Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Crafty Tablebase Probes

Author: Robert Hyatt

Date: 13:06:56 08/27/99

Go up one level in this thread


On August 27, 1999 at 15:24:50, blass uri wrote:

>On August 27, 1999 at 11:09:59, Robert Hyatt wrote:
>
>>On August 27, 1999 at 01:54:24, blass uri wrote:
>>
>>>
>>>On August 27, 1999 at 00:06:25, Robert Hyatt wrote:
>>>
>>>>On August 26, 1999 at 20:21:04, Mark Rawlings wrote:
>>>>
>>>>>I _finally_ got all of the 5-piece tablebases to use with Crafty and Winboard!
>>>>>These things are great for analyzing endgames.  Still waiting for the 6-piece
>>>>>files to be posted :)
>>>>>
>>>>>Anyway, what is a "successful" tablebase probe?  I thought when you have the
>>>>>whole set, all probes were successful?  Thanks,
>>>>>
>>>>>Mark
>>>>
>>>>
>>>>Note that you don't have the "whole set" because we didn't build the 4 vs 1
>>>>piece files.  You can, but they take up a whole lot more space, and I don't
>>>>think crafty needs those to win the positions.. :)
>>>
>>>There are some 3 vs 2 tablebases that crafty does not need to win the position.
>>>
>>>For example KQRvs KP,KQR vs KN
>>>
>>>Uri
>>
>>
>>There are many that it doesn't need... (KQ vs KR is another) however without all
>>of them, the search definitely gets confused.  And it is much better to _know_
>>when you are exchanging into a won ending, rather than into an ending you think
>>you can win but can't...
>>
>>We will have to do the 4 vs 1 files before we get to the 3 vs 3 files, which
>>are not too far into the future now.  So one day we will have all the 5's and
>>a bunch of the 6's...
>
>I do not think that you will have to do all the 4 vs 1 files.
>I think that using simple rules to decide if the position is drawn or not drawn
>is faster than getting to tablebases in most of the cases.
>
>For example I think that KQRP vs K is always winning for the side with the
>advantage if the position is not stalemate.
>
>KQRP vs K tablebases can help you to know the exact distance to mate but are not
>important for other practical purposes.
>
>Uri


Right.. but notice I was talking about _building_ the 6 piece files... and there
we need _all_ 5 piece files to compute distance-to-mate properly...



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.