Author: Jon Dart
Date: 22:00:06 05/30/03
Arasan had this position on ICC against messchess a while ago: [D] Q7/8/3k4/8/5p1P/4p2K/4Bq2/8 w - - Here Arasan played Qf3. This is a mistake because after Qxf3 Bxf3, the position is drawn. Seeing this requires knowledge. After the exchange, White will have the "wrong" bishop and a rook Pawn, and so cannot win despite being nominally ahead in material. Arasan had access to 5-man tablebases, but those don't help here because the position after the exchange has 6 total pieces. One way to test for the presence of this knowledge is this position: [D] 8/8/3k4/8/5p1P/4pQ1K/4Bq2/8 b - - where Black should play Qxf3. Crafty plays this right away and gets extra credit for showing a draw score. A number of other amateur programs also handle this correctly: Beowulf 2.0 and Bringer, for example. Some get it wrong: Quark 1.5 won't play Qxf3. Arasan had some knowledge about the "wrong" bishop/rook pawn draw, but only in very simple cases (e.g. when the opponent had no pawns). I've recently extended the code to cover more cases recently, and it does play Qxf3 now, with a close to even score. It will also avoid Qf3 in the first position. --Jon
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.