16 testpositions from the past.....

[ml]

Maybe it recognizes that the pieces are all trapped, and reduces their values correspondingly. The next benchmark would be to evaluate the position as a draw, but there are no engines, as far as I know, that are capable of this yet.

[Ovyron]

The next benchmark would be to evaluate the position as a draw, but there are no engines, as far as I know, that are capable of this yet.

Shredder 11 is close enough:

CPU0: AuthenticAMD x86 Family 6 Model 10 Stepping 0 2133 MHz
Engine: Shredder 11 UCI (64 MB)
by Stefan Meyer-Kahlen
1/01 0:00 -0.25 1.Kc2 (1) 0
2/03 0:00 -0.25 1.Kc2 Rd8 (17) 0
3/03 0:00 -0.25 1.Kc2 Rd8 2.Kc3 (64) 0
4/05 0:00 -0.25 1.Kc2 Ra6 2.Kd3 Rd8 3.Kc3 (180) 0
5/06 0:00 -0.25 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 (343) 1
6/07 0:00 -0.25 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 (625) 2
7/09 0:00 -0.25 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 (1.504) 5
8/10 0:00 -0.25 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Kh8 (1.971) 7
9/11 0:00 -0.26 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 (2.667) 9
10/13 0:00 -0.25 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Kh8 7.Kd3 (4.641) 16
11/14 0:00 -0.26 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Ra8 7.Kf4 Ba6 (5.978) 20
12/15 0:00 -0.26 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Ra8 7.Kf4 Ba6 8.Ke3 (7.562) 25
13/17 0:00 -0.26 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Ra8 7.Kf4 Ba6 8.Ke3 Rec8 9.Kd3 (9.735) 31
14/18 0:00 -0.25 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Ra8 7.Kf4 Ba6 8.Ke3 Rec8 9.Kf4 Kh8 (15.502) 48
15/19 0:00 -0.26 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Ra8 7.Kf4 Ba6 8.Ke3 Rec8 9.Kf4 Rf8 (20.678) 62
16/21 0:00 -0.26 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Ra8 7.Kf4 Ba6 8.Ke3 Rec8 9.Kf4 Rf8 (26.489) 75
17/22 0:00 -0.26 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Ra8 7.Kf4 Ba6 8.Ke3 Rec8 9.Kf4 Rf8 (34.872) 96
18/23 0:00 -0.26 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Ra8 7.Kf4 Ba6 8.Ke3 Rec8 9.Kf4 Rf8 (53.231) 136
19/25 0:00 -0.26 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Ra8 7.Kf4 Ba6 8.Ke3 Rec8 9.Kf4 Rf8 10.Ke3 Rfe8 11.Kd3 Rec8 12.Ke3 Rf8 (70.742) 168
20/26 0:00 -0.26 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Ra8 7.Kf4 Ba6 8.Ke3 Rec8 9.Kf4 Rf8 (91.254) 198
21/27 0:00 -0.26 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Ra8 7.Kf4 Ba6 8.Ke3 Rec8 9.Kf4 Rf8 (122.995) 236
22/29 0:00 -0.26 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Ra8 7.Kf4 Ba6 8.Ke3 Rec8 9.Kf4 Rf8 (184.377) 307
23/30 0:00 -0.25 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Ra8 7.Kf4 Ba6 8.Ke3 Rec8 9.Kf4 Rf8 10.Ke3 Rfe8 11.Kd3 Rec8 12.Ke3 Rf8 (267.765) 371
24/31 0:00 -0.25 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Ra8 7.Kf4 Ba6 8.Ke3 Rec8 9.Kf4 Rf8 (370.336) 420
25/33 0:01 -0.25 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Ra8 7.Kf4 Ba6 8.Ke3 Rec8 9.Kf4 Rf8 (526.269) 469
26/34 0:01 -0.25 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Ra8 7.Kf4 Ba6 8.Ke3 Rec8 9.Kf4 Rf8 (795.128) 509
27/36 0:02 -0.25 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Ra8 7.Kf4 Ba6 8.Ke3 Rec8 9.Kf4 Rf8 (1.088.017) 540
28/37 0:02 -0.25 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Ra8 7.Kf4 Ba6 8.Ke3 Rec8 9.Kf4 Rf8 (1.538.194) 560
29/40 0:03 -0.25 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Ra8 7.Kf4 Ba6 8.Ke3 Rec8 9.Kf4 Rf8 (2.111.627) 580
30/40 0:05 -0.25 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Ra8 7.Kf4 Ba6 8.Ke3 Rec8 9.Kf4 Rf8 (3.166.180) 616
31/42 0:07 -0.25 1.Kc2 Ra6 2.Kd3 Rd8 3.Kd4 Kg8 4.Ke3 Rf8 5.Kf4 Re8 6.Ke3 Ra8 7.Kf4 Ba6 8.Ke3 Rec8 9.Kf4 Rf8 (4.374.361) 621
best move: Kc3-c2 time: 0:09.243 min n/s: 612.857 CPU 90.2% n/s(1CPU): 679.442 nodes: 5.658.514

[ml]

Now there are only a few top engines that can avoid cxb5 to force a draw
[d]5brk/8/3p1p1p/1rpPpPpP/1pP1P1P1/pP6/P1K5/8 w - - 0 0

I think that this is not correct.

I believe that many engines that are not top engines can avoid cxb5 after some minutes of search.

You're right! I found some more engines that can avoid the rook capture: Crafty 20.14 w32 and Fritz 5.32, which take only a few seconds, and Comet B68 (rated 2497 on CCRL 40/40) in about 10 minutes. So contrary to my earlier assertion, there seem to be more than a few engines, including lower-rated ones, that can solve this.

[Michael Neish]

Out of interest, how about the alternative position:
[d]5brk/8/3p1p1p/1rp1pPpP/1pPPP1P1/pP6/P1K5/8 w - - 0 0

Do engines play d5 to obtain a draw? There could be any number of variations of this position made by moving Pawns back a square to see what the programs play.

Regards,

Mike.

[ml]

Very impressive. This practically qualifies as an evaluation as a draw. I wonder how other versions of Shredder would evaluate this position.

[ml]

This position appears to be a bit more difficult than the other one. I've tested a few engines on it: Fritz 10 finds d5 immediately; Crafty 20.14 w32 takes a few seconds; and Fritz 11 finds it in just under 3 minutes. Engines that can avoid cxb5 in the other position can probably find d5 here, except that it will take longer.

[Michael Neish]

This position appears to be a bit more difficult than the other one. I've tested a few engines on it: Fritz 10 finds d5 immediately; Crafty 20.14 w32 takes a few seconds; and Fritz 11 finds it in just under 3 minutes. Engines that can avoid cxb5 in the other position can probably find d5 here, except that it will take longer.

Hi, well I'm glad this position seems harder. I sort of came up with it spontaneously without testing it.

I guess you could keep setting the position back move by move to move the draw horizon further and further away, to see how engines cope with it.

Regards,

Mike.

P.S.: By the way, if you're using bitboards wouldn't it be trivial to insert a line in the evaluation to check if

1) all eight Pawns are present,
2) they're locked, and if so
3) check which side would benefit from unlocking Pawns (if it's possible to do so)?

Part 3 would be the costly part, but only rarely called because 1 and especially 2 will be rarely satisfied.

[ml]

In this case, I don't see how the position could be moved back further while also preventing black from unlocking the pawns.

1) and 2) wouldn't use up too many CPU cycles because they only have to be processed once for each move, rather than for each node in the search. But that would exclude positions like the last one, with the pawn at d4. Fritz 10 and Shredder 11, because they see a considerable advantage in not taking the rook, might be using an algorithm similar to this, or one that can identify trapped pieces that are unable to put the king in check or capture material.

Regards,
ml

[Ovyron]

Shredder 11 UCI:

CPU0: AuthenticAMD x86 Family 6 Model 10 Stepping 0 2133 MHz
by Stefan Meyer-Kahlen
1/01 0:00 -0.27 1.d5 (3) 0
2/03 0:00 -0.28 1.d5 Kg7 (23) 0
3/03 0:00 -0.28 1.d5 Kg7 2.Kd3 (73) 0
4/05 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 (193) 0
5/06 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 (274) 0
6/07 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 (474) 1
7/09 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 (799) 2
8/10 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 (1.181) 3
9/11 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 (1.887) 5
10/13 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 Kc7 7.Ke3 (2.480) 7
11/14 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 Kc7 7.Ke3 Kb7 (4.634) 13
12/15 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 Kc7 7.Ke3 Kb7 8.Kd3 (7.673) 21
13/17 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 Kc7 7.Ke3 Kb7 8.Kd3 Ka6 9.Kd2 (11.064) 30
14/18 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 Kc7 7.Ke3 Kb7 8.Kd3 Ka6 9.Kd2 Bg7 (14.619) 38
15/20 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 Kc7 7.Ke3 Kb7 8.Kd3 Ka6 9.Kd2 Bg7 10.Ke3 (18.924) 47
16/21 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 Kc7 7.Ke3 Kb7 8.Kd3 Ka6 9.Kd2 Bg7 10.Ke3 Kb7 11.Kd3 (25.034) 61
17/22 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 Kc7 7.Ke3 Kb7 8.Kd3 Ka6 9.Kd2 Bg7 10.Ke3 Kb7 11.Kd3 Raa8 (34.278) 79
18/23 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 Kc7 7.Ke3 Kb7 8.Kd3 Ka6 9.Kd2 Bg7 10.Ke3 Kb7 11.Kd3 Raa8 12.Kd2 (44.499) 98
19/25 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 Kc7 7.Ke3 Kb7 8.Kd3 Ka6 9.Kd2 Bg7 10.Ke3 Kb7 11.Kd3 Raa8 12.Kd2 Ra6 13.Ke3 (57.377) 122
20/26 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 Kc7 7.Ke3 Kb7 8.Kd3 Ka6 9.Kd2 Bg7 10.Ke3 Kb7 11.Kd3 Raa8 12.Kd2 Ra6 13.Ke3 Raa8 (76.303) 152
21/28 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 Kc7 7.Ke3 Kb7 8.Kd3 Ka6 9.Kd2 Bg7 10.Ke3 Kb7 11.Kd3 Raa8 12.Kd2 Ra6 13.Ke3 Rga8 14.Kd2 (101.244) 184
22/29 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 Kc7 7.Ke3 Kb7 8.Kd3 Ka6 9.Kd2 Bg7 10.Ke3 Kb7 11.Kd3 Raa8 12.Kd2 Ra6 13.Ke3 Rga8 14.Kd2 (139.775) 236
23/30 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 Kc7 7.Ke3 Kb7 8.Kd3 Ka6 9.Kd2 Bg7 10.Ke3 Kb7 11.Kd3 Raa8 12.Kd2 Ra6 13.Ke3 Rga8 14.Kd2 (186.032) 281
24/31 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 Kc7 7.Ke3 Kb7 8.Kd3 Ka6 9.Kd2 Bg7 10.Ke3 Kb7 11.Kd3 Raa8 12.Kd2 Ra6 13.Ke3 Rga8 14.Kd2 (236.836) 323
25/33 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 Kc7 7.Ke3 Kb7 8.Kd3 Ka6 9.Kd2 Bg7 10.Ke3 Kb7 11.Kd3 Raa8 12.Kd2 Ra6 13.Ke3 Rga8 14.Kd2 (307.517) 374
26/34 0:00 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 Kc7 7.Ke3 Kb7 8.Kd3 Ka6 9.Kd2 Bg7 10.Ke3 Kb7 11.Kd3 Raa8 12.Kd2 Ra6 13.Ke3 Rga8 14.Kd2 (398.565) 428
27/35 0:01 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 Kc7 7.Ke3 Kb7 8.Kd3 Ka6 9.Kd2 Bg7 10.Ke3 Kb7 11.Kd3 Raa8 12.Kd2 Ra6 13.Ke3 Rga8 14.Kd2 (629.308) 523
28/37 0:01 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 Kc7 7.Ke3 Kb7 8.Kd3 Ka6 9.Kd2 Bg7 10.Ke3 Kb7 11.Kd3 Raa8 12.Kd2 Ra6 13.Ke3 Rga8 14.Kd2 (836.957) 568
29/38 0:01 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 Kc7 7.Ke3 Kb7 8.Kd3 Ka6 9.Kd2 Bg7 10.Ke3 Kb7 11.Kd3 Raa8 12.Kd2 Ra6 13.Ke3 Rga8 14.Kd2 (1.042.649) 591
30/39 0:02 -0.28 1.d5 Ra5 2.Kd3 Kg7 3.Ke3 Kf7 4.Kd3 Ke7 5.Kd2 Kd7 6.Kd3 Kc7 7.Ke3 Kb7 8.Kd3 Ka6 9.Kd2 Bg7 10.Ke3 Kb7 11.Kd3 Raa8 12.Kd2 Ra6 13.Ke3 Rga8 14.Kd2 (1.336.958) 632
best move: d4-d5 time: 0:04.026 min n/s: 758.457 CPU 82.4% n/s(1CPU): 920.457 nodes: 3.037.623