Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Are there programs which can win this without tablebases?

Author: Günther Simon

Date: 09:03:58 10/22/05

Go up one level in this thread


On October 22, 2005 at 11:35:31, Dieter Buerssner wrote:

>On October 22, 2005 at 11:05:21, Günther Simon wrote:
>
>>On October 22, 2005 at 09:34:42, Uri Blass wrote:
>>
>>>On October 22, 2005 at 08:08:47, Günther Simon wrote:
>>>
>>>>This position really happened in one of my RWBC games today
>>>>and it revealed a bug in a program. 6 men say it is a win.
>>>>(Shredder Oracle : M27 with best play)
>>>>I am not sure though, if there are programs, which are able
>>>>to win this without tablebases at all? (in reasonable time ;)
>>>>
>>>>Guenther
>>>>
>>>>[D]4r3/8/8/8/8/6bk/7p/3R3K b - - 0 76
>>>
>>>I doubt if there are programs that can find it without tablebases at reasonable
>>>time because I expect KRB vs KR to be prune based on no threat but
>>>Yace with 5 piece tablebases agree with shredder  Oracle
>>>
>>>New game, 4320'/1
>>>4r3/8/8/8/8/6bk/7p/3R3K b - - 0 1
>>>
>>>Analysis by Yace 0.99.87:
>>>
>>...
>>
>>>1...Bf4 2.Rf1 Kg4 3.Rb1 Rc8 4.Rf1 Be3 5.Kxh2 Rd8 6.Kg2 Rd2+ 7.Kh1 Bf2 8.Kg2 Bg3+
>>>9.Kg1 Re2 10.Rd1 Bf4 11.Rd3 Be3+ 12.Kf1 Kf3 13.Rd7 Rh2 14.Rf7+ Bf4 15.Rg7 Rc2
>>>16.Re7 Rb2 17.Re8
>>>  -+  (-#27)   Depth: 22/51   00:29:39  2527052kN
>>>
>>>(,  22.10.2005)
>>>
>>>Uri
>>
>>Thanks for the cpu time Uri. I have one question though, did you
>>analyse this with 5 men + Yaces' bitbases or without bitbases?
>>(Those would probably speed it up a 'bit')
>
>
>Günther, I didn't try but I doubt that bitbases can help much here. Yace does
>not use KRBKR bitbase. 4-men bitbases cannot help at all, to see a huge score
>here - they can only help to prune some branches faster. Also note that the
>engine was not slowed down very much by TB-accesses. Yace found a significant
>score rather fast:
>
>[from Uri's post]
>1...Bf4 2.Rf1
>  -+  (-13.23)   Depth: 22/44   00:00:49  41473kN
>
>[over 800 knodes/s]
>
>Only showing a mate score needed very long. I guess, this is not easy to avoid,
>neither is it important for a game. Bitbases will typically not help, to show a
>mate score. Even when Yace had KRBKR bitbase, a winning position might be scored
>+15 (could show +100 or mate in 500 or something, but that is another thing).
>This would not be enough to show the real mate. It could help to find a really
>significant score faster.
>
>Regards,
>Dieter

Hi Dieter,

thanks for the extra infos. I think we have to agree though, what
a significant score means here?
I mean every engine will show here a huge plus, but nevertheless
it will not be capable of winning the game without tablebases,
or may be even not with tablebases below 6 men?
What score could be called as likely to win the game, when really
_playing_ this position? Will > +10 be sufficient for most cases?

Example with Amateur 2.86 @WB 4.27x, 128MB hash:
(had only my P3 650 cpu available)
+ 7.62, but quite stuck with the score and PV until d 26

16223 >first : setboard 4r3/8/8/8/8/6bk/7p/3R3K b - - 0 1
18927 >first : time 720000
otim 720000
18927 >first : go
20279 <first : 6 762 1 1497 Re2 Rc1 Rd2 Re1 Bf4
20299 <first : 7 763 3 3902 Re2 Rc1 Rd2 Re1 Rd3 Rf1 Rd2
20329 <first : 8 763 6 7231 Re2 Rc1 Rd2 Re1 Kg4 Rf1 Kh3
20429 <first : 9 761 17 21752 Re2 Rc1 Rd2 Rf1 Rc2 Re1 Rf2
20529 <first : 10 762 27 35102 Re2 Rc1 Rd2 Rf1 Bf2
20709 <first : 11 762 44 56377 Re2 Rc1 Rd2 Rf1 Rc2 Rd1 Rg2
20980 <first : 12 762 71 96680 Re2 Rc1 Rd2 Rf1 Rc2 Rd1 Rg2
21340 <first : 13 762 107 154460 Re2 Rc1 Rd2 Rf1 Rc2 Rd1 Rg2
21911 <first : 14 762 164 245651 Re2 Rc1 Rd2 Rf1 Rc2 Rd1 Rg2
22492 <first : 15 762 222 335421 Re2 Rc1 Rd2 Rf1 Rc2 Rd1 Rg2
23153 <first : 16 762 287 444767 Re2 Rc1 Rd2 Rf1 Rc2 Rd1 Rg2
24295 <first : 17 762 402 630626 Re2 Rc1 Rd2 Rf1 Rc2 Rd1 Rg2
25877 <first : 18 762 561 904945 Re2 Rc1 Rd2 Rf1 Rc2 Rd1 Rg2
28370 <first : 19 762 810 1323623 Re2 Rc1 Rd2 Rf1 Rc2 Rd1 Rg2
31996 <first : 20 762 1173 1956109 Re2 Rc1 Rd2 Rf1 Rc2 Rd1 Rg2
37614 <first : 21 762 1735 2953422 Re2 Rc1 Rd2 Rf1 Rc2 Rd1 Rg2
46787 <first : 22 762 2652 4628658 Re2 Rc1 Rd2 Rf1 Rc2 Rd1 Rg2
63952 <first : 23 762 4369 7291751 Re2 Rc1 Rd2 Rf1 Rc2 Rd1 Rg2
82138 <first : 24 762 6188 10221718 Re2 Rc1 Rd2 Rf1 Rc2 Rd1 Rg2
116758 <first : 25 762 9649 16349063 Re2 Rc1 Rd2 Rf1 Rc2 Rd1 Rg2
199997 <first : 26 762 17972 29900001 Re2 Rc1 Rd2 Rf1 Rc2 Rd1 Rg2

Regards,
Guenther



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.