Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Chess program improvement project (copy at Winboard::Programming)

Author: Robert Hyatt

Date: 19:05:29 03/08/06

Go up one level in this thread


On March 08, 2006 at 16:29:31, Stuart Cracraft wrote:

>On March 07, 2006 at 21:51:03, Robert Hyatt wrote:
>
>>On March 07, 2006 at 19:19:05, Stuart Cracraft wrote:
>>
>>>On March 07, 2006 at 15:49:27, Robert Hyatt wrote:
>>>
>>>>On March 07, 2006 at 02:27:24, Tony Werten wrote:
>>>>
>>>>>On March 07, 2006 at 00:35:43, Stuart Cracraft wrote:
>>>>>
>>>>>>On March 07, 2006 at 00:25:46, Stuart Cracraft wrote:
>>>>>>
>>>>>>>On March 06, 2006 at 22:50:44, Dann Corbit wrote:
>>>>>>>
>>>>>>>>Looks like king safety to me:
>>>>>>>>[D] r4rk1/1p2ppbp/p2pbnp1/q7/3BPPP1/2N2B2/PPP4P/R2Q1RK1 b - - bm Bxg4
>>>>>>>
>>>>>>>An extended search could not crack the 10th ply aspiration search fail high
>>>>>>>on this position. I do not know what was about to happen, only that it
>>>>>>>would take a long time.
>>>>>>>
>>>>>>>.sd 99
>>>>>>>maxdepth = 99 maxtime = 9999999 timed = 0
>>>>>>>.ts
>>>>>>>position file? [wac.epd] retry.epd
>>>>>>># of test positions to test? 1
>>>>>>>maxtime = 9999999
>>>>>>>Interrupt current ply and return move at timeout
>>>>>>>Testsuite: Tests/retry.epd 1 positions
>>>>>>>*** Problem   Solution(s): Bxg4 (bm)
>>>>>>>[D] r4rk1/1p2ppbp/p2pbnp1/q7/3BPPP1/2N2B2/PPP4P/R2Q1RK1 b - - bm Bxg4
>>>>>>>*** Problem   Solution(s): Bxg4 (bm)
>>>>>>>BR ** -- ** -- BR BK **
>>>>>>>** BP ** -- BP BP BB BP
>>>>>>>BP ** -- BP BB BN BP **
>>>>>>>BQ -- ** -- ** -- ** --
>>>>>>>-- ** -- WB WP WP WP **
>>>>>>>** -- WN -- ** WB ** --
>>>>>>>WP WP WP ** -- ** -- WP
>>>>>>>WR -- ** WQ ** WR WK --
>>>>>>>mv 1 stage 0, black to move, computer plays black
>>>>>>>hash=5076cc275e393397
>>>>>>>pawnhash=3d032756da421438
>>>>>>>0 0 0 0 0 0
>>>>>>>Alpha=999 Beta=1800 Maxdepth=99 MaxTime=9999999 xboard=0
>>>>>>>Itr/Max Mv   Score   Time Nodes PV
>>>>>>> 1/ 5  f6g4 1400  0.00       57 f6g4
>>>>>>> 2/10  f6g4 1347  0.01      385 f6g4 f3e2 g7d4 d1d4
>>>>>>> 3/10< f6g4 946  0.03     1959 f6g4 f3e2 g7d4 d1d4 f7f6
>>>>>>> 3/10  f6g4 945  0.05     2925 f6g4 d4g7 g8g7 f3g4
>>>>>>> 4/12< f6g4 544  0.10     6880 f6g4 d4g7 g8g7 f3g4 f7f6
>>>>>>> 4/15  a5d8 -241  0.25    17921 a5d8 g1g2
>>>>>>> 5/21  e6c4 -35  0.46    33988 e6c4 f3e2 a5c7
>>>>>>> 6/21  h7h5 -40  1.30   101881 h7h5 g4g5 f6d7 d4g7 g8g7
>>>>>>> 7/21  a5b4 -104  2.96   235870 a5b4 g4g5 f6h5 d4g7 h5g7 a1b1 a8c8
>>>>>>> 8/22  a5b4 -145  6.09   498480 a5b4 a2a3 b4b2 c3d5 b2b5 d5e7 g8h8 a1b1
>>>>>>> 9/24  e6c4 -79 11.79   991796 e6c4 f1f2 e7e5 d4e3 a5d8 a2a4
>>>>>>>10/24> e6c4 321 30.18  2576527 e6c4 f1f2 e7e5 d4e3 a5b4 g4g5 f6d7 f4e5 g7e5
>>>>>>>10/
>>>>>>
>>>>>>The search kept going and did find the move but it took almost 15 minutes.
>>>>>>
>>>>>>position file? [wac.epd] retry.epd
>>>>>># of test positions to test? 1
>>>>>>maxtime = 9999999
>>>>>>Interrupt current ply and return move at timeout
>>>>>>Testsuite: Tests/retry.epd 1 positions
>>>>>>*** Problem   Solution(s): Bxg4 (bm)
>>>>>>[D] r4rk1/1p2ppbp/p2pbnp1/q7/3BPPP1/2N2B2/PPP4P/R2Q1RK1 b - - bm Bxg4
>>>>>>*** Problem   Solution(s): Bxg4 (bm)
>>>>>>BR ** -- ** -- BR BK **
>>>>>>** BP ** -- BP BP BB BP
>>>>>>BP ** -- BP BB BN BP **
>>>>>>BQ -- ** -- ** -- ** --
>>>>>>-- ** -- WB WP WP WP **
>>>>>>** -- WN -- ** WB ** --
>>>>>>WP WP WP ** -- ** -- WP
>>>>>>WR -- ** WQ ** WR WK --
>>>>>>mv 1 stage 0, black to move, computer plays black
>>>>>>hash=5076cc275e393397
>>>>>>pawnhash=3d032756da421438
>>>>>>0 0 0 0 0 0
>>>>>>Alpha=999 Beta=1800 Maxdepth=99 MaxTime=9999999 xboard=0
>>>>>>Itr/Max Mv   Score   Time Nodes PV
>>>>>> 1/ 5  f6g4 1400  0.00       57 f6g4
>>>>>> 2/10  f6g4 1347  0.01      385 f6g4 f3e2 g7d4 d1d4
>>>>>> 3/10< f6g4 946  0.03     1959 f6g4 f3e2 g7d4 d1d4 f7f6
>>>>>> 3/10  f6g4 945  0.05     2925 f6g4 d4g7 g8g7 f3g4
>>>>>> 4/12< f6g4 544  0.10     6880 f6g4 d4g7 g8g7 f3g4 f7f6
>>>>>> 4/15  a5d8 -241  0.25    17921 a5d8 g1g2
>>>>>> 5/21  e6c4 -35  0.46    33988 e6c4 f3e2 a5c7
>>>>>> 6/21  h7h5 -40  1.30   101881 h7h5 g4g5 f6d7 d4g7 g8g7
>>>>>> 7/21  a5b4 -104  2.96   235870 a5b4 g4g5 f6h5 d4g7 h5g7 a1b1 a8c8
>>>>>> 8/22  a5b4 -145  6.09   498480 a5b4 a2a3 b4b2 c3d5 b2b5 d5e7 g8h8 a1b1
>>>>>> 9/24  e6c4 -79 11.79   991796 e6c4 f1f2 e7e5 d4e3 a5d8 a2a4
>>>>>>10/24> e6c4 321 30.18  2576527 e6c4 f1f2 e7e5 d4e3 a5b4 g4g5 f6d7 f4e5 g7e5
>>>>>>10/25  e6g4 1344 863.20  4198040 e6g4 e4e5 d6e5 f4e5 g4f3 d1f3 a5b4 f3e3 f6g4
>>>>>>11/26  e6g4 1032 889.10  6506793 e6g4 e4e5 d6e5 f4e5 g4f3 d1f3 a5b4 a1d1 f6h5
>>>>>>b2b3
>>>>>>12/30  e6g4 1261 957.00 12603674 e6g4 f3g4 f6g4 c3d5 g7d4 d1d4 a8e8 h2h3 g4h6
>>>>>>b2b4 a5a3
>>>>>>13/32  e6g4 1285 1110.45 26675821 e6g4 f3g4 f6g4 c3d5 g7d4 d1d4 a8e8 h2h3 g4f6
>>>>>>d5f6 e7f6 d4d6 e8e4 b
>>>>>>2b4 f8d8
>>>>>>14/
>>>>>
>>>>>My guess is something is wrong with your move ordering. The failhigh on ply 10
>>>>>took 30 times the previous. That should not happen.
>>>>
>>>>There's more to it.  Look at the node counts.  A factor of less than 2X more
>>>>nodes, which suggests that the machine is doing other things which makes time
>>>>interpretation meaningless on such tests....
>>>>
>>>>
>>>
>>>This last paragraph went completely over my head. I see the node counts
>>>not going up 2x on between iteration. But "makes time interpretation
>>>meaningless" -- can you elaborate?
>>
>>
>>Time to fail high was 30x, but node count was only 2x.   Where did the extra 15X
>>time go if not searching more nodes?
>>
>>Something is fishy, most likely the cpu was not 100% busy playing chess, but was
>>running something else also...
>
>
>Very likely - the box is not dedicated.
>
>Thanks.


You can't test on a box like that and draw _any_ conclusions about the time
taken to find something.



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.