Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Waht is the biggest positional score of your program?(Crafty >37 pawns)

Author: Robert Hyatt

Date: 10:03:02 09/27/01

Go up one level in this thread


On September 26, 2001 at 16:02:26, Andrew Dados wrote:

>On September 26, 2001 at 10:20:00, Robert Hyatt wrote:
>
>>On September 25, 2001 at 07:01:52, Uri Blass wrote:
>>
>>>On September 25, 2001 at 06:58:07, Uri Blass wrote:
>>>
>>>>The champion seems to be little goliath
>>>>
>>>>It does not want to tell me information at depth 1 but it does not see a forced
>>>>mate based on it's score and even if it can see 8 queens the positional score is
>>>>more than 200 pawns.
>>>>
>>>>Goliath Light 1.5 - Blass,U
>>>>8/8/PPPPPPPP/8/8/4K3/8/4k3 w - - 0 1
>>>>
>>>>Analysis by Goliath Light 1.5:
>>>>
>>>>1.h7
>>>>  +-  (290.17)   Depth: 4/4   00:00:00
>>>>1.h7 Kf1
>>>>  +-  (292.17)   Depth: 4/4   00:00:00
>>>>1.h7 Kf1
>>>>  +-  (292.17)   Depth: 4/5   00:00:00
>>>>
>>>>(Blass, Tel-aviv 25.09.2001)
>>>>
>>>>Here is another anlaysis by goliath light
>>>>goliath light claims that it can see mate in 307 but unfortunately does not give
>>>>the full 613 plies
>>>>
>>>>Goliath Light 1.5 - Blass,U
>>>>8/7P/PPPPPPP1/8/8/4K3/8/5k2 w - - 0 1
>>>>
>>>>Analysis by Goliath Light 1.5:
>>>>
>>>>2.h8Q
>>>>  +-  (290.17)   Depth: 4/4   00:00:00
>>>>2.h8Q
>>>>  +-  (#6)   Depth: 4/4   00:00:00
>>>>2.h8Q Kg1 3.g7
>>>>  +-  (292.28)   Depth: 4/4   00:00:00
>>>>2.g7
>>>>  +-  (#307)   Depth: 4/4   00:00:00
>>>>2.g7 Ke1 3.h8Q Kd1
>>>>  +-  (#307)   Depth: 4/4   00:00:00
>>>>
>>>>(Blass, Tel-aviv 25.09.2001)
>>>>
>>>>Uri
>>>I downloaded another engine and downloaded goliath again.
>>>
>>>I give goliath to analyze again the same position
>>>It can now see mate in 313 with another move and the score improves to mate in
>>>312
>>>Goliath Light 1.5 - Blass,U
>>>[D]8/7P/PPPPPPP1/8/8/4K3/8/5k2 w - - 0 1
>>>
>>>Analysis by Goliath Light 1.5:
>>>
>>>2.h8Q
>>>  +-  (290.17)   Depth: 4/4   00:00:00
>>>2.h8Q Kg2
>>>  +-  (292.17)   Depth: 4/4   00:00:00
>>>2.d7
>>>  +-  (#313)   Depth: 4/4   00:00:00
>>>2.d7 Kg2
>>>  +-  (#313)   Depth: 4/4   00:00:00
>>>2.c7 Kg2
>>>  +-  (#312)   Depth: 4/4   00:00:00
>>>
>>>(Blass, Tel-aviv 25.09.2001)
>>>
>>>Uri
>>
>>
>>That is broken of course.  There are no such mates in the 3-4-5 piece
>>tablebases.
>
>What you see can be explained by scores overlapping 'mating score limit'.
>Let's say 292 'pawn umits' is internally very close to score which will be
>interpreted as 'mate in x' score. Hard to fit it all in 16 bits, which is
>probably the case here....
>
>-Andrew-


That is certainly possible.  In reality you can have 9 queens (81) plus
two rooks, two bishops and two knights.  Total is 103.  10300 fits in a
16 bit signed int just fine.  And should allow another 22000 in positional
scores as well...

Perhaps his "mate score" is simply too small if that is what happened...

IE I use MATE = 32768, since all mate scores will be that number - some number
of plies, which will fit just fine..



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.