Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Go Cake, go!! Extra! Extra! EGTB error claims one victim.

Author: martin fierz

Date: 18:38:57 08/10/02

Go up one level in this thread


On August 10, 2002 at 13:13:56, Uri Blass wrote:

>On August 10, 2002 at 12:30:26, martin fierz wrote:
>
>>On August 09, 2002 at 06:11:01, David Dory wrote:
>>
>>>On August 08, 2002 at 16:43:55, martin fierz wrote:
>>>
>>>>i think schaeffer wrote something about how unlikely it was that an EGTB error
>>>>would have turned up in his match against tinsley (they had bugs in 1992, but
>>>>they thought they had none in 1996 - but they had some there, too). well, it
>>>>seems it turned up here! we're not 100% certain yet, there's another guy in
>>>>england using schaeffers database, if he also gets that error with his lookup
>>>>code it's definitely the chinook database.
>>>>
>>>>aloha
>>>>  martin
>>>
>>>Those damn EGTB's have been hexed for Jonathan. He mentions in his book "One
>>>Jump Ahead", how they found errors in them, again and again. Much to team
>>>Chinook's dismay.
>>>
>>>I thought after the last EGTB errors that every one of the files had been
>>>"proofed", 100% accurate. Very interesting, if the error is in Chinook's files.
>>
>>schaeffer sent me an email saying the EGTB was correct. which leaves a corrupt
>>file on ed's harddisk as one possible cause, or an error in his access code as
>>the other possibility for what happened. we don't know yet and until he has a
>>chance of checking this back home, i can't tell you what really happened.
>>
>>
>>>After running thru a game on your web site, I'm POSITIVE I won't be playing
>>>checkers with you guys! Wow! Your programs are GOOD!
>>>
>>>Great web site, and best wishes for a great tournament, Martin.
>>>
>>>Dave
>>
>>thanks :-)
>>
>>aloha
>>  martin
>
>I suspect that this checker 8*8 game is practically solved and programs only
>need hardware that is slightly faster to get 100% draws instead of more than 90%
>draws.

i think you suspect wrong. the reason for the high number of draws is that we
have close-to-perfect opening books, which take our programs until positions
where they will play perfectly thanks to the 8-piece EGTB. without those books,
there would have been many more decisive games - even on much faster hardware.

>I am going to say the same about chess if I see more than 90% draws but it does
>not seem to happen to chess in the near future.
i agree, since in chess there is no equivalent to the 8-piece db for checkers.

>I wonder if there are positions from practical checkers games when programs
>cannot find the right move after an hour of search.
lots of them! i will check how long it takes for my program to avoid the two
losses it had here, but these positions were so deep, i do not think it will
find the right move with days of search time. my opponents had those two games
in book...

aloha
  martin
>Uri



This page took 0.01 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.