Author: Roy Eassa
Date: 07:14:33 06/03/02
Go up one level in this thread
On June 01, 2002 at 16:22:57, Christophe Theron wrote: >On June 01, 2002 at 11:16:44, Robert Hyatt wrote: > >>On June 01, 2002 at 06:23:34, Gian-Carlo Pascutto wrote: >> >>>On June 01, 2002 at 00:12:53, Robert Hyatt wrote: >>> >>>>Give crafty 1 second per move. It solves it easily. I discovered this >>>>a few years ago when talking about the 4 piece files. This is hard for >>>>humans. It is _very_ easy for computers. Far easier than I would have >>>>thought possible... >>> >>>Hmm, I just tested and mine can't do it at 1 sec per move, >>>so I guess you do need some kind of 'mopping up' code. >>> >>>-- >>>GCP >> >> >>Yes. All you really need is code to (a) drive the losing king to the >>edge of the board and (b) get the winning king close to the losing king >>to threaten mate. The search will see how to prevent the rook checks and >>eventually by threatening mate it will win the rook... > > >I think Tiger has both heuristics and that's why it manages to win, but that's >only by chance. > >Actually these heuristics are always activated in the late endgame as I noticed >it was useful so often. > >I did not know they were also working in KQKR and that nothing else was needed. > > From what I've seen, Chess Tiger plays a very strong endgame in the absense of tablebases. I wonder: has anybody devised a Nunn-like set of positions to test chess programs' endgame abilities? I could imagine a set of endgames where programs (which would not be allowed to use tablebases) would face off against each other twice -- once as Black and once as White. A tournament of such positions could help determine which programs REALLY are best in the endgame in the absense of TBs.
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.