 7.h5 Kf8
8.Qc8+ Ke7 9.Qc2 Kf8 10.Qc6 Kg7
11.Qd7 (378.958.126) 4331

28/46 1:31 0.00++ 1...Nf7 2.Kh3 N5h6 3.Qc8 Ne5 4.Qc7+ Nhf7
5.Qa7 Nxf3 6.Qa4 N7e5 7.Qd1 Kf7
8.Qd5+ Kg7 9.Qd6 Nc4 10.Qc5 Ncd2 (394.458.331) 4333

28/49 2:40 -5.81 1...Nf7 2.Kh3 N5h6 3.Qc8 Ne5 4.Qc7+ Nef7
5.Qd7 Nf5 6.Qa7 g5 7.hxg5 fxg5 8.Qd7 Kf6
9.Qd5 Ne5 10.Qe4 g4+ 11.Kg2 (703.459.015) 4389

29/49 2:44 0.00++ 1...Nf7 2.Kh3 N5h6 3.Qa7 Ng8 4.g4 Ngh6
5.Qe7 Ng8 6.Qa7 (720.955.637) 4388

29/51 4:45 -5.97 1...Nf7 2.Kh3 N5h6 3.Qa7 Nf5 4.g4 hxg4+
5.fxg4 N5d6 6.Qb6 Nc4 7.Qc5 Nce5
8.Qe7 Nc6 9.Qc7 Nce5 10.Qb6 g5
11.Qg1 (1.243.652.484) 4355

30/51 4:56 0.00++ 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.Kg3 Nf5+
5.Kf4 gxf3 6.Qa3 Nxh4 7.Qa6 Ne5
8.Qf1 g5+ (1.285.336.221) 4338

30/78 16:38 -6.13 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qd7 Ng8 8.Qe8 Ne5
9.Kg3 Nh6 10.Qe7+ Nef7 11.Kh3 f5
12.gxf5 g4+ 13.Kh2 Nxf5 14.Qe6 N7h6 (4.441.193.127) 4449

31/78 29:04 -6.13 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qa7 Kh7 8.Qd7 Kg7
9.Qe6 Nd8 10.Qe3 Kf7 11.Qd4 Nb7
12.Qc4+ (7.709.416.148) 4419

32/78 29:37 0.00++ 1...Nf7 2.Kh3 N5d6 3.g4 g5 4.gxh5 Kh6
5.hxg5+ Nxg5+ (7.852.757.183) 4418

32/96 58:32 -6.13 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.Qe7 gxh4 6.Qc7 Kg6 (15.713.889.858) 4473

33/96 60:31 0.00++ 1...Nf7 2.Kh3 N5h6 3.Qa7 Nf5 4.g4 N5h6
5.Kg3 hxg4 6.fxg4 Kf8 7.Qc5+ Kg7
8.Qe7 Ng8 9.Qb7 Ngh6 10.Qc7 Kg8
11.Qc4 (16.422.596.643) 4522

33/108 87:17 -5.85 1...Nf7 2.Kh3 N5h6 3.Qa7 Kg8 4.Qa2 Kg7
5.Qc4 Nf5 6.g4 hxg4+ 7.fxg4 N5h6
8.Qc5 g5 9.h5 Ne5 10.Qe7+ Nef7
11.Qe8 Ne5 12.Qe7+ (23.603.240.602) 4506

34/108 89:11 0.00++ 1...Nf7 2.Kh3 Ng5+ 3.hxg5 Kh7 4.gxf6 (24.113.304.475) 4506

34/178 194:26 -5.96 1...Nf7 2.Kh3 N5h6 3.Qc7 Nf5 4.Qa5 Ne7
5.Qc5 Ng8 6.Qb6 f5 7.Qe6 Nfh6 8.Qd6 Nf7
9.Qc5 Nf6 10.Kg2 Nh6 11.Qe7+ Nf7
12.g4 hxg4 13.Qc7 (51.373.020.306) 4403

35/178 197:10 0.00++ 1...Nf7 2.Kh3 N5h6 3.Qa7 Kh7 4.f4 Nf5
5.Qxf7+ Kh6 6.Qxf6 (52.407.562.042) 4430

35/178 204:28 -5.65 1...Nf7 2.Kh3 N5h6 3.Qa7 Kh7 4.Qe7 Kg7
5.Qa7 (54.472.650.215) 4440

best move: Nh8-f7 time: 228:47.464 min n/s: 4.440.208 nodes: 54.472.650.215

Build 170 has again a very different 'time to depth' pattern also known as the 'branching factor'. This is because I tried to do the re-searches inside the TT-search block without gratuitous extensions, so just like Stockfish does them, only adapting the size of the search window in the upwards or downwards direction, but not the searchdepth. I did not think it would have this much effect on the branching factor but I also took them out of the PV Leaf Search in build 167 and at least it seems to help if you want to see the program do more iterations. This version has not been running very long but I just post what I have now. Depth 37 has not been reached at the moment so it is nor sure this version is really able to go much deeper than depth 36. Still I like the branching compared to build 167 and also compared to 169.

Eelco

7n/6k1/5pp1/2Q2n1p/7P/5PP1/6K1/8 b - -

Engine: Rainbow Serpent 2.2.4 Build 170 (4 threads @ 2672 MHz, 512 MB)
by Tord Romstad, Marco Costalba and Joona Kiiski

15/24 0:00 -5.13 1...Nf7 2.Qa7 g5 3.Qd7 Kg6 4.hxg5 Nxg5
5.Qe8+ Nf7 6.Kf2 Kg7 7.Qe6 Kg6 8.Qd7 Ne5
9.Qe8+ Kh6 10.Qd8 Kg6 11.Qh8 Kf7 (2.864.481) 2867

16/26 0:02 -5.21 1...Nf7 2.Kh3 Ne5 3.g4 hxg4+ 4.fxg4 Nh6
5.Qe7+ Nhf7 6.Kg3 Nc6 7.Qc7 Nce5
8.Kf4 Kf8 9.g5 fxg5+ 10.hxg5 Ke8
11.Ke4 (8.524.912) 3436

17/29 0:05 -5.46-- 1...Nf7 2.Kh3 N5d6 3.g4 hxg4+ 4.fxg4 Kh7
5.Qd5 Kg7 6.Qd3 f5 7.Qd4+ Kh7 8.Qf6 fxg4+
9.Kxg4 Nh6+ 10.Kf4 Nhf7 11.Ke3 Ne5
12.Kd4 Ndc4 13.Qe6 (19.941.259) 3463

17/29 0:05 -5.43 1...Nf7 2.Kh3 N5d6 3.g4 hxg4+ 4.fxg4 Kh7
5.Qd5 Kg7 6.Qd3 Ne8 7.Qd5 Ned6 (20.679.188) 3460

18/29 0:06 +0.13++ 1...Nf7 2.Kh3 N5d6 3.g4 hxg4+ 4.fxg4 Ne8
5.Qd5 Ned6 6.Qd4 Nb5 7.Qe3 Kg8 8.Qe7 g5 (21.231.903) 3454

18/29 0:06 -5.53 1...Nf7 2.Kh3 N5d6 3.g4 hxg4+ 4.fxg4 Ne8
5.Qd5 Ned6 6.Kg3 f5 7.Qd4+ Kh6 8.g5+ Kh7
9.Kf3 Ne4 10.Qd7 Kg7 11.Qd4+ Kf8
12.Qd7 Ne5+ 13.Kg2 (21.658.721) 3453

19/29 0:06 +0.02++ 1...Nf7 2.Kh3 Ne5 3.g4 Nh6 4.gxh5 gxh5
5.f4 Ng6 6.Kg3 f5 7.Qd5 Ne7 8.Qd8 Nhg8 (23.303.860) 3434

19/32 0:07 -5.49 1...Nf7 2.Kh3 N5d6 3.g4 hxg4+ 4.fxg4 Ne4
5.Qb4 Nfd6 6.Qb8 Nf7 7.Qc7 Nf2+
8.Kg3 Ne4+ 9.Kf4 Ned6 10.Qc5 g5+
11.hxg5 fxg5+ 12.Ke3 Kf6 13.Kd4 Ke6
14.Qd5+ Kf6 (24.581.464) 3440

20/49 0:25 -5.32 1...Nf7 2.Qc7 Kf8 3.Qc6 Kg7 4.Qb6 g5
5.Qa5 Ne3+ 6.Kh3 Kg6 7.Qb6 Nf5 8.Qb3 (103.993.647) 4040

21/49 0:28 -5.60 1...Nf7 2.Kh3 N5h6 3.Qa7 g5 4.hxg5 fxg5
5.Qd4+ Kg6 6.Kg2 Nf5 7.Qb6+ N5d6
8.Qb1+ Kf6 9.Qb2+ Kf5 10.Qe2 Ne5
11.g4+ hxg4 12.Qc2+ Kf6 13.Kf2 (113.052.504) 3981

22/49 0:33 -5.65 1...Nf7 2.Kh3 N5h6 3.Qa7 g5 4.hxg5 fxg5
5.Qd4+ Kg6 6.Kg2 Ng8 7.Qe4+ Kg7
8.Qc2 Nf6 9.f4 Ng4 10.Qd3 Kf6
11.Qd4+ Kg6 12.Qe4+ Kg7 13.Qd3 (133.284.237) 3977

23/49 0:38 -5.65 1...Nf7 2.Kh3 N5h6 3.Qa7 g5 4.hxg5 fxg5
5.Qd4+ Kg6 6.Kg2 Ng8 7.Qe4+ Kg7
8.Qc2 Nf6 9.f4 Ng4 10.Qd3 Kf6
11.Qd4+ Kg6 12.Qe4+ Kg7 13.Qd3 (153.953.862) 4024

24/49 1:05 -5.75 1...Nf7 2.Kh3 N5h6 3.Qa7 g5 4.hxg5 fxg5
5.Qd4+ Kg6 6.Kg2 Nf5 7.Qb6+ Kg7
8.Qe6 N5h6 9.f4 Ng4 10.Qc6 Ne3+
11.Kf3 Nf5 12.Qc5 g4+ 13.Kg2 Kf6
14.Qa5 (260.400.494) 3968

25/49 1:06 -5.75 1...Nf7 2.Kh3 N5h6 3.Qa7 g5 4.hxg5 fxg5
5.Qd4+ Kg6 6.Kg2 Nf5 7.Qb6+ Kg7
8.Qe6 N5h6 9.f4 Ng4 10.Qc6 Ne3+
11.Kf3 Nf5 12.Qc5 g4+ 13.Kg2 Kf6
14.Qa5 (261.796.279) 3966

26/49 1:06 -5.75 1...Nf7 2.Kh3 N5h6 3.Qa7 g5 4.hxg5 fxg5
5.Qd4+ Kg6 6.Kg2 Nf5 7.Qb6+ Kg7
8.Qe6 N5h6 9.f4 Ng4 10.Qc6 Ne3+
11.Kf3 Nf5 12.Qc5 g4+ 13.Kg2 Kf6
14.Qa5 (262.347.269) 3965

27/49 1:27 -5.75 1...Nf7 2.Kh3 N5h6 3.Qa7 Nf5 4.g4 N5d6
5.gxh5 gxh5 6.Qg1+ Kh7 7.Qb1+ Kh6
8.Qc2 f5 9.Qc3 Kg6 10.Qd3 Kf6 11.Qd1 Kg6
12.Qg1+ Kf6 (354.291.053) 4067

28/62 2:45 -6.09 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 Ne5
5.Qc7+ Nhf7 6.Qe7 Nc6 7.Qe6 Nce5
8.Kg3 g5 9.h5 Nh6 10.Qe7+ Nef7
11.Qd7 Kg8 12.Qe6 Kg7 13.Kh3 Ne5
14.Qe7+ Nef7 (693.618.902) 4187

29/62 2:57 -6.09 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 Ne5
5.Qc7+ Nhf7 6.Qe7 Nc6 7.Qe6 Nce5
8.Kg3 g5 9.h5 Nh6 10.Qe7+ Nef7
11.Qd7 Kg8 12.Qe6 Kg7 13.Kh3 Ne5
14.Qe7+ Nef7 (749.683.898) 4216

30/62 3:40 -6.08 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 Ne5
5.Qc7+ Nhf7 6.Qe7 Nc6 7.Qe6 Nce5
8.Kg3 g5 9.h5 Nh6 10.Qe7+ Nef7
11.Qd7 Kg8 12.Qc6 Kg7 13.Qc7 Kf8
14.Qc5+ Ke8 (921.093.465) 4178

31/62 3:40 -6.08 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 Ne5
5.Qc7+ Nhf7 6.Qe7 Nc6 7.Qe6 Nce5
8.Kg3 g5 9.h5 Nh6 10.Qe7+ Nef7
11.Qd7 Kg8 12.Qc6 Kg7 13.Qc7 Kf8
14.Qc5+ Ke8 (922.448.088) 4177

32/62 3:51 -6.08 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 Ne5
5.Qc7+ Nhf7 6.Qe7 Nc6 7.Qe6 Nce5
8.Kg3 g5 9.h5 Nh6 10.Qe7+ Nef7
11.Qd7 Kg8 12.Qc6 Kg7 13.Qc7 Kf8
14.Qc5+ Ke8 (967.745.485) 4178

33/69 5:32 -6.08 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 Ne5
5.Qc7+ Nhf7 6.Qe7 Nc6 7.Qe6 Nce5
8.Kg3 g5 9.h5 Nh6 10.Qe7+ Nef7
11.Qd7 Kg8 12.Qc6 Kg7 13.Qc7 Kf8
14.Qc5+ Ke8 (1.401.171.051) 4212

34/92 6:47 -6.08 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 Ne5
5.Qc7+ Nhf7 6.Qe7 Nc6 7.Qe6 Nce5
8.Kg3 g5 9.h5 Nh6 10.Qe7+ Nef7
11.Qd7 Kg8 12.Qc6 Kg7 13.Qc7 Kf8
14.Qb6 Ke7 (1.730.118.418) 4250

35/97 11:48 -6.08 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 Ne5
5.Qc7+ Nhf7 6.Qe7 Nc6 7.Qe6 Nce5
8.Kg3 g5 9.h5 Nh6 10.Qe7+ Nef7
11.Qd7 Kg8 12.Qc6 Kg7 13.Qc7 Kf8
14.Qb6 Ke7 (2.979.518.881) 4204

36/97 11:54 -6.08 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 Ne5
5.Qc7+ Nhf7 6.Qe7 Nc6 7.Qe6 Nce5
8.Kg3 g5 9.h5 Nh6 10.Qe7+ Nef7
11.Qd7 Kg8 12.Qc6 Kg7 13.Qc7 Kf8
14.Qb6 Ke7 (3.003.001.424) 4201

[ernest]

Thanks for testing, Eelco!
So, White win, or draw?

[Eelco de Groot]

Hello Ernest,

I would like it if it turned out to be won, then I could say the evals weren't wrong after all, but at the moment I'm not very sure... Rainbow Serpent isn't helping much because it can't look deeper than 36 plies at the moment. I do think Peter Martan usually gets it right and he thinks it's a win! Maybe somebody else will post a deep analysis. With such a big score I don't think it is enough just to trust in the no change in evaluation as convincing evidence...

Regards, Eelco

[peter]

I do think Peter Martan usually gets it right and he thinks it's a win! Maybe somebody else will post a deep analysis. With such a big score I don't think it is enough just to trust in the no change in evaluation as convincing evidence...

Hi Eelco!
If you have such an opionion about me, I have to live up to it!

Problem is, the mate in about 40 is too far away for engines even with 6MOB, to keep the exact score in hash in backward analysis to starting position. Stockfish is best once again, especially with gtbs support, pity there are no Gaviota's 6MOB, are there?
I just thought it to be a good example to show that eval progress is not to be counted in any position at all by letting engine compute by its own at the starting position only, even endlessly, when outcome, even a forced mate, ist out of their horizon.
Evalcurves over time and plies are the very best meaning of testing to me, but at least one relevant variant has to be known by the tester, cause if you don't know more than the engines, how do you want to test it?


[Event "Forum"]
[Site "?"]
[Date "2012.09.04"]
[Round "?"]
[White "Haupt"]
[Black "Bauer"]
[Result "1-0"]
[Annotator "Martan,Peter"]
[SetUp "1"]
[FEN "7n/6k1/5pp1/2Q2n1p/7P/5PP1/6K1/8 b - - 0 1"]
[PlyCount "54"]

1... Nh6 (1... Nf7 2. Kh3 N5h6 3. g4 hxg4+ 4. fxg4 g5 5. h5 Ne5 6. Qe7+ Nef7 7.
Qd7 Kf8 8. Qe6 Ng8 9. Qc8+ Kg7 10. Qc7 Kf8 11. Kg3 Ne7 12. Qc5 Ke8 13. Qc3 Ng8
14. Qc8+ Nd8 15. Qb8 Nh6 16. Qb5+ Ke7 17. Qb4+ Kf7 (17... Ke8 18. Qa4+ Kf8 19.
Qd4 Ke7 20. Qa7+ Kd6 21. Qg7 Ndf7 22. Qxf6+ Kc5 23. Qe6 Kd4 24. Qc6 Ke5 25. Kf3
Kd4 26. Qe4+ Kc5 27. Qe3+ Kc4 28. Qd2 Ne5+ 29. Ke4 Nhf7 30. Qd4+ Kb3 31. Qxe5
Nxe5 32. Kxe5 {#11}) 18. Qd6 Ke8 19. Qxf6 Ndf7 20. Kf3 Kf8 21. Qc3 Ke7 22. Ke4
Nd6+ 23. Kd5 Ndf7 24. Qd4 Ke8 25. Ke6 Kf8 26. Kf6 Ng8+ 27. Ke6 Ngh6 28. Qf6 Kg8
29. Ke7 Ne5 30. Qxe5 Nf7 31. Qb8+ Kg7 32. Qf8+ Kh7 33. Kf6 Ne5 34. Qg7#) 2.
Qa7+ N8f7 3. Kh3 Kf8 (3... Kh7 4. g4 hxg4+ 5. fxg4 g5 6. Qd7 gxh4 7. Kxh4 Kg6
8. Qe8 Kg7 9. Qe7 Ng8 10. Qd7 Kg6 11. Qe8 Kg7 12. Kh5 Ne5 13. Qe6 Nf7 14. Qd7
Kf8 15. Qc8+ Kg7 16. Qe6 Ng5 17. Qd7+ Nf7 18. Qa7 Ngh6 19. g5 fxg5 20. Qd4+ {
#27}) 4. g4 hxg4+ 5. fxg4 g5 6. Kg3 Kg7 7. h5 Kf8 8. Kf3 Nd6 (8... Kg8 9. Ke4
Kg7 10. Qc7 Nxg4 11. Kf5 Ngh6+ 12. Ke6 g4 13. Qe7 f5 14. Qc7 Kf8 15. Kd7 Ng8
16. Qc5+ Kg7 17. Qxf5 Kf8 18. Qf4 Ngh6 19. Qd4 g3 20. Qc5+ Kg7 21. Qc3+ Kg8 22.
Qxg3+ Kf8 {#10}) 9. Qc7 Nhf7 (9... Ne8 10. Qh7 Nf7 11. h6 Ne5+ 12. Ke4 Nf7 13.
Kd4 Nc7 14. Qg6 Ke7 15. h7 Ne6+ 16. Ke4 Nd6+ 17. Kf3 Nf7 18. Qxf7+ Kxf7 19.
h8=Q Ke7 20. Ke4 Nf4 21. Qc8 Kd6 22. Qc3 Kd7 23. Qxf6 {#9}) 10. h6 Ne8 11. Qh2
Ne5+ 12. Kg3 Nf7 13. Qh5 Nh8 14. Kf3 Nd6 15. Qh2 Ndf7 16. Qb8+ Ke7 17. Qc7+ Ke6
18. Qb6+ Nd6 19. h7 Nhf7 20. Qb8 Ne5+ 21. Ke3 Ndf7 22. Ke4 Nd6+ (22... Ke7 23.
Kd5 Ng6 24. Qb4+ Kd7 25. Qa4+ Ke7 26. Qe4+ Nfe5 27. Qb4+ Kf7 28. Qb7+ Ke8 29.
Qc8+ Kf7 30. Qg8+ Ke7 31. Qe6+ Kf8 32. Qxf6+ Nf7 33. Qxg6 Ke7 34. Qg7 Ke8 35.
h8=Q+ Nxh8 36. Ke6 Ng6 37. Qd7+ Kf8 38. Qf7#) 23. Qxd6+ Kxd6 24. h8=Q Nd7 25.
Qd8 Kc6 26. Kf5 Kd6 27. Qe8 Ne5 28. Kxf6 {#13} 1-0

[peter]

... the mate in about 40

Reviewing MV from .pgn above, found a shorter way with 12.Ke2 instead of Kg3:



1... Nh6 2. Qa7+ N8f7 3. Kh3 Kf8 4. g4 hxg4+ 5. fxg4 g5 6. Kg3 Kg7 7. h5 Kf8 8. Kf3 Nd6 9. Qc7 Nhf7 10. h6 Ne8 11. Qh2 Ne5+ 12. Ke2 Nxg4 13. Qh3 Nxh6 14. Qxh6+
{#19}


So this was mate in 33,
but then you would have to review the variants after 1...Nf7 instead of Nh6 again to be sure, if there wasn't any longer there.

Up to 11. move you can take back the mate score with Houdini and 6MOB:

81: Haupt - Bauer, Forum
4nk2/2Q2n2/5p1P/6p1/6P1/5K2/8/8 w - - 0 1

Analysis by Houdini 2.0c Pro x64L:

11.Dh2 Se5+ 12.Ke2 Sxg4 13.Dh3 Sxh6 14.Dxh6+ Ke7
+- (#23) Tiefe: 38/60 00:09:52 16932mN, tb=36387

[Houdini]

Found this in the German CSS-Forum
http://forum.computerschach.de/cgi-bin/ ... l?tid=4956

How long does it take for your program (and machine) to see a progress in evaluation - if any - in this endgame position?

[d]7n/6k1/5pp1/2Q2n1p/7P/5PP1/6K1/8 b - - 0 0

It takes very long, but in the end there is progress - it looks like a win.

Code: Select all

7n/6k1/5pp1/2Q2n1p/7P/5PP1/6K1/8 b - -

Engine: Houdini 3 (4096 MB)
by Robert Houdart

40/79  26:33  +4.71    1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5 
                       5.h5 Ne5 6.Qe7+ Nef7 7.Qd7 Kf8 8.Qe6 Kg7 
                       9.Kg3 Ne5 10.Qe7+ Nef7 11.Kf3 f5 
                       12.gxf5 Nxf5 13.Qb7 Kf6 14.Qc6+ Ke5 (11.409.100.936) 7158 

41/79  33:14  +4.71    1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5 
                       5.h5 Ne5 6.Qe7+ Nef7 7.Qd7 Kf8 8.Qe6 Kg7 
                       9.Kg3 Ne5 10.Qe7+ Nef7 11.Kf3 f5 
                       12.gxf5 Nxf5 13.Qb7 Kf6 14.Qc6+ Ke5 (14.538.768.473) 7288 

42/79  48:42  +4.71    1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5 
                       5.h5 Ne5 6.Qe7+ Nef7 7.Qd7 Kf8 8.Qe6 Kg7 
                       9.Kg3 Ne5 10.Qe7+ Nef7 11.Kf3 f5 
                       12.gxf5 Nxf5 13.Qb7 Kf6 14.Qc6+ Ke5 (21.761.904.559) 7446 

43/81  74:29  +4.73    1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5 
                       5.h5 Ne5 6.Qe7+ Nef7 7.Qd7 Kf8 8.Qe6 Kg7 
                       9.Kg3 Ne5 10.Qe7+ Nef7 11.Kf3 f5 
                       12.gxf5 Nxf5 13.Qb7 Kf6 14.Qc6+ Ke5 (33.869.186.216) 7577 

44/93  93:01  +4.93--  1...Nf7 2.Kh3 (42.294.463.637) 7577 

44/93  136:11 +5.32--  1...Nf7 2.Kh3 (64.204.083.722) 7856 

44/99  344:59 +6.36    1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5 
                       5.h5 Ne5 6.Qe7+ Nef7 7.Qd7 Kf8 8.Qe6 Ng8 
                       9.Qc8+ Kg7 10.Qd7 Kf8 11.Kg3 Ngh6 
                       12.Qe6 Kg7 13.Kg2 Kf8 14.Kf2 Kg7 (162.934.850.135) 7871 

[Eelco de Groot]

Hi Peter, thanks very much for all the analysis!

I think that makes the case for a win for White, and that interactive analysis is still sometimes better than just letting an engine compute everything from the start position. I will have to look at the variations deeper to see if the engines do things differently here and there. Now we also have the output from Houdini 3 (Or is it Telescopini 3?) to compare. Strangely how it looks very similar to what I just got on my monitor In this case I agree with Chris Whittington that sometimes the engines look at positions in such identical ways that they will probably both miss better continuations. Well that is just like in human chess, only there you have more short term blunders and probably a more human positional understanding when the game is not yet decided by those blunders.

I finally got a build that could take the search over the new barrier of 36 plies that seemed to have replaced the barrier of 28 plies Iteration 42 took quite a while and confirmed the eval of iteration 41, the nodes per second seemed suddenly lower again, very strange, I thought it possibly could mean that every now and then we are a long time in quiescence search (higher nodes per second, less hash hits) But the position is not so tactically complex that I see why there would be any very long quiescence searches at all So that is probably not what I am seeing. I have no good explanation for this, but I also have not really studied how Stockfish does count its nodes in a multithreaded search.

This is my build 171, as you can see a very similar evaluation to Houdini's 6.36 but here taking a negative eval from Black's point of view. Less than 1% difference in eval! It is just a coincidence of course, for instance there is no scaling of Rainbow Serpents output towards the value of a pawn in the endgame that I think Houdini has.

Eelco

7n/6k1/5pp1/2Q2n1p/7P/5PP1/6K1/8 b - -

Engine: Rainbow Serpent 2.2.4 Build 171 (Q6700 4 threads, 512 MB)
by Tord Romstad, Marco Costalba and Joona Kiiski

15/23 0:01 +0.19++ 1...Nf7 2.Kh3 N5d6 3.g4 hxg4+ 4.fxg4 f5
5.Qd4+ Kh7 6.Qf4 fxg4+ 7.Kxg4 Nf5
8.h5 N7d6 9.Kf3 (2.697.700) 2614

15/26 0:01 -5.17 1...Nf7 2.Kh3 N5h6 3.Qc3 Ne5 4.Qc7+ Nhf7
5.Kg2 Kf8 6.f4 Ng4 7.Kf3 f5 8.Ke2 Kg7
9.Qe7 Nf6 (5.224.615) 3034

16/28 0:03 -5.28 1...Nf7 2.Qc7 Kf8 3.Qc8+ Kg7 4.Qc6 Ne5
5.Qe6 Nf7 6.Qd7 N5d6 7.Kf2 Nf5 8.Qc7 Kf8
9.Qc8+ Ke7 10.g4 hxg4 11.fxg4 Nxh4
12.Qc7+ Ke6 13.Qc4+ Ke7 14.Qc5+ Nd6 (11.038.733) 3480

17/28 0:03 +0.10++ 1...Nf7 2.Qc7 Kf8 3.Kh3 Ne5 4.g4 hxg4+
5.fxg4 Ne7 6.h5 Nf5 7.gxf5 gxh5 (11.346.951) 3478

17/39 0:06 -5.17 1...Nf7 2.Qc7 N5h6 3.Qa7 Nf5 4.Kh3 N5d6
5.Qc5 Kh7 6.Qd4 Kg7 7.g4 Nb5 8.Qb6 Nfd6
9.Qd8 Nf7 10.Qe7 hxg4+ 11.fxg4 Nbd6
12.Kg3 f5 13.Qa7 fxg4 14.Qa1+ Kf8 (21.304.366) 3514

18/39 0:09 -5.43-- 1...Nf7 2.Kh3 N5d6 3.g4 hxg4+ 4.fxg4 Ne4
5.Qd5 Nfd6 6.Kg2 f5 7.gxf5 gxf5
8.Qe6 Nb5 9.Qxf5 Ng5 (33.326.090) 3524

18/39 0:09 -5.43 1...Nf7 2.Kh3 N5d6 3.g4 hxg4+ 4.fxg4 Ne4
5.Qd5 Ned6 6.Kg3 f5 7.Qd4+ Kf8 8.Kh3 fxg4+
9.Kxg4 Ke7 10.Kf4 Nf5 11.Qe4+ Kf6
12.h5 g5+ 13.Kf3 Ne5+ 14.Kg2 (33.437.443) 3521

19/39 0:10 +0.01++ 1...Nf7 2.Kh3 Ne5 3.Qa7+ Kf8 4.g4 Ne7
5.gxh5 gxh5 6.Qc5 N5g6 7.Kg3 Nh8
8.Qxh5 Ng8 (35.541.684) 3499

19/39 0:12 -5.39 1...Nf7 2.Kh3 N5h6 3.Qa7 g5 4.hxg5 fxg5
5.Qd4+ Kg6 6.Qb6+ Kg7 7.Qa6 Ng8
8.Qc6 Nf6 9.f4 Ng4 10.fxg5 Nge5
11.Qf6+ Kg8 (43.935.847) 3452

20/39 0:13 +0.06++ 1...Nf7 2.Kh3 N5h6 3.Qa7 g5 4.hxg5 fxg5
5.Qd4+ Kg6 6.Qb6+ Kg7 7.Qc5 Kg6
8.Qc6+ Kh7 9.Qf6 Kg8 10.Qg6+ Kf8
11.Qxh5 g4+ 12.fxg4 Nf5 (47.485.217) 3467

20/39 0:14 -5.55 1...Nf7 2.Kh3 Ne5 3.g4 hxg4+ 4.fxg4 Nh6
5.Qe7+ Nhf7 6.Kg3 Nc6 7.Qc7 Nce5
8.Qc5 Nd3 9.Qd5 Nde5 10.Qb5 g5
11.Qb7 gxh4+ 12.Kh2 (50.865.472) 3506

21/39 0:14 -5.65 1...Nf7 2.Kh3 Ne5 3.g4 hxg4+ 4.fxg4 Nh6
5.Qe7+ Nhf7 6.Kg3 Nc6 7.Qc7 Nce5
8.Qb7 Kf8 9.Qc8+ Kg7 10.Qe6 Nd3
11.Qe7 Nde5 12.g5 fxg5 13.hxg5 Kg8
14.Kf4 Kg7 (52.233.861) 3513

22/39 0:18 -5.61 1...Nf7 2.Kh3 Ne5 3.g4 hxg4+ 4.fxg4 Nh6
5.Qe7+ Nhf7 6.Kg3 Nc6 7.Qc7 Nce5
8.Qb7 Kf8 9.Qc8+ Kg7 10.Qe6 Nd3
11.Qc6 Nde5 12.Qc5 Nd8 13.Qe7+ Ndf7
14.Qe6 (67.000.059) 3566

23/40 0:31 -5.61 1...Nf7 2.Kh3 Ne5 3.g4 hxg4+ 4.fxg4 Nh6
5.Qe7+ Nhf7 6.Kg3 Nc6 7.Qc7 Nce5
8.Qb7 Kf8 9.Qc8+ Kg7 10.Qe6 Nd3
11.Qc6 Nde5 12.Qc5 Nd8 13.Qe7+ Ndf7
14.Qe6 (119.506.333) 3792

24/40 0:32 -5.61 1...Nf7 2.Kh3 Ne5 3.g4 hxg4+ 4.fxg4 Nh6
5.Qe7+ Nhf7 6.Kg3 Nc6 7.Qc7 Nce5
8.Qb7 Kf8 9.Qc8+ Kg7 10.Qe6 Nd3
11.Qc6 Nfe5 12.Qb7+ Nf7 13.Qd5 Nde5
14.Qe6 (122.368.699) 3797

25/40 0:33 -5.61 1...Nf7 2.Kh3 Ne5 3.g4 hxg4+ 4.fxg4 Nh6
5.Qe7+ Nhf7 6.Kg3 Nc6 7.Qc7 Nce5
8.Qb7 Kf8 9.Qc8+ Kg7 10.Qe6 Nd3
11.Qc6 Nfe5 12.Qb7+ Nf7 13.Qe7 Nde5
14.Qe6 (128.391.807) 3811

26/46 0:45 -5.95-- 1...Nf7 2.Kh3 Ne5 3.g4 hxg4+ 4.fxg4 Nh6
5.Qc7+ Nef7 6.Qe7 g5 7.h5 f5 8.Qe6 f4
9.Kg2 Kf8 10.Kf2 Kg7 11.Kf3 Ng8
12.Qf5 Ngh6 13.Qg6+ Kf8 14.Qe6 Kg7 (178.860.152) 3972

26/46 0:46 -5.93 1...Nf7 2.Kh3 Ne5 3.g4 hxg4+ 4.fxg4 Nh6
5.Qc7+ Nef7 6.Qe7 Ng8 7.Qe8 Ne5
8.Qe6 Nf7 9.Kg3 Nd8 10.Qd7+ Nf7
11.Qd4 Ngh6 12.Qa7 Kh8 13.Qc5 Ne5
14.Qc8+ Kh7 (182.348.125) 3963

27/46 0:48 -5.93 1...Nf7 2.Kh3 Ne5 3.g4 hxg4+ 4.fxg4 Nh6
5.Qc7+ Nef7 6.Qe7 Ng8 7.Qe8 g5 8.h5 Ne5
9.Kg3 Nf7 10.Qd7 Ngh6 11.Qe6 Ne5
12.Qe7+ Nef7 13.Kf3 Ng8 14.Qe6 Ngh6 (194.440.158) 3970

28/47 1:04 -5.93 1...Nf7 2.Kh3 Ne5 3.g4 hxg4+ 4.fxg4 Nh6
5.Qc7+ Nef7 6.Qe7 Ng8 7.Qe8 g5 8.h5 Ne5
9.Kg3 Nf7 10.Qd7 Ngh6 11.Qe6 Ne5
12.Qe7+ Nef7 13.Kf3 Ng8 14.Qe6 Ngh6 (263.303.478) 4085

29/57 5:10 -5.93 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qc7 Kf8 8.Qb6 Kg7
9.Qa7 Kh7 10.Qc7 Kg7 11.Kg3 Kf8
12.Qc6 Kg7 13.Qc5 Ng8 14.Qc7 Ngh6 (1.329.377.073) 4278

30/57 5:18 -5.93 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qc7 Kf8 8.Qb6 Kg7
9.Qa7 Kh7 10.Qc7 Kg7 11.Kg3 Ng8
12.Qd7 Ngh6 13.Qe6 Ne5 14.Qe7+ Nef7 (1.375.979.833) 4317

31/57 5:21 -5.93 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qc7 Kf8 8.Qb6 Kg7
9.Qa7 Kh7 10.Qc7 Kg7 11.Kg3 Ng8
12.Qd7 Ngh6 13.Qe6 Ne5 14.Qe7+ Nef7 (1.385.599.037) 4315

32/57 6:31 -5.93 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qc7 Kf8 8.Qb6 Kg7
9.Qa7 Kf8 10.Qb6 (1.908.641.072) 4873

33/58 7:42 -5.93 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qc7 Kf8 8.Qb6 Kg7
9.Qa7 Kf8 10.Qb6 (2.443.816.559) 5278

34/65 9:38 -5.93 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qc7 Kf8 8.Qb6 Kg7
9.Qa7 Kg8 10.Qa8+ Kg7 11.Qb7 Kf8
12.Qb6 (3.060.731.945) 5293

35/65 9:46 -5.93 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qc7 Kf8 8.Qb6 Kg7
9.Qa7 Kf8 10.Qb6 (3.098.332.223) 5284

36/68 11:14 -5.93 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qc7 Kf8 8.Qb6 Kg7
9.Qa7 Kg8 10.Qb8+ Kg7 11.Qc7 (3.621.822.569) 5371

37/69 13:22 -5.93 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qc7 Kf8 8.Qb6 Kg7
9.Qa7 Kg8 10.Qc5 Ne5 11.Qe7 Nexg4
12.Qe1 Ne5 13.Qd1 (4.438.386.537) 5527

38/69 18:21 -5.93 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qc7 Kf8 8.Qb6 Kg7
9.Qa7 Kf8 10.Qc5+ Ke8 (6.511.491.928) 5912

39/69 18:27 -5.93 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qc7 Kf8 8.Qb6 Kg7
9.Qa7 Nf5 10.gxf5 (6.537.449.980) 5902

40/69 21:27 -5.93 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qc7 Kf8 8.Qb6 Kg7
9.Qb4 Ne5 10.Qe7+ (7.804.531.522) 6060

41/75 34:22 -6.30 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qc7 Kf8 8.Qc6 Kg7
9.Qd7 Kg8 10.Qe6 Kg7 11.Kg3 Ne5 (12.831.274.458) 6222

42/80 277:48 -6.30 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5
5.h5 Ne5 6.Qe7+ Nef7 7.Qc7 Kf8 8.Qc6 Kg7
9.Qd7 (82.496.227.702) 4949

[Eelco de Groot]

I did let it run a bit longer but that is more interesting for myself to test that engine still has not crashed. Further progress (of the evaluation I mean) is probably beyond the horizon.

43/89 1392:30 -6.30 1...Nf7 2.Kh3 N5h6 3.g4 hxg4+ 4.fxg4 g5 5.h5 Ne5 6.Qe7+ Nef7 7.Qc7 Kf8 8.Qc6 Kg7 (408.881.393.642) 4893
44/89 2111:04 -6.06 1...Nf7 2.Kh3 N5d6 3.Qd4 Nb5 4.Qe3 Nbd6 5.Qe6 f5 6.Qe2 Nc8 7.Qe8 Ncd6 8.Qe6 g5 9.Qd5 g4+ 10.fxg4 hxg4+ 11.Kh2 (628.395.195.453) 4961
45/89 2305:40 -6.14 1...Nf7 2.Kh3 N5d6 3.Qd4 Nf5 4.Qb4 g5 5.hxg5 Nxg5+ 6.Kg2 Kf7 7.Qc5 Ng7 8.Qa7+ Kg6 9.Qe7 Nf7 10.Qe4+ Kh6 11.Qe3+ Kg6 12.Qd3+ Kh6 13.Kh3 (686.186.506.865) 4960
46/89 2745:36 -6.18 1...Nf7 2.Kh3 N5d6 3.Qd4 Nf5 4.Qb4 g5 5.hxg5 Nxg5+ 6.Kg2 Kf7 7.Qc5 Ng7 8.Qa7+ Kg6 9.Qe7 N5e6 10.Qb7 (807.132.076.749) 4899
47/89 3901:54 -6.21 1...Nf7 2.Kh3 N5d6 3.g4 hxg4+ 4.fxg4 Kg8 5.Qe3 f5 6.Qe6 fxg4+ (1291.237.432.952) 5515

Eelco

[peter]

I did let it run a bit longer

I think this could be of interest for you too, here is Stockfish2.2.1 Ofry using 5mob Gaviota's:

Code: Select all

81: Haupt - Bauer, Forum
7n/6k1/5pp1/2Q2n1p/7P/5PP1/6K1/8 b - - 0 1

Analysis by Stockfish 2.2.1 ofry 64bit SSE4.2T:

1...Sf7 2.Da7 g5 3.hxg5 fxg5 4.Dc5 
  +-  (4.87 --)   Tiefe: 7/13   00:00:13  4kN
...
1...Sf7 2.Da7 Kg8 3.f4 Kg7 4.Kf3 Kf8 5.Kg2 Kg8 6.Kf2 Kg7 7.Kf3 
  +-  (5.01)   Tiefe: 10/16   00:00:13  41kN
...
1...Sf7 2.Kh3 S5d6 3.g4 hxg4+ 4.fxg4 Kf8 5.Kg3 Kg8 6.Kf4 Kg7 7.Dd4 Sb5 8.Dd3 Sc7 9.Dd7 Sa6 10.De7 
  +-  (6.06 --)   Tiefe: 17/29   00:00:16  1256kN, tb=33
...
1...Sf7 2.Kh3 S5d6 3.g4 hxg4+ 4.fxg4 Se4 5.Dd4 Sfd6 6.g5 Kf7 7.gxf6 Ke6 8.Kg4 Sxf6+ 9.Kg5 Sf7+ 10.Kxg6 Se5+ 11.Kh6 Kf5 12.Df2+ Ke6 13.De2 Kf5 14.Kg7 Se8+ 15.Kf8 Sd6 16.h5 Sdc4 
  +-  (6.40)   Tiefe: 20/47   00:00:28  5645kN, tb=55
1...Sf7 2.Kh3 S5h6 3.g4 hxg4+ 4.fxg4 g5 5.h5 Se5 6.De7+ Sef7 7.Dd7 Sg8 8.Dc7 Sgh6 9.Db6 Se5 10.Db7+ Sef7 11.Dc7 Sg8 12.Db6 Sgh6 13.Kg3 Sg8 
  +-  (6.32 ++)   Tiefe: 21/47   00:00:30  6221kN, tb=62
...
1...Sf7 2.Kh3 S5h6 3.g4 hxg4+ 4.fxg4 g5 5.h5 Se5 6.De7+ Sef7 7.Dd7 Sg8 8.Db7 Sgh6 9.De7 Sg8 10.Dc7 Sgh6 11.Dd7 
  +-  (6.38 --)   Tiefe: 40/58   00:06:41  166mN, tb=125
1...Sf7 2.Kh3 S5d6 3.g4 hxg4+ 4.fxg4 Se4 5.Dc2 Sfd6 6.Dc6 Sf2+ 7.Kg2 Sfe4 8.Dd7+ Kh6 9.De7 f5 10.Df8+ Kh7 11.g5 
  +-  (6.46 --)   Tiefe: 40/58   00:07:51  199mN, tb=125
1...Sf7 2.Kh3 Se5 3.g4 Sh6 4.g5 Shf7 5.De7 fxg5 6.hxg5 Sd3 7.Df6+ Kf8 8.f4 Sf2+ 9.Kh4 Sd3 10.Dd4 
  +-  (6.58 --)   Tiefe: 40/61   00:09:59  252mN, tb=271
1...Sf7 2.Kh3 S5h6 3.g4 hxg4+ 4.fxg4 g5 5.h5 Se5 6.De7+ Sef7 7.Dc7 Sg8 8.Dd7 Sgh6 9.Kg3 Sg8 10.Kf3 Sgh6 11.Ke3 Kf8 12.De6 Kg7 13.Kd4 Sg8 14.Dd7 Sgh6 15.Kd5 
  +-  (6.76 --)   Tiefe: 40/61   00:11:58  307mN, tb=1828
1...Sf7 2.Kh3 S5h6 3.g4 hxg4+ 4.fxg4 g5 5.h5 Se5 6.De7+ Sef7 7.Dc7 Sg8 8.Dd7 Sgh6 9.Kg3 Kh8 10.De6 Kg7 11.Kg2 Se5 12.De7+ Sef7 13.Kf3 f5 14.gxf5 Sxf5 15.Dd7 Kf6 16.Dc6+ S5d6 17.Da6 Ke5 18.Kg4 Kf6 19.Df1+ Ke7 20.De2+ Kf8 21.h6 Se8 22.De6 
  +-  (6.76)   Tiefe: 40/62   00:12:18  325mN, tb=1901
1...Sh6 2.Kh3 f5 3.De5+ Kh7 4.Kg2 S8f7 5.Df6 Sh8 6.Kf2 Sg8 7.De6 Kg7 8.g4 fxg4 9.fxg4 Sf7 
  +-  (6.68 ++)   Tiefe: 41/73   00:22:13  559mN, tb=1911
1...Sh6 2.Kh3 f5 3.De5+ Kh7 4.Kg2 S8f7 5.Df6 Sh8 6.Kf2 S8f7 7.Ke3 Sh8 8.Dd4 S8f7 9.Kf4 Sg8 10.g4 fxg4 11.fxg4 hxg4 12.Kxg4 Sgh6+ 13.Kf4 Sg8 14.Dc3 Kh6 15.Kg4 Kh7 16.h5 g5 17.Dd4 Sgh6+ 18.Kf3 Kg8 19.Df6 Kf8 20.Kg3 Sg8 21.Dc3 Sgh6 22.Df6 
  +-  (6.84 --)   Tiefe: 41/73   00:22:31  572mN, tb=2063
1...Sh6 2.Kh3 f5 3.De5+ Kh7 4.Kg2 S6f7 5.Dc7 Kh6 6.Dd7 Se5 7.De7 Sef7 8.De6 Kh7 9.Dd7 Kg7 10.Dd4+ Kh7 11.Kf2 Sh6 12.Ke3 S6f7 13.Da7 Kg7 14.Dd7 Kh7 15.Dc7 
  +-  (6.96 --)   Tiefe: 41/73   00:24:07  623mN, tb=2183
1...Sh6 2.Kh3 Kf7 3.Dc7+ Kf8 4.g4 hxg4+ 5.fxg4 S8f7 6.g5 fxg5 7.Dc5+ Kg7 8.hxg5 Sf5 9.Kg4 Kh7 10.Db6 S7d6 11.Dc7+ Kg8 12.Dd8+ Kh7 13.Dd7+ Kh8 14.De6 Kh7 15.Kf4 Kg7 16.Df6+ Kh7 17.De6 
  +-  (7.15 --)   Tiefe: 41/73   00:27:11  700mN, tb=2197
1...Sh6 2.Kh3 S8f7 3.g4 hxg4+ 4.fxg4 g5 5.h5 Se5 6.De7+ Sef7 7.Dd7 Kf8 8.De6 Sg8 9.Dc8+ Kg7 10.Dd7 Kf8 11.Kg3 Se5 12.Dh7 Sf7 13.De4 Sgh6 14.De6 Kg7 15.De8 Sg8 16.Dc6 Se5 17.Dc7+ Kf8 18.Dc8+ Kg7 19.Db7+ Kh8 20.Kh3 Sd3 21.Dc8 Se5 22.Df8 Sd7 23.Db4 Se5 24.Dc5 Kg7 25.Kg3 Kh8 26.Dc7 Sh6 27.Dd8+ Sg8 28.Df8 Kh7 29.Kh3 Sd3 30.Df7+ Kh8 31.h6 Sxh6 32.Dxf6+ Kh7 33.De7+ Kh8 34.Dd8+ 
  +-  (6.66)   Tiefe: 41/73   00:28:48  763mN, tb=2197
1...Sh6 2.Kh3 S8f7 3.g4 hxg4+ 4.fxg4 g5 5.h5 Se5 6.De7+ Sef7 7.Dd7 Kf8 8.De6 Sg8 9.Kg3 Kg7 10.Dd7 Kf8 11.De6 
  +-  (6.74 --)   Tiefe: 42/73   00:29:01  773mN, tb=2198
1...Sh6 2.Kh3 S8f7 3.g4 hxg4+ 4.fxg4 g5 5.h5 Kg8 6.Dc8+ Kg7 7.Dd7 Kf8 8.De6 Kg7 9.De7 Sg8 10.Dc7 Kf8 11.Kg3 Sgh6 12.Db8+ Kg7 13.Da7 Kg8 14.De7 Kg7 15.Kf3 f5 16.gxf5 Sxf5 17.Db7 Sd4+ 18.Kf2 Sf5 19.Db2+ Kf8 20.Kf3 Sg7 21.Db8+ Ke7 22.Db4+ Ke6 23.Kg4 
  +-  (6.82 --)   Tiefe: 42/73   00:29:13  785mN, tb=2198
1...Sh6 2.Kh3 S8f7 3.g4 hxg4+ 4.fxg4 g5 5.h5 Sg8 6.Dc7 Sgh6 7.Kg3 Sg8 8.Kf3 Sgh6 9.De7 f5 10.gxf5 Sxf5 11.Dc5 S5h6 12.Dd4+ Kg8 13.Ke4 g4 14.Dc3 Sd6+ 15.Kf4 Sdf7 16.Db2 Kf8 17.Db1 Sd6 18.Db8+ Se8 19.Db6 Sf7 20.Kxg4 Se5+ 21.Kf4 
  +-  (6.94 --)   Tiefe: 42/73   00:29:30  799mN, tb=2207
1...Sh6 2.Kh3 S8f7 3.g4 hxg4+ 4.fxg4 g5 5.h5 Se5 6.De7+ Sef7 7.Dc7 Sg8 8.Kg2 Sgh6 9.Kf3 Kf8 10.Dc5+ Kg7 11.De7 f5 12.gxf5 Sxf5 13.Dc5 Kf6 14.Db6+ S5d6 15.Dd4+ Se5+ 16.Kg2 Ke6 17.Dc5 Sef7 18.Kh3 Sf5 19.Dc4+ Ke7 20.De4+ Kf6 21.Dc6+ Ke7 22.Dg6 S5h6 23.De4+ Kf8 24.De6 Kg7 25.Dg6+ Kf8 26.De6 
  +-  (7.13 --)   Tiefe: 42/73   00:30:17  836mN, tb=2218
1...Sh6 2.Kh3 S8f7 3.g4 Se5 4.g5 Shf7 5.De7 fxg5 6.hxg5 Kg8 7.De6 h4 8.f4 
  +-  (7.40 --)   Tiefe: 42/73   00:32:54  926mN, tb=2218
1...Sh6 2.Kh3 S8f7 3.g4 hxg4+ 4.fxg4 g5 5.h5 Se5 6.De7+ Sef7 7.Dc7 Kf8 8.Db8+ Kg7 9.Dc7 
  +-  (7.23)   Tiefe: 42/75   00:35:38  1060mN, tb=5219

And Rybka Cluster 64 cores:

Code: Select all

Haupt - Bauer, Forum 2012
7n/6k1/5pp1/2Q2n1p/7P/5PP1/6K1/8 b - - 0 1

Analysis by Rybka Cluster 64 Cores:

1...Sh6 2.Da7+ S8f7 3.Kh3 Kg8 4.g4 hxg4+ 5.fxg4 g5 6.De7 Kg7 7.h5 Sg8 8.Dd7 Kf8 9.Kg3 Se5 10.Dh7 
  +-  (4.57)   Tiefe: 37   00:09:49  2146mN
1...Sh6 2.Da7+ S8f7 3.Kh3 Kf8 4.g4 hxg4+ 5.fxg4 g5 6.Dc5+ Kg8 7.De7 Kg7 8.h5 Sg8 9.Dd7 Kf8 10.Kg3 Se5 
  +-  (4.57)   Tiefe: 37   00:11:37  2147mN
1...Sh6 2.Da7+ S8f7 3.Kh3 Kf8 4.g4 hxg4+ 5.fxg4 g5 6.Dc5+ Kg8 7.De7 Kg7 8.h5 Sg8 9.Dd7 Kf8 10.Kg3 
  +-  (4.57)   Tiefe: 37   00:11:55  2147mN
1...Sh6 2.Da7+ S8f7 3.Kh3 Kf8 4.g4 hxg4+ 5.fxg4 g5 6.Dc5+ Kg8 7.De7 Kg7 8.h5 Sg8 9.Dd7 Kf8 
  +-  (4.57)   Tiefe: 37   00:11:57  2147mN
1...Sh6 2.Da7+ S8f7 3.Kh3 Kf8 4.g4 hxg4+ 5.fxg4 g5 6.Dc5+ Kg8 7.De7 Kg7 8.h5 Sg8 9.Dd7 
  +-  (4.57)   Tiefe: 37   00:11:58  2147mN
1...Sh6 2.Da7+ S8f7 3.Kh3 Kf8 4.g4 hxg4+ 5.fxg4 g5 6.Dc5+ Kg8 7.De7 Kg7 8.h5 Sg8 9.Dd7 
  +-  (4.57)   Tiefe: 38   00:20:32  2147mN
1...Sf7 2.Dc7 Kf8 3.Kh3 S5h6 4.Dd7 Sg8 5.g4 hxg4+ 6.fxg4 g5 7.Dc8+ Kg7 8.h5 Se5 9.De8 Kh8 10.Df8 
  +-  (4.64)   Tiefe: 38   00:23:35  2147mN
1...Sf7 2.Dc7 Kf8 3.Kh3 S5h6 4.Dd7 Sg8 5.g4 hxg4+ 6.fxg4 g5 7.Dc8+ Kg7 8.h5 Se5 9.De8 Kh8 10.Df8 
  +-  (4.65)   Tiefe: 38   00:29:16  2147mN
1...Sf7 2.Dc7 Kf8 3.Kh3 S5h6 4.Dd7 Sg8 5.g4 hxg4+ 6.fxg4 g5 7.Dc8+ Kg7 8.h5 Se5 9.De8 Kh8 
  +-  (4.67)   Tiefe: 38   00:29:54  2147mN
1...Sf7 2.Dc7 Kf8 3.Kh3 S5h6 4.Dd7 Sg8 5.g4 hxg4+ 6.fxg4 g5 7.Dc8+ Kg7 8.h5 Se5 9.De8 
  +-  (4.67)   Tiefe: 38   00:30:18  2147mN
1...Sf7 2.Dc7 Kf8 3.Kh3 S5h6 4.Dd7 Sg8 5.g4 hxg4+ 6.fxg4 g5 7.Dc8+ Kg7 8.h5 Se5 9.De8 
  +-  (4.70)   Tiefe: 38   00:30:50  2147mN

Pity I don't remember the exact eval of the Cluster at the beginning and switched to list view only later on, then it would be interesting to know of course, which tbs it was using but seeing the very low node counts I'd guess it could well be 6mob.

That all shows one thing to me once again: evals staying at a certain range for a long time computing show just one thing: engine's horizon is reached and should be widened with variants along the search tree to get more significant eval-curves, sometimes they'll get up, sometimes down, but if horizon gets really wider, one of the both will occure.
Measuring this over time and plies in forward and backward solving might be the only test for engines's strength comparison for certain positions nowadays when Celo has come to its end as for significant measurements comparing engines too near to each other in style and strength at modern levels.
By the way did I mention lately that there is no such thing as overall playing strength in reality and it's always to be measured only relatively to certain positions?