Author: Dann Corbit
Date: 09:45:03 02/11/99
Go up one level in this thread
On February 11, 1999 at 12:05:27, Jon Dart wrote: >I for one would be quite interested in an endgame test suite >that had verified solutions. The existing published ones (e.g. >ECE) are full of errors. I have culled some tests out of >Fine's Basic Chess Endings, but that has quite a few errors, too. > >(To be interesting the tests should probably be out of the range >that can be solved directly with existing tablebases). > >B.t.w. you might take a look at the tests I have published on >my web page (ftp://shell3.ba.best.com/pub/jdart/games.epd). >Some of these are hard enough to be interesting. For example, >here is one position encountered in a game my program >had against GnuChess: > >6k1/p5bp/8/1N4p1/P7/B4pP1/2b1rP1P/5RK1 w - - bm Bc5; id "test.84"; I had all of those up through test 77. I have added 78 through 84 to my test set. (Sets me back one day, darn it!). Once all of the analysis is complete, I will make it publicly available. I have a notion about test sets. Lots of times, there is a stated answer (bm)which is wrong. It is a sort of surface move you would get by looking 6 moves ahead, but which is refuted at 8 or 9 fullmoves. I think that sort of information could be very interesting to include in the tests. That way, we can get an even better idea about how deeply a program is thinking.
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.