Author: Amir Ban
Date: 04:13:00 06/15/98
Go up one level in this thread
On June 15, 1998 at 06:56:27, Kai Lübke wrote: >Experimenting with Junior 4.6, I found bugs in the way the program >handles the following test positions: > >Albillo 20: Ng4? with immediate "draw" though it's a mate in 9 (false >KNNKP endgame knowledge) >[I don't understand why Junior doesn't even *try* to search the position >here.] > >Albillo 27: Be4 with immediate "mate in 2" though it's mate in 9 >(doesn't see promotion to knight which >prolongs mate) > >Killer 01 (8/5P1k/7p/6pP/5pP1/5PpK/6P1/8 w - - ): plays f8Q?? which is a >stalemate. Does not see >f8N! (for example) which easily wins for white > >=> doesn't even consider underpromotion, just as if it didn't know it at >all! > >--- >Shep These are not bugs, they are features :) More exactly they are conscious choices I make. The "insufficient material" and underpromotion behavior of Junior were already discussed here and elsewhere. In Junior 5.0, the user will be able to configure the engine in several ways, and one of the options will be to disable the insufficient-material heuristic. Amir
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.